1 /*- 2 * Copyright (c) 2014, Bryan Venteicher <bryanv@FreeBSD.org> 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice unmodified, this list of conditions, and the following 10 * disclaimer. 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 16 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 17 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 18 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, 19 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 20 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 21 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 22 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 23 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 24 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 25 */ 26 27 #include "opt_inet.h" 28 #include "opt_inet6.h" 29 30 #include <sys/cdefs.h> 31 __FBSDID("$FreeBSD$"); 32 33 #include <sys/param.h> 34 #include <sys/eventhandler.h> 35 #include <sys/kernel.h> 36 #include <sys/lock.h> 37 #include <sys/hash.h> 38 #include <sys/malloc.h> 39 #include <sys/mbuf.h> 40 #include <sys/module.h> 41 #include <sys/refcount.h> 42 #include <sys/rmlock.h> 43 #include <sys/priv.h> 44 #include <sys/proc.h> 45 #include <sys/queue.h> 46 #include <sys/sbuf.h> 47 #include <sys/socket.h> 48 #include <sys/socketvar.h> 49 #include <sys/sockio.h> 50 #include <sys/sysctl.h> 51 #include <sys/systm.h> 52 53 #include <net/bpf.h> 54 #include <net/ethernet.h> 55 #include <net/if.h> 56 #include <net/if_var.h> 57 #include <net/if_clone.h> 58 #include <net/if_dl.h> 59 #include <net/if_media.h> 60 #include <net/if_types.h> 61 #include <net/if_vxlan.h> 62 #include <net/netisr.h> 63 64 #include <netinet/in.h> 65 #include <netinet/in_systm.h> 66 #include <netinet/in_var.h> 67 #include <netinet/in_pcb.h> 68 #include <netinet/ip.h> 69 #include <netinet/ip6.h> 70 #include <netinet/ip_var.h> 71 #include <netinet/udp.h> 72 #include <netinet/udp_var.h> 73 74 #include <netinet6/ip6_var.h> 75 #include <netinet6/scope6_var.h> 76 77 struct vxlan_softc; 78 LIST_HEAD(vxlan_softc_head, vxlan_softc); 79 80 struct vxlan_socket_mc_info { 81 union vxlan_sockaddr vxlsomc_saddr; 82 union vxlan_sockaddr vxlsomc_gaddr; 83 int vxlsomc_ifidx; 84 int vxlsomc_users; 85 }; 86 87 #define VXLAN_SO_MC_MAX_GROUPS 32 88 89 #define VXLAN_SO_VNI_HASH_SHIFT 6 90 #define VXLAN_SO_VNI_HASH_SIZE (1 << VXLAN_SO_VNI_HASH_SHIFT) 91 #define VXLAN_SO_VNI_HASH(_vni) ((_vni) % VXLAN_SO_VNI_HASH_SIZE) 92 93 struct vxlan_socket { 94 struct socket *vxlso_sock; 95 struct rmlock vxlso_lock; 96 u_int vxlso_refcnt; 97 union vxlan_sockaddr vxlso_laddr; 98 LIST_ENTRY(vxlan_socket) vxlso_entry; 99 struct vxlan_softc_head vxlso_vni_hash[VXLAN_SO_VNI_HASH_SIZE]; 100 struct vxlan_socket_mc_info vxlso_mc[VXLAN_SO_MC_MAX_GROUPS]; 101 }; 102 103 #define VXLAN_SO_RLOCK(_vso, _p) rm_rlock(&(_vso)->vxlso_lock, (_p)) 104 #define VXLAN_SO_RUNLOCK(_vso, _p) rm_runlock(&(_vso)->vxlso_lock, (_p)) 105 #define VXLAN_SO_WLOCK(_vso) rm_wlock(&(_vso)->vxlso_lock) 106 #define VXLAN_SO_WUNLOCK(_vso) rm_wunlock(&(_vso)->vxlso_lock) 107 #define VXLAN_SO_LOCK_ASSERT(_vso) \ 108 rm_assert(&(_vso)->vxlso_lock, RA_LOCKED) 109 #define VXLAN_SO_LOCK_WASSERT(_vso) \ 110 rm_assert(&(_vso)->vxlso_lock, RA_WLOCKED) 111 112 #define VXLAN_SO_ACQUIRE(_vso) refcount_acquire(&(_vso)->vxlso_refcnt) 113 #define VXLAN_SO_RELEASE(_vso) refcount_release(&(_vso)->vxlso_refcnt) 114 115 struct vxlan_ftable_entry { 116 LIST_ENTRY(vxlan_ftable_entry) vxlfe_hash; 117 uint16_t vxlfe_flags; 118 uint8_t vxlfe_mac[ETHER_ADDR_LEN]; 119 union vxlan_sockaddr vxlfe_raddr; 120 time_t vxlfe_expire; 121 }; 122 123 #define VXLAN_FE_FLAG_DYNAMIC 0x01 124 #define VXLAN_FE_FLAG_STATIC 0x02 125 126 #define VXLAN_FE_IS_DYNAMIC(_fe) \ 127 ((_fe)->vxlfe_flags & VXLAN_FE_FLAG_DYNAMIC) 128 129 #define VXLAN_SC_FTABLE_SHIFT 9 130 #define VXLAN_SC_FTABLE_SIZE (1 << VXLAN_SC_FTABLE_SHIFT) 131 #define VXLAN_SC_FTABLE_MASK (VXLAN_SC_FTABLE_SIZE - 1) 132 #define VXLAN_SC_FTABLE_HASH(_sc, _mac) \ 133 (vxlan_mac_hash(_sc, _mac) % VXLAN_SC_FTABLE_SIZE) 134 135 LIST_HEAD(vxlan_ftable_head, vxlan_ftable_entry); 136 137 struct vxlan_statistics { 138 uint32_t ftable_nospace; 139 uint32_t ftable_lock_upgrade_failed; 140 }; 141 142 struct vxlan_softc { 143 struct ifnet *vxl_ifp; 144 struct vxlan_socket *vxl_sock; 145 uint32_t vxl_vni; 146 union vxlan_sockaddr vxl_src_addr; 147 union vxlan_sockaddr vxl_dst_addr; 148 uint32_t vxl_flags; 149 #define VXLAN_FLAG_INIT 0x0001 150 #define VXLAN_FLAG_TEARDOWN 0x0002 151 #define VXLAN_FLAG_LEARN 0x0004 152 153 uint32_t vxl_port_hash_key; 154 uint16_t vxl_min_port; 155 uint16_t vxl_max_port; 156 uint8_t vxl_ttl; 157 158 /* Lookup table from MAC address to forwarding entry. */ 159 uint32_t vxl_ftable_cnt; 160 uint32_t vxl_ftable_max; 161 uint32_t vxl_ftable_timeout; 162 uint32_t vxl_ftable_hash_key; 163 struct vxlan_ftable_head *vxl_ftable; 164 165 /* Derived from vxl_dst_addr. */ 166 struct vxlan_ftable_entry vxl_default_fe; 167 168 struct ip_moptions *vxl_im4o; 169 struct ip6_moptions *vxl_im6o; 170 171 struct rmlock vxl_lock; 172 volatile u_int vxl_refcnt; 173 174 int vxl_unit; 175 int vxl_vso_mc_index; 176 struct vxlan_statistics vxl_stats; 177 struct sysctl_oid *vxl_sysctl_node; 178 struct sysctl_ctx_list vxl_sysctl_ctx; 179 struct callout vxl_callout; 180 uint8_t vxl_hwaddr[ETHER_ADDR_LEN]; 181 int vxl_mc_ifindex; 182 struct ifnet *vxl_mc_ifp; 183 struct ifmedia vxl_media; 184 char vxl_mc_ifname[IFNAMSIZ]; 185 LIST_ENTRY(vxlan_softc) vxl_entry; 186 LIST_ENTRY(vxlan_softc) vxl_ifdetach_list; 187 }; 188 189 #define VXLAN_RLOCK(_sc, _p) rm_rlock(&(_sc)->vxl_lock, (_p)) 190 #define VXLAN_RUNLOCK(_sc, _p) rm_runlock(&(_sc)->vxl_lock, (_p)) 191 #define VXLAN_WLOCK(_sc) rm_wlock(&(_sc)->vxl_lock) 192 #define VXLAN_WUNLOCK(_sc) rm_wunlock(&(_sc)->vxl_lock) 193 #define VXLAN_LOCK_WOWNED(_sc) rm_wowned(&(_sc)->vxl_lock) 194 #define VXLAN_LOCK_ASSERT(_sc) rm_assert(&(_sc)->vxl_lock, RA_LOCKED) 195 #define VXLAN_LOCK_WASSERT(_sc) rm_assert(&(_sc)->vxl_lock, RA_WLOCKED) 196 #define VXLAN_UNLOCK(_sc, _p) do { \ 197 if (VXLAN_LOCK_WOWNED(_sc)) \ 198 VXLAN_WUNLOCK(_sc); \ 199 else \ 200 VXLAN_RUNLOCK(_sc, _p); \ 201 } while (0) 202 203 #define VXLAN_ACQUIRE(_sc) refcount_acquire(&(_sc)->vxl_refcnt) 204 #define VXLAN_RELEASE(_sc) refcount_release(&(_sc)->vxl_refcnt) 205 206 #define satoconstsin(sa) ((const struct sockaddr_in *)(sa)) 207 #define satoconstsin6(sa) ((const struct sockaddr_in6 *)(sa)) 208 209 struct vxlanudphdr { 210 struct udphdr vxlh_udp; 211 struct vxlan_header vxlh_hdr; 212 } __packed; 213 214 static int vxlan_ftable_addr_cmp(const uint8_t *, const uint8_t *); 215 static void vxlan_ftable_init(struct vxlan_softc *); 216 static void vxlan_ftable_fini(struct vxlan_softc *); 217 static void vxlan_ftable_flush(struct vxlan_softc *, int); 218 static void vxlan_ftable_expire(struct vxlan_softc *); 219 static int vxlan_ftable_update_locked(struct vxlan_softc *, 220 const union vxlan_sockaddr *, const uint8_t *, 221 struct rm_priotracker *); 222 static int vxlan_ftable_learn(struct vxlan_softc *, 223 const struct sockaddr *, const uint8_t *); 224 static int vxlan_ftable_sysctl_dump(SYSCTL_HANDLER_ARGS); 225 226 static struct vxlan_ftable_entry * 227 vxlan_ftable_entry_alloc(void); 228 static void vxlan_ftable_entry_free(struct vxlan_ftable_entry *); 229 static void vxlan_ftable_entry_init(struct vxlan_softc *, 230 struct vxlan_ftable_entry *, const uint8_t *, 231 const struct sockaddr *, uint32_t); 232 static void vxlan_ftable_entry_destroy(struct vxlan_softc *, 233 struct vxlan_ftable_entry *); 234 static int vxlan_ftable_entry_insert(struct vxlan_softc *, 235 struct vxlan_ftable_entry *); 236 static struct vxlan_ftable_entry * 237 vxlan_ftable_entry_lookup(struct vxlan_softc *, 238 const uint8_t *); 239 static void vxlan_ftable_entry_dump(struct vxlan_ftable_entry *, 240 struct sbuf *); 241 242 static struct vxlan_socket * 243 vxlan_socket_alloc(const union vxlan_sockaddr *); 244 static void vxlan_socket_destroy(struct vxlan_socket *); 245 static void vxlan_socket_release(struct vxlan_socket *); 246 static struct vxlan_socket * 247 vxlan_socket_lookup(union vxlan_sockaddr *vxlsa); 248 static void vxlan_socket_insert(struct vxlan_socket *); 249 static int vxlan_socket_init(struct vxlan_socket *, struct ifnet *); 250 static int vxlan_socket_bind(struct vxlan_socket *, struct ifnet *); 251 static int vxlan_socket_create(struct ifnet *, int, 252 const union vxlan_sockaddr *, struct vxlan_socket **); 253 static void vxlan_socket_ifdetach(struct vxlan_socket *, 254 struct ifnet *, struct vxlan_softc_head *); 255 256 static struct vxlan_socket * 257 vxlan_socket_mc_lookup(const union vxlan_sockaddr *); 258 static int vxlan_sockaddr_mc_info_match( 259 const struct vxlan_socket_mc_info *, 260 const union vxlan_sockaddr *, 261 const union vxlan_sockaddr *, int); 262 static int vxlan_socket_mc_join_group(struct vxlan_socket *, 263 const union vxlan_sockaddr *, const union vxlan_sockaddr *, 264 int *, union vxlan_sockaddr *); 265 static int vxlan_socket_mc_leave_group(struct vxlan_socket *, 266 const union vxlan_sockaddr *, 267 const union vxlan_sockaddr *, int); 268 static int vxlan_socket_mc_add_group(struct vxlan_socket *, 269 const union vxlan_sockaddr *, const union vxlan_sockaddr *, 270 int, int *); 271 static void vxlan_socket_mc_release_group_by_idx(struct vxlan_socket *, 272 int); 273 274 static struct vxlan_softc * 275 vxlan_socket_lookup_softc_locked(struct vxlan_socket *, 276 uint32_t); 277 static struct vxlan_softc * 278 vxlan_socket_lookup_softc(struct vxlan_socket *, uint32_t); 279 static int vxlan_socket_insert_softc(struct vxlan_socket *, 280 struct vxlan_softc *); 281 static void vxlan_socket_remove_softc(struct vxlan_socket *, 282 struct vxlan_softc *); 283 284 static struct ifnet * 285 vxlan_multicast_if_ref(struct vxlan_softc *, int); 286 static void vxlan_free_multicast(struct vxlan_softc *); 287 static int vxlan_setup_multicast_interface(struct vxlan_softc *); 288 289 static int vxlan_setup_multicast(struct vxlan_softc *); 290 static int vxlan_setup_socket(struct vxlan_softc *); 291 static void vxlan_setup_interface(struct vxlan_softc *); 292 static int vxlan_valid_init_config(struct vxlan_softc *); 293 static void vxlan_init_wait(struct vxlan_softc *); 294 static void vxlan_init_complete(struct vxlan_softc *); 295 static void vxlan_init(void *); 296 static void vxlan_release(struct vxlan_softc *); 297 static void vxlan_teardown_wait(struct vxlan_softc *); 298 static void vxlan_teardown_complete(struct vxlan_softc *); 299 static void vxlan_teardown_locked(struct vxlan_softc *); 300 static void vxlan_teardown(struct vxlan_softc *); 301 static void vxlan_ifdetach(struct vxlan_softc *, struct ifnet *, 302 struct vxlan_softc_head *); 303 static void vxlan_timer(void *); 304 305 static int vxlan_ctrl_get_config(struct vxlan_softc *, void *); 306 static int vxlan_ctrl_set_vni(struct vxlan_softc *, void *); 307 static int vxlan_ctrl_set_local_addr(struct vxlan_softc *, void *); 308 static int vxlan_ctrl_set_remote_addr(struct vxlan_softc *, void *); 309 static int vxlan_ctrl_set_local_port(struct vxlan_softc *, void *); 310 static int vxlan_ctrl_set_remote_port(struct vxlan_softc *, void *); 311 static int vxlan_ctrl_set_port_range(struct vxlan_softc *, void *); 312 static int vxlan_ctrl_set_ftable_timeout(struct vxlan_softc *, void *); 313 static int vxlan_ctrl_set_ftable_max(struct vxlan_softc *, void *); 314 static int vxlan_ctrl_set_multicast_if(struct vxlan_softc * , void *); 315 static int vxlan_ctrl_set_ttl(struct vxlan_softc *, void *); 316 static int vxlan_ctrl_set_learn(struct vxlan_softc *, void *); 317 static int vxlan_ctrl_ftable_entry_add(struct vxlan_softc *, void *); 318 static int vxlan_ctrl_ftable_entry_rem(struct vxlan_softc *, void *); 319 static int vxlan_ctrl_flush(struct vxlan_softc *, void *); 320 static int vxlan_ioctl_drvspec(struct vxlan_softc *, 321 struct ifdrv *, int); 322 static int vxlan_ioctl_ifflags(struct vxlan_softc *); 323 static int vxlan_ioctl(struct ifnet *, u_long, caddr_t); 324 325 #if defined(INET) || defined(INET6) 326 static uint16_t vxlan_pick_source_port(struct vxlan_softc *, struct mbuf *); 327 static void vxlan_encap_header(struct vxlan_softc *, struct mbuf *, 328 int, uint16_t, uint16_t); 329 #endif 330 static int vxlan_encap4(struct vxlan_softc *, 331 const union vxlan_sockaddr *, struct mbuf *); 332 static int vxlan_encap6(struct vxlan_softc *, 333 const union vxlan_sockaddr *, struct mbuf *); 334 static int vxlan_transmit(struct ifnet *, struct mbuf *); 335 static void vxlan_qflush(struct ifnet *); 336 static void vxlan_rcv_udp_packet(struct mbuf *, int, struct inpcb *, 337 const struct sockaddr *, void *); 338 static int vxlan_input(struct vxlan_socket *, uint32_t, struct mbuf **, 339 const struct sockaddr *); 340 341 static void vxlan_set_default_config(struct vxlan_softc *); 342 static int vxlan_set_user_config(struct vxlan_softc *, 343 struct ifvxlanparam *); 344 static int vxlan_clone_create(struct if_clone *, int, caddr_t); 345 static void vxlan_clone_destroy(struct ifnet *); 346 347 static uint32_t vxlan_mac_hash(struct vxlan_softc *, const uint8_t *); 348 static void vxlan_fakeaddr(struct vxlan_softc *); 349 static int vxlan_media_change(struct ifnet *); 350 static void vxlan_media_status(struct ifnet *, struct ifmediareq *); 351 352 static int vxlan_sockaddr_cmp(const union vxlan_sockaddr *, 353 const struct sockaddr *); 354 static void vxlan_sockaddr_copy(union vxlan_sockaddr *, 355 const struct sockaddr *); 356 static int vxlan_sockaddr_in_equal(const union vxlan_sockaddr *, 357 const struct sockaddr *); 358 static void vxlan_sockaddr_in_copy(union vxlan_sockaddr *, 359 const struct sockaddr *); 360 static int vxlan_sockaddr_supported(const union vxlan_sockaddr *, int); 361 static int vxlan_sockaddr_in_any(const union vxlan_sockaddr *); 362 static int vxlan_sockaddr_in_multicast(const union vxlan_sockaddr *); 363 static int vxlan_sockaddr_in6_embedscope(union vxlan_sockaddr *); 364 365 static int vxlan_can_change_config(struct vxlan_softc *); 366 static int vxlan_check_vni(uint32_t); 367 static int vxlan_check_ttl(int); 368 static int vxlan_check_ftable_timeout(uint32_t); 369 static int vxlan_check_ftable_max(uint32_t); 370 371 static void vxlan_sysctl_setup(struct vxlan_softc *); 372 static void vxlan_sysctl_destroy(struct vxlan_softc *); 373 static int vxlan_tunable_int(struct vxlan_softc *, const char *, int); 374 375 static void vxlan_ifdetach_event(void *, struct ifnet *); 376 static void vxlan_load(void); 377 static void vxlan_unload(void); 378 static int vxlan_modevent(module_t, int, void *); 379 380 static const char vxlan_name[] = "vxlan"; 381 static MALLOC_DEFINE(M_VXLAN, vxlan_name, 382 "Virtual eXtensible LAN Interface"); 383 static struct if_clone *vxlan_cloner; 384 385 static struct mtx vxlan_list_mtx; 386 #define VXLAN_LIST_LOCK() mtx_lock(&vxlan_list_mtx) 387 #define VXLAN_LIST_UNLOCK() mtx_unlock(&vxlan_list_mtx) 388 389 static LIST_HEAD(, vxlan_socket) vxlan_socket_list; 390 391 static eventhandler_tag vxlan_ifdetach_event_tag; 392 393 SYSCTL_DECL(_net_link); 394 SYSCTL_NODE(_net_link, OID_AUTO, vxlan, CTLFLAG_RW, 0, 395 "Virtual eXtensible Local Area Network"); 396 397 static int vxlan_legacy_port = 0; 398 TUNABLE_INT("net.link.vxlan.legacy_port", &vxlan_legacy_port); 399 static int vxlan_reuse_port = 0; 400 TUNABLE_INT("net.link.vxlan.reuse_port", &vxlan_reuse_port); 401 402 /* Default maximum number of addresses in the forwarding table. */ 403 #ifndef VXLAN_FTABLE_MAX 404 #define VXLAN_FTABLE_MAX 2000 405 #endif 406 407 /* Timeout (in seconds) of addresses learned in the forwarding table. */ 408 #ifndef VXLAN_FTABLE_TIMEOUT 409 #define VXLAN_FTABLE_TIMEOUT (20 * 60) 410 #endif 411 412 /* 413 * Maximum timeout (in seconds) of addresses learned in the forwarding 414 * table. 415 */ 416 #ifndef VXLAN_FTABLE_MAX_TIMEOUT 417 #define VXLAN_FTABLE_MAX_TIMEOUT (60 * 60 * 24) 418 #endif 419 420 /* Number of seconds between pruning attempts of the forwarding table. */ 421 #ifndef VXLAN_FTABLE_PRUNE 422 #define VXLAN_FTABLE_PRUNE (5 * 60) 423 #endif 424 425 static int vxlan_ftable_prune_period = VXLAN_FTABLE_PRUNE; 426 427 struct vxlan_control { 428 int (*vxlc_func)(struct vxlan_softc *, void *); 429 int vxlc_argsize; 430 int vxlc_flags; 431 #define VXLAN_CTRL_FLAG_COPYIN 0x01 432 #define VXLAN_CTRL_FLAG_COPYOUT 0x02 433 #define VXLAN_CTRL_FLAG_SUSER 0x04 434 }; 435 436 static const struct vxlan_control vxlan_control_table[] = { 437 [VXLAN_CMD_GET_CONFIG] = 438 { vxlan_ctrl_get_config, sizeof(struct ifvxlancfg), 439 VXLAN_CTRL_FLAG_COPYOUT 440 }, 441 442 [VXLAN_CMD_SET_VNI] = 443 { vxlan_ctrl_set_vni, sizeof(struct ifvxlancmd), 444 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER, 445 }, 446 447 [VXLAN_CMD_SET_LOCAL_ADDR] = 448 { vxlan_ctrl_set_local_addr, sizeof(struct ifvxlancmd), 449 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER, 450 }, 451 452 [VXLAN_CMD_SET_REMOTE_ADDR] = 453 { vxlan_ctrl_set_remote_addr, sizeof(struct ifvxlancmd), 454 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER, 455 }, 456 457 [VXLAN_CMD_SET_LOCAL_PORT] = 458 { vxlan_ctrl_set_local_port, sizeof(struct ifvxlancmd), 459 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER, 460 }, 461 462 [VXLAN_CMD_SET_REMOTE_PORT] = 463 { vxlan_ctrl_set_remote_port, sizeof(struct ifvxlancmd), 464 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER, 465 }, 466 467 [VXLAN_CMD_SET_PORT_RANGE] = 468 { vxlan_ctrl_set_port_range, sizeof(struct ifvxlancmd), 469 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER, 470 }, 471 472 [VXLAN_CMD_SET_FTABLE_TIMEOUT] = 473 { vxlan_ctrl_set_ftable_timeout, sizeof(struct ifvxlancmd), 474 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER, 475 }, 476 477 [VXLAN_CMD_SET_FTABLE_MAX] = 478 { vxlan_ctrl_set_ftable_max, sizeof(struct ifvxlancmd), 479 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER, 480 }, 481 482 [VXLAN_CMD_SET_MULTICAST_IF] = 483 { vxlan_ctrl_set_multicast_if, sizeof(struct ifvxlancmd), 484 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER, 485 }, 486 487 [VXLAN_CMD_SET_TTL] = 488 { vxlan_ctrl_set_ttl, sizeof(struct ifvxlancmd), 489 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER, 490 }, 491 492 [VXLAN_CMD_SET_LEARN] = 493 { vxlan_ctrl_set_learn, sizeof(struct ifvxlancmd), 494 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER, 495 }, 496 497 [VXLAN_CMD_FTABLE_ENTRY_ADD] = 498 { vxlan_ctrl_ftable_entry_add, sizeof(struct ifvxlancmd), 499 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER, 500 }, 501 502 [VXLAN_CMD_FTABLE_ENTRY_REM] = 503 { vxlan_ctrl_ftable_entry_rem, sizeof(struct ifvxlancmd), 504 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER, 505 }, 506 507 [VXLAN_CMD_FLUSH] = 508 { vxlan_ctrl_flush, sizeof(struct ifvxlancmd), 509 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER, 510 }, 511 }; 512 513 static const int vxlan_control_table_size = nitems(vxlan_control_table); 514 515 static int 516 vxlan_ftable_addr_cmp(const uint8_t *a, const uint8_t *b) 517 { 518 int i, d; 519 520 for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) 521 d = ((int)a[i]) - ((int)b[i]); 522 523 return (d); 524 } 525 526 static void 527 vxlan_ftable_init(struct vxlan_softc *sc) 528 { 529 int i; 530 531 sc->vxl_ftable = malloc(sizeof(struct vxlan_ftable_head) * 532 VXLAN_SC_FTABLE_SIZE, M_VXLAN, M_ZERO | M_WAITOK); 533 534 for (i = 0; i < VXLAN_SC_FTABLE_SIZE; i++) 535 LIST_INIT(&sc->vxl_ftable[i]); 536 sc->vxl_ftable_hash_key = arc4random(); 537 } 538 539 static void 540 vxlan_ftable_fini(struct vxlan_softc *sc) 541 { 542 int i; 543 544 for (i = 0; i < VXLAN_SC_FTABLE_SIZE; i++) { 545 KASSERT(LIST_EMPTY(&sc->vxl_ftable[i]), 546 ("%s: vxlan %p ftable[%d] not empty", __func__, sc, i)); 547 } 548 MPASS(sc->vxl_ftable_cnt == 0); 549 550 free(sc->vxl_ftable, M_VXLAN); 551 sc->vxl_ftable = NULL; 552 } 553 554 static void 555 vxlan_ftable_flush(struct vxlan_softc *sc, int all) 556 { 557 struct vxlan_ftable_entry *fe, *tfe; 558 int i; 559 560 for (i = 0; i < VXLAN_SC_FTABLE_SIZE; i++) { 561 LIST_FOREACH_SAFE(fe, &sc->vxl_ftable[i], vxlfe_hash, tfe) { 562 if (all || VXLAN_FE_IS_DYNAMIC(fe)) 563 vxlan_ftable_entry_destroy(sc, fe); 564 } 565 } 566 } 567 568 static void 569 vxlan_ftable_expire(struct vxlan_softc *sc) 570 { 571 struct vxlan_ftable_entry *fe, *tfe; 572 int i; 573 574 VXLAN_LOCK_WASSERT(sc); 575 576 for (i = 0; i < VXLAN_SC_FTABLE_SIZE; i++) { 577 LIST_FOREACH_SAFE(fe, &sc->vxl_ftable[i], vxlfe_hash, tfe) { 578 if (VXLAN_FE_IS_DYNAMIC(fe) && 579 time_uptime >= fe->vxlfe_expire) 580 vxlan_ftable_entry_destroy(sc, fe); 581 } 582 } 583 } 584 585 static int 586 vxlan_ftable_update_locked(struct vxlan_softc *sc, 587 const union vxlan_sockaddr *vxlsa, const uint8_t *mac, 588 struct rm_priotracker *tracker) 589 { 590 struct vxlan_ftable_entry *fe; 591 int error __unused; 592 593 VXLAN_LOCK_ASSERT(sc); 594 595 again: 596 /* 597 * A forwarding entry for this MAC address might already exist. If 598 * so, update it, otherwise create a new one. We may have to upgrade 599 * the lock if we have to change or create an entry. 600 */ 601 fe = vxlan_ftable_entry_lookup(sc, mac); 602 if (fe != NULL) { 603 fe->vxlfe_expire = time_uptime + sc->vxl_ftable_timeout; 604 605 if (!VXLAN_FE_IS_DYNAMIC(fe) || 606 vxlan_sockaddr_in_equal(&fe->vxlfe_raddr, &vxlsa->sa)) 607 return (0); 608 if (!VXLAN_LOCK_WOWNED(sc)) { 609 VXLAN_RUNLOCK(sc, tracker); 610 VXLAN_WLOCK(sc); 611 sc->vxl_stats.ftable_lock_upgrade_failed++; 612 goto again; 613 } 614 vxlan_sockaddr_in_copy(&fe->vxlfe_raddr, &vxlsa->sa); 615 return (0); 616 } 617 618 if (!VXLAN_LOCK_WOWNED(sc)) { 619 VXLAN_RUNLOCK(sc, tracker); 620 VXLAN_WLOCK(sc); 621 sc->vxl_stats.ftable_lock_upgrade_failed++; 622 goto again; 623 } 624 625 if (sc->vxl_ftable_cnt >= sc->vxl_ftable_max) { 626 sc->vxl_stats.ftable_nospace++; 627 return (ENOSPC); 628 } 629 630 fe = vxlan_ftable_entry_alloc(); 631 if (fe == NULL) 632 return (ENOMEM); 633 634 vxlan_ftable_entry_init(sc, fe, mac, &vxlsa->sa, VXLAN_FE_FLAG_DYNAMIC); 635 636 /* The prior lookup failed, so the insert should not. */ 637 error = vxlan_ftable_entry_insert(sc, fe); 638 MPASS(error == 0); 639 640 return (0); 641 } 642 643 static int 644 vxlan_ftable_learn(struct vxlan_softc *sc, const struct sockaddr *sa, 645 const uint8_t *mac) 646 { 647 struct rm_priotracker tracker; 648 union vxlan_sockaddr vxlsa; 649 int error; 650 651 /* 652 * The source port may be randomly selected by the remote host, so 653 * use the port of the default destination address. 654 */ 655 vxlan_sockaddr_copy(&vxlsa, sa); 656 vxlsa.in4.sin_port = sc->vxl_dst_addr.in4.sin_port; 657 658 if (VXLAN_SOCKADDR_IS_IPV6(&vxlsa)) { 659 error = vxlan_sockaddr_in6_embedscope(&vxlsa); 660 if (error) 661 return (error); 662 } 663 664 VXLAN_RLOCK(sc, &tracker); 665 error = vxlan_ftable_update_locked(sc, &vxlsa, mac, &tracker); 666 VXLAN_UNLOCK(sc, &tracker); 667 668 return (error); 669 } 670 671 static int 672 vxlan_ftable_sysctl_dump(SYSCTL_HANDLER_ARGS) 673 { 674 struct rm_priotracker tracker; 675 struct sbuf sb; 676 struct vxlan_softc *sc; 677 struct vxlan_ftable_entry *fe; 678 size_t size; 679 int i, error; 680 681 /* 682 * This is mostly intended for debugging during development. It is 683 * not practical to dump an entire large table this way. 684 */ 685 686 sc = arg1; 687 size = PAGE_SIZE; /* Calculate later. */ 688 689 sbuf_new(&sb, NULL, size, SBUF_FIXEDLEN); 690 sbuf_putc(&sb, '\n'); 691 692 VXLAN_RLOCK(sc, &tracker); 693 for (i = 0; i < VXLAN_SC_FTABLE_SIZE; i++) { 694 LIST_FOREACH(fe, &sc->vxl_ftable[i], vxlfe_hash) { 695 if (sbuf_error(&sb) != 0) 696 break; 697 vxlan_ftable_entry_dump(fe, &sb); 698 } 699 } 700 VXLAN_RUNLOCK(sc, &tracker); 701 702 if (sbuf_len(&sb) == 1) 703 sbuf_setpos(&sb, 0); 704 705 sbuf_finish(&sb); 706 error = sysctl_handle_string(oidp, sbuf_data(&sb), sbuf_len(&sb), req); 707 sbuf_delete(&sb); 708 709 return (error); 710 } 711 712 static struct vxlan_ftable_entry * 713 vxlan_ftable_entry_alloc(void) 714 { 715 struct vxlan_ftable_entry *fe; 716 717 fe = malloc(sizeof(*fe), M_VXLAN, M_ZERO | M_NOWAIT); 718 719 return (fe); 720 } 721 722 static void 723 vxlan_ftable_entry_free(struct vxlan_ftable_entry *fe) 724 { 725 726 free(fe, M_VXLAN); 727 } 728 729 static void 730 vxlan_ftable_entry_init(struct vxlan_softc *sc, struct vxlan_ftable_entry *fe, 731 const uint8_t *mac, const struct sockaddr *sa, uint32_t flags) 732 { 733 734 fe->vxlfe_flags = flags; 735 fe->vxlfe_expire = time_uptime + sc->vxl_ftable_timeout; 736 memcpy(fe->vxlfe_mac, mac, ETHER_ADDR_LEN); 737 vxlan_sockaddr_copy(&fe->vxlfe_raddr, sa); 738 } 739 740 static void 741 vxlan_ftable_entry_destroy(struct vxlan_softc *sc, 742 struct vxlan_ftable_entry *fe) 743 { 744 745 sc->vxl_ftable_cnt--; 746 LIST_REMOVE(fe, vxlfe_hash); 747 vxlan_ftable_entry_free(fe); 748 } 749 750 static int 751 vxlan_ftable_entry_insert(struct vxlan_softc *sc, 752 struct vxlan_ftable_entry *fe) 753 { 754 struct vxlan_ftable_entry *lfe; 755 uint32_t hash; 756 int dir; 757 758 VXLAN_LOCK_WASSERT(sc); 759 hash = VXLAN_SC_FTABLE_HASH(sc, fe->vxlfe_mac); 760 761 lfe = LIST_FIRST(&sc->vxl_ftable[hash]); 762 if (lfe == NULL) { 763 LIST_INSERT_HEAD(&sc->vxl_ftable[hash], fe, vxlfe_hash); 764 goto out; 765 } 766 767 do { 768 dir = vxlan_ftable_addr_cmp(fe->vxlfe_mac, lfe->vxlfe_mac); 769 if (dir == 0) 770 return (EEXIST); 771 if (dir > 0) { 772 LIST_INSERT_BEFORE(lfe, fe, vxlfe_hash); 773 goto out; 774 } else if (LIST_NEXT(lfe, vxlfe_hash) == NULL) { 775 LIST_INSERT_AFTER(lfe, fe, vxlfe_hash); 776 goto out; 777 } else 778 lfe = LIST_NEXT(lfe, vxlfe_hash); 779 } while (lfe != NULL); 780 781 out: 782 sc->vxl_ftable_cnt++; 783 784 return (0); 785 } 786 787 static struct vxlan_ftable_entry * 788 vxlan_ftable_entry_lookup(struct vxlan_softc *sc, const uint8_t *mac) 789 { 790 struct vxlan_ftable_entry *fe; 791 uint32_t hash; 792 int dir; 793 794 VXLAN_LOCK_ASSERT(sc); 795 hash = VXLAN_SC_FTABLE_HASH(sc, mac); 796 797 LIST_FOREACH(fe, &sc->vxl_ftable[hash], vxlfe_hash) { 798 dir = vxlan_ftable_addr_cmp(mac, fe->vxlfe_mac); 799 if (dir == 0) 800 return (fe); 801 if (dir > 0) 802 break; 803 } 804 805 return (NULL); 806 } 807 808 static void 809 vxlan_ftable_entry_dump(struct vxlan_ftable_entry *fe, struct sbuf *sb) 810 { 811 char buf[64]; 812 const union vxlan_sockaddr *sa; 813 const void *addr; 814 int i, len, af, width; 815 816 sa = &fe->vxlfe_raddr; 817 af = sa->sa.sa_family; 818 len = sbuf_len(sb); 819 820 sbuf_printf(sb, "%c 0x%02X ", VXLAN_FE_IS_DYNAMIC(fe) ? 'D' : 'S', 821 fe->vxlfe_flags); 822 823 for (i = 0; i < ETHER_ADDR_LEN - 1; i++) 824 sbuf_printf(sb, "%02X:", fe->vxlfe_mac[i]); 825 sbuf_printf(sb, "%02X ", fe->vxlfe_mac[i]); 826 827 if (af == AF_INET) { 828 addr = &sa->in4.sin_addr; 829 width = INET_ADDRSTRLEN - 1; 830 } else { 831 addr = &sa->in6.sin6_addr; 832 width = INET6_ADDRSTRLEN - 1; 833 } 834 inet_ntop(af, addr, buf, sizeof(buf)); 835 sbuf_printf(sb, "%*s ", width, buf); 836 837 sbuf_printf(sb, "%08jd", (intmax_t)fe->vxlfe_expire); 838 839 sbuf_putc(sb, '\n'); 840 841 /* Truncate a partial line. */ 842 if (sbuf_error(sb) != 0) 843 sbuf_setpos(sb, len); 844 } 845 846 static struct vxlan_socket * 847 vxlan_socket_alloc(const union vxlan_sockaddr *sa) 848 { 849 struct vxlan_socket *vso; 850 int i; 851 852 vso = malloc(sizeof(*vso), M_VXLAN, M_WAITOK | M_ZERO); 853 rm_init(&vso->vxlso_lock, "vxlansorm"); 854 refcount_init(&vso->vxlso_refcnt, 0); 855 for (i = 0; i < VXLAN_SO_VNI_HASH_SIZE; i++) 856 LIST_INIT(&vso->vxlso_vni_hash[i]); 857 vso->vxlso_laddr = *sa; 858 859 return (vso); 860 } 861 862 static void 863 vxlan_socket_destroy(struct vxlan_socket *vso) 864 { 865 struct socket *so; 866 #ifdef INVARIANTS 867 int i; 868 struct vxlan_socket_mc_info *mc; 869 870 for (i = 0; i < VXLAN_SO_MC_MAX_GROUPS; i++) { 871 mc = &vso->vxlso_mc[i]; 872 KASSERT(mc->vxlsomc_gaddr.sa.sa_family == AF_UNSPEC, 873 ("%s: socket %p mc[%d] still has address", 874 __func__, vso, i)); 875 } 876 877 for (i = 0; i < VXLAN_SO_VNI_HASH_SIZE; i++) { 878 KASSERT(LIST_EMPTY(&vso->vxlso_vni_hash[i]), 879 ("%s: socket %p vni_hash[%d] not empty", 880 __func__, vso, i)); 881 } 882 #endif 883 so = vso->vxlso_sock; 884 if (so != NULL) { 885 vso->vxlso_sock = NULL; 886 soclose(so); 887 } 888 889 rm_destroy(&vso->vxlso_lock); 890 free(vso, M_VXLAN); 891 } 892 893 static void 894 vxlan_socket_release(struct vxlan_socket *vso) 895 { 896 int destroy; 897 898 VXLAN_LIST_LOCK(); 899 destroy = VXLAN_SO_RELEASE(vso); 900 if (destroy != 0) 901 LIST_REMOVE(vso, vxlso_entry); 902 VXLAN_LIST_UNLOCK(); 903 904 if (destroy != 0) 905 vxlan_socket_destroy(vso); 906 } 907 908 static struct vxlan_socket * 909 vxlan_socket_lookup(union vxlan_sockaddr *vxlsa) 910 { 911 struct vxlan_socket *vso; 912 913 VXLAN_LIST_LOCK(); 914 LIST_FOREACH(vso, &vxlan_socket_list, vxlso_entry) { 915 if (vxlan_sockaddr_cmp(&vso->vxlso_laddr, &vxlsa->sa) == 0) { 916 VXLAN_SO_ACQUIRE(vso); 917 break; 918 } 919 } 920 VXLAN_LIST_UNLOCK(); 921 922 return (vso); 923 } 924 925 static void 926 vxlan_socket_insert(struct vxlan_socket *vso) 927 { 928 929 VXLAN_LIST_LOCK(); 930 VXLAN_SO_ACQUIRE(vso); 931 LIST_INSERT_HEAD(&vxlan_socket_list, vso, vxlso_entry); 932 VXLAN_LIST_UNLOCK(); 933 } 934 935 static int 936 vxlan_socket_init(struct vxlan_socket *vso, struct ifnet *ifp) 937 { 938 struct thread *td; 939 int error; 940 941 td = curthread; 942 943 error = socreate(vso->vxlso_laddr.sa.sa_family, &vso->vxlso_sock, 944 SOCK_DGRAM, IPPROTO_UDP, td->td_ucred, td); 945 if (error) { 946 if_printf(ifp, "cannot create socket: %d\n", error); 947 return (error); 948 } 949 950 error = udp_set_kernel_tunneling(vso->vxlso_sock, 951 vxlan_rcv_udp_packet, NULL, vso); 952 if (error) { 953 if_printf(ifp, "cannot set tunneling function: %d\n", error); 954 return (error); 955 } 956 957 if (vxlan_reuse_port != 0) { 958 struct sockopt sopt; 959 int val = 1; 960 961 bzero(&sopt, sizeof(sopt)); 962 sopt.sopt_dir = SOPT_SET; 963 sopt.sopt_level = IPPROTO_IP; 964 sopt.sopt_name = SO_REUSEPORT; 965 sopt.sopt_val = &val; 966 sopt.sopt_valsize = sizeof(val); 967 error = sosetopt(vso->vxlso_sock, &sopt); 968 if (error) { 969 if_printf(ifp, 970 "cannot set REUSEADDR socket opt: %d\n", error); 971 return (error); 972 } 973 } 974 975 return (0); 976 } 977 978 static int 979 vxlan_socket_bind(struct vxlan_socket *vso, struct ifnet *ifp) 980 { 981 union vxlan_sockaddr laddr; 982 struct thread *td; 983 int error; 984 985 td = curthread; 986 laddr = vso->vxlso_laddr; 987 988 error = sobind(vso->vxlso_sock, &laddr.sa, td); 989 if (error) { 990 if (error != EADDRINUSE) 991 if_printf(ifp, "cannot bind socket: %d\n", error); 992 return (error); 993 } 994 995 return (0); 996 } 997 998 static int 999 vxlan_socket_create(struct ifnet *ifp, int multicast, 1000 const union vxlan_sockaddr *saddr, struct vxlan_socket **vsop) 1001 { 1002 union vxlan_sockaddr laddr; 1003 struct vxlan_socket *vso; 1004 int error; 1005 1006 laddr = *saddr; 1007 1008 /* 1009 * If this socket will be multicast, then only the local port 1010 * must be specified when binding. 1011 */ 1012 if (multicast != 0) { 1013 if (VXLAN_SOCKADDR_IS_IPV4(&laddr)) 1014 laddr.in4.sin_addr.s_addr = INADDR_ANY; 1015 #ifdef INET6 1016 else 1017 laddr.in6.sin6_addr = in6addr_any; 1018 #endif 1019 } 1020 1021 vso = vxlan_socket_alloc(&laddr); 1022 if (vso == NULL) 1023 return (ENOMEM); 1024 1025 error = vxlan_socket_init(vso, ifp); 1026 if (error) 1027 goto fail; 1028 1029 error = vxlan_socket_bind(vso, ifp); 1030 if (error) 1031 goto fail; 1032 1033 /* 1034 * There is a small window between the bind completing and 1035 * inserting the socket, so that a concurrent create may fail. 1036 * Let's not worry about that for now. 1037 */ 1038 vxlan_socket_insert(vso); 1039 *vsop = vso; 1040 1041 return (0); 1042 1043 fail: 1044 vxlan_socket_destroy(vso); 1045 1046 return (error); 1047 } 1048 1049 static void 1050 vxlan_socket_ifdetach(struct vxlan_socket *vso, struct ifnet *ifp, 1051 struct vxlan_softc_head *list) 1052 { 1053 struct rm_priotracker tracker; 1054 struct vxlan_softc *sc; 1055 int i; 1056 1057 VXLAN_SO_RLOCK(vso, &tracker); 1058 for (i = 0; i < VXLAN_SO_VNI_HASH_SIZE; i++) { 1059 LIST_FOREACH(sc, &vso->vxlso_vni_hash[i], vxl_entry) 1060 vxlan_ifdetach(sc, ifp, list); 1061 } 1062 VXLAN_SO_RUNLOCK(vso, &tracker); 1063 } 1064 1065 static struct vxlan_socket * 1066 vxlan_socket_mc_lookup(const union vxlan_sockaddr *vxlsa) 1067 { 1068 union vxlan_sockaddr laddr; 1069 struct vxlan_socket *vso; 1070 1071 laddr = *vxlsa; 1072 1073 if (VXLAN_SOCKADDR_IS_IPV4(&laddr)) 1074 laddr.in4.sin_addr.s_addr = INADDR_ANY; 1075 #ifdef INET6 1076 else 1077 laddr.in6.sin6_addr = in6addr_any; 1078 #endif 1079 1080 vso = vxlan_socket_lookup(&laddr); 1081 1082 return (vso); 1083 } 1084 1085 static int 1086 vxlan_sockaddr_mc_info_match(const struct vxlan_socket_mc_info *mc, 1087 const union vxlan_sockaddr *group, const union vxlan_sockaddr *local, 1088 int ifidx) 1089 { 1090 1091 if (!vxlan_sockaddr_in_any(local) && 1092 !vxlan_sockaddr_in_equal(&mc->vxlsomc_saddr, &local->sa)) 1093 return (0); 1094 if (!vxlan_sockaddr_in_equal(&mc->vxlsomc_gaddr, &group->sa)) 1095 return (0); 1096 if (ifidx != 0 && ifidx != mc->vxlsomc_ifidx) 1097 return (0); 1098 1099 return (1); 1100 } 1101 1102 static int 1103 vxlan_socket_mc_join_group(struct vxlan_socket *vso, 1104 const union vxlan_sockaddr *group, const union vxlan_sockaddr *local, 1105 int *ifidx, union vxlan_sockaddr *source) 1106 { 1107 struct sockopt sopt; 1108 int error; 1109 1110 *source = *local; 1111 1112 if (VXLAN_SOCKADDR_IS_IPV4(group)) { 1113 struct ip_mreq mreq; 1114 1115 mreq.imr_multiaddr = group->in4.sin_addr; 1116 mreq.imr_interface = local->in4.sin_addr; 1117 1118 bzero(&sopt, sizeof(sopt)); 1119 sopt.sopt_dir = SOPT_SET; 1120 sopt.sopt_level = IPPROTO_IP; 1121 sopt.sopt_name = IP_ADD_MEMBERSHIP; 1122 sopt.sopt_val = &mreq; 1123 sopt.sopt_valsize = sizeof(mreq); 1124 error = sosetopt(vso->vxlso_sock, &sopt); 1125 if (error) 1126 return (error); 1127 1128 /* 1129 * BMV: Ideally, there would be a formal way for us to get 1130 * the local interface that was selected based on the 1131 * imr_interface address. We could then update *ifidx so 1132 * vxlan_sockaddr_mc_info_match() would return a match for 1133 * later creates that explicitly set the multicast interface. 1134 * 1135 * If we really need to, we can of course look in the INP's 1136 * membership list: 1137 * sotoinpcb(vso->vxlso_sock)->inp_moptions-> 1138 * imo_membership[]->inm_ifp 1139 * similarly to imo_match_group(). 1140 */ 1141 source->in4.sin_addr = local->in4.sin_addr; 1142 1143 } else if (VXLAN_SOCKADDR_IS_IPV6(group)) { 1144 struct ipv6_mreq mreq; 1145 1146 mreq.ipv6mr_multiaddr = group->in6.sin6_addr; 1147 mreq.ipv6mr_interface = *ifidx; 1148 1149 bzero(&sopt, sizeof(sopt)); 1150 sopt.sopt_dir = SOPT_SET; 1151 sopt.sopt_level = IPPROTO_IPV6; 1152 sopt.sopt_name = IPV6_JOIN_GROUP; 1153 sopt.sopt_val = &mreq; 1154 sopt.sopt_valsize = sizeof(mreq); 1155 error = sosetopt(vso->vxlso_sock, &sopt); 1156 if (error) 1157 return (error); 1158 1159 /* 1160 * BMV: As with IPv4, we would really like to know what 1161 * interface in6p_lookup_mcast_ifp() selected. 1162 */ 1163 } else 1164 error = EAFNOSUPPORT; 1165 1166 return (error); 1167 } 1168 1169 static int 1170 vxlan_socket_mc_leave_group(struct vxlan_socket *vso, 1171 const union vxlan_sockaddr *group, const union vxlan_sockaddr *source, 1172 int ifidx) 1173 { 1174 struct sockopt sopt; 1175 int error; 1176 1177 bzero(&sopt, sizeof(sopt)); 1178 sopt.sopt_dir = SOPT_SET; 1179 1180 if (VXLAN_SOCKADDR_IS_IPV4(group)) { 1181 struct ip_mreq mreq; 1182 1183 mreq.imr_multiaddr = group->in4.sin_addr; 1184 mreq.imr_interface = source->in4.sin_addr; 1185 1186 sopt.sopt_level = IPPROTO_IP; 1187 sopt.sopt_name = IP_DROP_MEMBERSHIP; 1188 sopt.sopt_val = &mreq; 1189 sopt.sopt_valsize = sizeof(mreq); 1190 error = sosetopt(vso->vxlso_sock, &sopt); 1191 1192 } else if (VXLAN_SOCKADDR_IS_IPV6(group)) { 1193 struct ipv6_mreq mreq; 1194 1195 mreq.ipv6mr_multiaddr = group->in6.sin6_addr; 1196 mreq.ipv6mr_interface = ifidx; 1197 1198 sopt.sopt_level = IPPROTO_IPV6; 1199 sopt.sopt_name = IPV6_LEAVE_GROUP; 1200 sopt.sopt_val = &mreq; 1201 sopt.sopt_valsize = sizeof(mreq); 1202 error = sosetopt(vso->vxlso_sock, &sopt); 1203 1204 } else 1205 error = EAFNOSUPPORT; 1206 1207 return (error); 1208 } 1209 1210 static int 1211 vxlan_socket_mc_add_group(struct vxlan_socket *vso, 1212 const union vxlan_sockaddr *group, const union vxlan_sockaddr *local, 1213 int ifidx, int *idx) 1214 { 1215 union vxlan_sockaddr source; 1216 struct vxlan_socket_mc_info *mc; 1217 int i, empty, error; 1218 1219 /* 1220 * Within a socket, the same multicast group may be used by multiple 1221 * interfaces, each with a different network identifier. But a socket 1222 * may only join a multicast group once, so keep track of the users 1223 * here. 1224 */ 1225 1226 VXLAN_SO_WLOCK(vso); 1227 for (empty = 0, i = 0; i < VXLAN_SO_MC_MAX_GROUPS; i++) { 1228 mc = &vso->vxlso_mc[i]; 1229 1230 if (mc->vxlsomc_gaddr.sa.sa_family == AF_UNSPEC) { 1231 empty++; 1232 continue; 1233 } 1234 1235 if (vxlan_sockaddr_mc_info_match(mc, group, local, ifidx)) 1236 goto out; 1237 } 1238 VXLAN_SO_WUNLOCK(vso); 1239 1240 if (empty == 0) 1241 return (ENOSPC); 1242 1243 error = vxlan_socket_mc_join_group(vso, group, local, &ifidx, &source); 1244 if (error) 1245 return (error); 1246 1247 VXLAN_SO_WLOCK(vso); 1248 for (i = 0; i < VXLAN_SO_MC_MAX_GROUPS; i++) { 1249 mc = &vso->vxlso_mc[i]; 1250 1251 if (mc->vxlsomc_gaddr.sa.sa_family == AF_UNSPEC) { 1252 vxlan_sockaddr_copy(&mc->vxlsomc_gaddr, &group->sa); 1253 vxlan_sockaddr_copy(&mc->vxlsomc_saddr, &source.sa); 1254 mc->vxlsomc_ifidx = ifidx; 1255 goto out; 1256 } 1257 } 1258 VXLAN_SO_WUNLOCK(vso); 1259 1260 error = vxlan_socket_mc_leave_group(vso, group, &source, ifidx); 1261 MPASS(error == 0); 1262 1263 return (ENOSPC); 1264 1265 out: 1266 mc->vxlsomc_users++; 1267 VXLAN_SO_WUNLOCK(vso); 1268 1269 *idx = i; 1270 1271 return (0); 1272 } 1273 1274 static void 1275 vxlan_socket_mc_release_group_by_idx(struct vxlan_socket *vso, int idx) 1276 { 1277 union vxlan_sockaddr group, source; 1278 struct vxlan_socket_mc_info *mc; 1279 int ifidx, leave; 1280 1281 KASSERT(idx >= 0 && idx < VXLAN_SO_MC_MAX_GROUPS, 1282 ("%s: vso %p idx %d out of bounds", __func__, vso, idx)); 1283 1284 leave = 0; 1285 mc = &vso->vxlso_mc[idx]; 1286 1287 VXLAN_SO_WLOCK(vso); 1288 mc->vxlsomc_users--; 1289 if (mc->vxlsomc_users == 0) { 1290 group = mc->vxlsomc_gaddr; 1291 source = mc->vxlsomc_saddr; 1292 ifidx = mc->vxlsomc_ifidx; 1293 bzero(mc, sizeof(*mc)); 1294 leave = 1; 1295 } 1296 VXLAN_SO_WUNLOCK(vso); 1297 1298 if (leave != 0) { 1299 /* 1300 * Our socket's membership in this group may have already 1301 * been removed if we joined through an interface that's 1302 * been detached. 1303 */ 1304 vxlan_socket_mc_leave_group(vso, &group, &source, ifidx); 1305 } 1306 } 1307 1308 static struct vxlan_softc * 1309 vxlan_socket_lookup_softc_locked(struct vxlan_socket *vso, uint32_t vni) 1310 { 1311 struct vxlan_softc *sc; 1312 uint32_t hash; 1313 1314 VXLAN_SO_LOCK_ASSERT(vso); 1315 hash = VXLAN_SO_VNI_HASH(vni); 1316 1317 LIST_FOREACH(sc, &vso->vxlso_vni_hash[hash], vxl_entry) { 1318 if (sc->vxl_vni == vni) { 1319 VXLAN_ACQUIRE(sc); 1320 break; 1321 } 1322 } 1323 1324 return (sc); 1325 } 1326 1327 static struct vxlan_softc * 1328 vxlan_socket_lookup_softc(struct vxlan_socket *vso, uint32_t vni) 1329 { 1330 struct rm_priotracker tracker; 1331 struct vxlan_softc *sc; 1332 1333 VXLAN_SO_RLOCK(vso, &tracker); 1334 sc = vxlan_socket_lookup_softc_locked(vso, vni); 1335 VXLAN_SO_RUNLOCK(vso, &tracker); 1336 1337 return (sc); 1338 } 1339 1340 static int 1341 vxlan_socket_insert_softc(struct vxlan_socket *vso, struct vxlan_softc *sc) 1342 { 1343 struct vxlan_softc *tsc; 1344 uint32_t vni, hash; 1345 1346 vni = sc->vxl_vni; 1347 hash = VXLAN_SO_VNI_HASH(vni); 1348 1349 VXLAN_SO_WLOCK(vso); 1350 tsc = vxlan_socket_lookup_softc_locked(vso, vni); 1351 if (tsc != NULL) { 1352 VXLAN_SO_WUNLOCK(vso); 1353 vxlan_release(tsc); 1354 return (EEXIST); 1355 } 1356 1357 VXLAN_ACQUIRE(sc); 1358 LIST_INSERT_HEAD(&vso->vxlso_vni_hash[hash], sc, vxl_entry); 1359 VXLAN_SO_WUNLOCK(vso); 1360 1361 return (0); 1362 } 1363 1364 static void 1365 vxlan_socket_remove_softc(struct vxlan_socket *vso, struct vxlan_softc *sc) 1366 { 1367 1368 VXLAN_SO_WLOCK(vso); 1369 LIST_REMOVE(sc, vxl_entry); 1370 VXLAN_SO_WUNLOCK(vso); 1371 1372 vxlan_release(sc); 1373 } 1374 1375 static struct ifnet * 1376 vxlan_multicast_if_ref(struct vxlan_softc *sc, int ipv4) 1377 { 1378 struct ifnet *ifp; 1379 1380 VXLAN_LOCK_ASSERT(sc); 1381 1382 if (ipv4 && sc->vxl_im4o != NULL) 1383 ifp = sc->vxl_im4o->imo_multicast_ifp; 1384 else if (!ipv4 && sc->vxl_im6o != NULL) 1385 ifp = sc->vxl_im6o->im6o_multicast_ifp; 1386 else 1387 ifp = NULL; 1388 1389 if (ifp != NULL) 1390 if_ref(ifp); 1391 1392 return (ifp); 1393 } 1394 1395 static void 1396 vxlan_free_multicast(struct vxlan_softc *sc) 1397 { 1398 1399 if (sc->vxl_mc_ifp != NULL) { 1400 if_rele(sc->vxl_mc_ifp); 1401 sc->vxl_mc_ifp = NULL; 1402 sc->vxl_mc_ifindex = 0; 1403 } 1404 1405 if (sc->vxl_im4o != NULL) { 1406 free(sc->vxl_im4o, M_VXLAN); 1407 sc->vxl_im4o = NULL; 1408 } 1409 1410 if (sc->vxl_im6o != NULL) { 1411 free(sc->vxl_im6o, M_VXLAN); 1412 sc->vxl_im6o = NULL; 1413 } 1414 } 1415 1416 static int 1417 vxlan_setup_multicast_interface(struct vxlan_softc *sc) 1418 { 1419 struct ifnet *ifp; 1420 1421 ifp = ifunit_ref(sc->vxl_mc_ifname); 1422 if (ifp == NULL) { 1423 if_printf(sc->vxl_ifp, "multicast interface %s does " 1424 "not exist\n", sc->vxl_mc_ifname); 1425 return (ENOENT); 1426 } 1427 1428 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 1429 if_printf(sc->vxl_ifp, "interface %s does not support " 1430 "multicast\n", sc->vxl_mc_ifname); 1431 if_rele(ifp); 1432 return (ENOTSUP); 1433 } 1434 1435 sc->vxl_mc_ifp = ifp; 1436 sc->vxl_mc_ifindex = ifp->if_index; 1437 1438 return (0); 1439 } 1440 1441 static int 1442 vxlan_setup_multicast(struct vxlan_softc *sc) 1443 { 1444 const union vxlan_sockaddr *group; 1445 int error; 1446 1447 group = &sc->vxl_dst_addr; 1448 error = 0; 1449 1450 if (sc->vxl_mc_ifname[0] != '\0') { 1451 error = vxlan_setup_multicast_interface(sc); 1452 if (error) 1453 return (error); 1454 } 1455 1456 /* 1457 * Initialize an multicast options structure that is sufficiently 1458 * populated for use in the respective IP output routine. This 1459 * structure is typically stored in the socket, but our sockets 1460 * may be shared among multiple interfaces. 1461 */ 1462 if (VXLAN_SOCKADDR_IS_IPV4(group)) { 1463 sc->vxl_im4o = malloc(sizeof(struct ip_moptions), M_VXLAN, 1464 M_ZERO | M_WAITOK); 1465 sc->vxl_im4o->imo_multicast_ifp = sc->vxl_mc_ifp; 1466 sc->vxl_im4o->imo_multicast_ttl = sc->vxl_ttl; 1467 sc->vxl_im4o->imo_multicast_vif = -1; 1468 } else if (VXLAN_SOCKADDR_IS_IPV6(group)) { 1469 sc->vxl_im6o = malloc(sizeof(struct ip6_moptions), M_VXLAN, 1470 M_ZERO | M_WAITOK); 1471 sc->vxl_im6o->im6o_multicast_ifp = sc->vxl_mc_ifp; 1472 sc->vxl_im6o->im6o_multicast_hlim = sc->vxl_ttl; 1473 } 1474 1475 return (error); 1476 } 1477 1478 static int 1479 vxlan_setup_socket(struct vxlan_softc *sc) 1480 { 1481 struct vxlan_socket *vso; 1482 struct ifnet *ifp; 1483 union vxlan_sockaddr *saddr, *daddr; 1484 int multicast, error; 1485 1486 vso = NULL; 1487 ifp = sc->vxl_ifp; 1488 saddr = &sc->vxl_src_addr; 1489 daddr = &sc->vxl_dst_addr; 1490 1491 multicast = vxlan_sockaddr_in_multicast(daddr); 1492 MPASS(multicast != -1); 1493 sc->vxl_vso_mc_index = -1; 1494 1495 /* 1496 * Try to create the socket. If that fails, attempt to use an 1497 * existing socket. 1498 */ 1499 error = vxlan_socket_create(ifp, multicast, saddr, &vso); 1500 if (error) { 1501 if (multicast != 0) 1502 vso = vxlan_socket_mc_lookup(saddr); 1503 else 1504 vso = vxlan_socket_lookup(saddr); 1505 1506 if (vso == NULL) { 1507 if_printf(ifp, "cannot create socket (error: %d), " 1508 "and no existing socket found\n", error); 1509 goto out; 1510 } 1511 } 1512 1513 if (multicast != 0) { 1514 error = vxlan_setup_multicast(sc); 1515 if (error) 1516 goto out; 1517 1518 error = vxlan_socket_mc_add_group(vso, daddr, saddr, 1519 sc->vxl_mc_ifindex, &sc->vxl_vso_mc_index); 1520 if (error) 1521 goto out; 1522 } 1523 1524 sc->vxl_sock = vso; 1525 error = vxlan_socket_insert_softc(vso, sc); 1526 if (error) { 1527 sc->vxl_sock = NULL; 1528 if_printf(ifp, "network identifier %d already exists in " 1529 "this socket\n", sc->vxl_vni); 1530 goto out; 1531 } 1532 1533 return (0); 1534 1535 out: 1536 if (vso != NULL) { 1537 if (sc->vxl_vso_mc_index != -1) { 1538 vxlan_socket_mc_release_group_by_idx(vso, 1539 sc->vxl_vso_mc_index); 1540 sc->vxl_vso_mc_index = -1; 1541 } 1542 if (multicast != 0) 1543 vxlan_free_multicast(sc); 1544 vxlan_socket_release(vso); 1545 } 1546 1547 return (error); 1548 } 1549 1550 static void 1551 vxlan_setup_interface(struct vxlan_softc *sc) 1552 { 1553 struct ifnet *ifp; 1554 1555 ifp = sc->vxl_ifp; 1556 ifp->if_hdrlen = ETHER_HDR_LEN + sizeof(struct vxlanudphdr); 1557 1558 if (VXLAN_SOCKADDR_IS_IPV4(&sc->vxl_dst_addr) != 0) 1559 ifp->if_hdrlen += sizeof(struct ip); 1560 else if (VXLAN_SOCKADDR_IS_IPV6(&sc->vxl_dst_addr) != 0) 1561 ifp->if_hdrlen += sizeof(struct ip6_hdr); 1562 } 1563 1564 static int 1565 vxlan_valid_init_config(struct vxlan_softc *sc) 1566 { 1567 const char *reason; 1568 1569 if (vxlan_check_vni(sc->vxl_vni) != 0) { 1570 reason = "invalid virtual network identifier specified"; 1571 goto fail; 1572 } 1573 1574 if (vxlan_sockaddr_supported(&sc->vxl_src_addr, 1) == 0) { 1575 reason = "source address type is not supported"; 1576 goto fail; 1577 } 1578 1579 if (vxlan_sockaddr_supported(&sc->vxl_dst_addr, 0) == 0) { 1580 reason = "destination address type is not supported"; 1581 goto fail; 1582 } 1583 1584 if (vxlan_sockaddr_in_any(&sc->vxl_dst_addr) != 0) { 1585 reason = "no valid destination address specified"; 1586 goto fail; 1587 } 1588 1589 if (vxlan_sockaddr_in_multicast(&sc->vxl_dst_addr) == 0 && 1590 sc->vxl_mc_ifname[0] != '\0') { 1591 reason = "can only specify interface with a group address"; 1592 goto fail; 1593 } 1594 1595 if (vxlan_sockaddr_in_any(&sc->vxl_src_addr) == 0) { 1596 if (VXLAN_SOCKADDR_IS_IPV4(&sc->vxl_src_addr) ^ 1597 VXLAN_SOCKADDR_IS_IPV4(&sc->vxl_dst_addr)) { 1598 reason = "source and destination address must both " 1599 "be either IPv4 or IPv6"; 1600 goto fail; 1601 } 1602 } 1603 1604 if (sc->vxl_src_addr.in4.sin_port == 0) { 1605 reason = "local port not specified"; 1606 goto fail; 1607 } 1608 1609 if (sc->vxl_dst_addr.in4.sin_port == 0) { 1610 reason = "remote port not specified"; 1611 goto fail; 1612 } 1613 1614 return (0); 1615 1616 fail: 1617 if_printf(sc->vxl_ifp, "cannot initialize interface: %s\n", reason); 1618 return (EINVAL); 1619 } 1620 1621 static void 1622 vxlan_init_wait(struct vxlan_softc *sc) 1623 { 1624 1625 VXLAN_LOCK_WASSERT(sc); 1626 while (sc->vxl_flags & VXLAN_FLAG_INIT) 1627 rm_sleep(sc, &sc->vxl_lock, 0, "vxlint", hz); 1628 } 1629 1630 static void 1631 vxlan_init_complete(struct vxlan_softc *sc) 1632 { 1633 1634 VXLAN_WLOCK(sc); 1635 sc->vxl_flags &= ~VXLAN_FLAG_INIT; 1636 wakeup(sc); 1637 VXLAN_WUNLOCK(sc); 1638 } 1639 1640 static void 1641 vxlan_init(void *xsc) 1642 { 1643 static const uint8_t empty_mac[ETHER_ADDR_LEN]; 1644 struct vxlan_softc *sc; 1645 struct ifnet *ifp; 1646 1647 sc = xsc; 1648 ifp = sc->vxl_ifp; 1649 1650 VXLAN_WLOCK(sc); 1651 if (ifp->if_drv_flags & IFF_DRV_RUNNING) { 1652 VXLAN_WUNLOCK(sc); 1653 return; 1654 } 1655 sc->vxl_flags |= VXLAN_FLAG_INIT; 1656 VXLAN_WUNLOCK(sc); 1657 1658 if (vxlan_valid_init_config(sc) != 0) 1659 goto out; 1660 1661 vxlan_setup_interface(sc); 1662 1663 if (vxlan_setup_socket(sc) != 0) 1664 goto out; 1665 1666 /* Initialize the default forwarding entry. */ 1667 vxlan_ftable_entry_init(sc, &sc->vxl_default_fe, empty_mac, 1668 &sc->vxl_dst_addr.sa, VXLAN_FE_FLAG_STATIC); 1669 1670 VXLAN_WLOCK(sc); 1671 ifp->if_drv_flags |= IFF_DRV_RUNNING; 1672 callout_reset(&sc->vxl_callout, vxlan_ftable_prune_period * hz, 1673 vxlan_timer, sc); 1674 VXLAN_WUNLOCK(sc); 1675 1676 if_link_state_change(ifp, LINK_STATE_UP); 1677 out: 1678 vxlan_init_complete(sc); 1679 } 1680 1681 static void 1682 vxlan_release(struct vxlan_softc *sc) 1683 { 1684 1685 /* 1686 * The softc may be destroyed as soon as we release our reference, 1687 * so we cannot serialize the wakeup with the softc lock. We use a 1688 * timeout in our sleeps so a missed wakeup is unfortunate but not 1689 * fatal. 1690 */ 1691 if (VXLAN_RELEASE(sc) != 0) 1692 wakeup(sc); 1693 } 1694 1695 static void 1696 vxlan_teardown_wait(struct vxlan_softc *sc) 1697 { 1698 1699 VXLAN_LOCK_WASSERT(sc); 1700 while (sc->vxl_flags & VXLAN_FLAG_TEARDOWN) 1701 rm_sleep(sc, &sc->vxl_lock, 0, "vxltrn", hz); 1702 } 1703 1704 static void 1705 vxlan_teardown_complete(struct vxlan_softc *sc) 1706 { 1707 1708 VXLAN_WLOCK(sc); 1709 sc->vxl_flags &= ~VXLAN_FLAG_TEARDOWN; 1710 wakeup(sc); 1711 VXLAN_WUNLOCK(sc); 1712 } 1713 1714 static void 1715 vxlan_teardown_locked(struct vxlan_softc *sc) 1716 { 1717 struct ifnet *ifp; 1718 struct vxlan_socket *vso; 1719 1720 ifp = sc->vxl_ifp; 1721 1722 VXLAN_LOCK_WASSERT(sc); 1723 MPASS(sc->vxl_flags & VXLAN_FLAG_TEARDOWN); 1724 1725 ifp->if_flags &= ~IFF_UP; 1726 ifp->if_drv_flags &= ~IFF_DRV_RUNNING; 1727 callout_stop(&sc->vxl_callout); 1728 vso = sc->vxl_sock; 1729 sc->vxl_sock = NULL; 1730 1731 VXLAN_WUNLOCK(sc); 1732 if_link_state_change(ifp, LINK_STATE_DOWN); 1733 1734 if (vso != NULL) { 1735 vxlan_socket_remove_softc(vso, sc); 1736 1737 if (sc->vxl_vso_mc_index != -1) { 1738 vxlan_socket_mc_release_group_by_idx(vso, 1739 sc->vxl_vso_mc_index); 1740 sc->vxl_vso_mc_index = -1; 1741 } 1742 } 1743 1744 VXLAN_WLOCK(sc); 1745 while (sc->vxl_refcnt != 0) 1746 rm_sleep(sc, &sc->vxl_lock, 0, "vxldrn", hz); 1747 VXLAN_WUNLOCK(sc); 1748 1749 callout_drain(&sc->vxl_callout); 1750 1751 vxlan_free_multicast(sc); 1752 if (vso != NULL) 1753 vxlan_socket_release(vso); 1754 1755 vxlan_teardown_complete(sc); 1756 } 1757 1758 static void 1759 vxlan_teardown(struct vxlan_softc *sc) 1760 { 1761 1762 VXLAN_WLOCK(sc); 1763 if (sc->vxl_flags & VXLAN_FLAG_TEARDOWN) { 1764 vxlan_teardown_wait(sc); 1765 VXLAN_WUNLOCK(sc); 1766 return; 1767 } 1768 1769 sc->vxl_flags |= VXLAN_FLAG_TEARDOWN; 1770 vxlan_teardown_locked(sc); 1771 } 1772 1773 static void 1774 vxlan_ifdetach(struct vxlan_softc *sc, struct ifnet *ifp, 1775 struct vxlan_softc_head *list) 1776 { 1777 1778 VXLAN_WLOCK(sc); 1779 1780 if (sc->vxl_mc_ifp != ifp) 1781 goto out; 1782 if (sc->vxl_flags & VXLAN_FLAG_TEARDOWN) 1783 goto out; 1784 1785 sc->vxl_flags |= VXLAN_FLAG_TEARDOWN; 1786 LIST_INSERT_HEAD(list, sc, vxl_ifdetach_list); 1787 1788 out: 1789 VXLAN_WUNLOCK(sc); 1790 } 1791 1792 static void 1793 vxlan_timer(void *xsc) 1794 { 1795 struct vxlan_softc *sc; 1796 1797 sc = xsc; 1798 VXLAN_LOCK_WASSERT(sc); 1799 1800 vxlan_ftable_expire(sc); 1801 callout_schedule(&sc->vxl_callout, vxlan_ftable_prune_period * hz); 1802 } 1803 1804 static int 1805 vxlan_ioctl_ifflags(struct vxlan_softc *sc) 1806 { 1807 struct ifnet *ifp; 1808 1809 ifp = sc->vxl_ifp; 1810 1811 if (ifp->if_flags & IFF_UP) { 1812 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) 1813 vxlan_init(sc); 1814 } else { 1815 if (ifp->if_drv_flags & IFF_DRV_RUNNING) 1816 vxlan_teardown(sc); 1817 } 1818 1819 return (0); 1820 } 1821 1822 static int 1823 vxlan_ctrl_get_config(struct vxlan_softc *sc, void *arg) 1824 { 1825 struct rm_priotracker tracker; 1826 struct ifvxlancfg *cfg; 1827 1828 cfg = arg; 1829 bzero(cfg, sizeof(*cfg)); 1830 1831 VXLAN_RLOCK(sc, &tracker); 1832 cfg->vxlc_vni = sc->vxl_vni; 1833 memcpy(&cfg->vxlc_local_sa, &sc->vxl_src_addr, 1834 sizeof(union vxlan_sockaddr)); 1835 memcpy(&cfg->vxlc_remote_sa, &sc->vxl_dst_addr, 1836 sizeof(union vxlan_sockaddr)); 1837 cfg->vxlc_mc_ifindex = sc->vxl_mc_ifindex; 1838 cfg->vxlc_ftable_cnt = sc->vxl_ftable_cnt; 1839 cfg->vxlc_ftable_max = sc->vxl_ftable_max; 1840 cfg->vxlc_ftable_timeout = sc->vxl_ftable_timeout; 1841 cfg->vxlc_port_min = sc->vxl_min_port; 1842 cfg->vxlc_port_max = sc->vxl_max_port; 1843 cfg->vxlc_learn = (sc->vxl_flags & VXLAN_FLAG_LEARN) != 0; 1844 cfg->vxlc_ttl = sc->vxl_ttl; 1845 VXLAN_RUNLOCK(sc, &tracker); 1846 1847 #ifdef INET6 1848 if (VXLAN_SOCKADDR_IS_IPV6(&cfg->vxlc_local_sa)) 1849 sa6_recoverscope(&cfg->vxlc_local_sa.in6); 1850 if (VXLAN_SOCKADDR_IS_IPV6(&cfg->vxlc_remote_sa)) 1851 sa6_recoverscope(&cfg->vxlc_remote_sa.in6); 1852 #endif 1853 1854 return (0); 1855 } 1856 1857 static int 1858 vxlan_ctrl_set_vni(struct vxlan_softc *sc, void *arg) 1859 { 1860 struct ifvxlancmd *cmd; 1861 int error; 1862 1863 cmd = arg; 1864 1865 if (vxlan_check_vni(cmd->vxlcmd_vni) != 0) 1866 return (EINVAL); 1867 1868 VXLAN_WLOCK(sc); 1869 if (vxlan_can_change_config(sc)) { 1870 sc->vxl_vni = cmd->vxlcmd_vni; 1871 error = 0; 1872 } else 1873 error = EBUSY; 1874 VXLAN_WUNLOCK(sc); 1875 1876 return (error); 1877 } 1878 1879 static int 1880 vxlan_ctrl_set_local_addr(struct vxlan_softc *sc, void *arg) 1881 { 1882 struct ifvxlancmd *cmd; 1883 union vxlan_sockaddr *vxlsa; 1884 int error; 1885 1886 cmd = arg; 1887 vxlsa = &cmd->vxlcmd_sa; 1888 1889 if (!VXLAN_SOCKADDR_IS_IPV46(vxlsa)) 1890 return (EINVAL); 1891 if (vxlan_sockaddr_in_multicast(vxlsa) != 0) 1892 return (EINVAL); 1893 if (VXLAN_SOCKADDR_IS_IPV6(vxlsa)) { 1894 error = vxlan_sockaddr_in6_embedscope(vxlsa); 1895 if (error) 1896 return (error); 1897 } 1898 1899 VXLAN_WLOCK(sc); 1900 if (vxlan_can_change_config(sc)) { 1901 vxlan_sockaddr_in_copy(&sc->vxl_src_addr, &vxlsa->sa); 1902 error = 0; 1903 } else 1904 error = EBUSY; 1905 VXLAN_WUNLOCK(sc); 1906 1907 return (error); 1908 } 1909 1910 static int 1911 vxlan_ctrl_set_remote_addr(struct vxlan_softc *sc, void *arg) 1912 { 1913 struct ifvxlancmd *cmd; 1914 union vxlan_sockaddr *vxlsa; 1915 int error; 1916 1917 cmd = arg; 1918 vxlsa = &cmd->vxlcmd_sa; 1919 1920 if (!VXLAN_SOCKADDR_IS_IPV46(vxlsa)) 1921 return (EINVAL); 1922 if (VXLAN_SOCKADDR_IS_IPV6(vxlsa)) { 1923 error = vxlan_sockaddr_in6_embedscope(vxlsa); 1924 if (error) 1925 return (error); 1926 } 1927 1928 VXLAN_WLOCK(sc); 1929 if (vxlan_can_change_config(sc)) { 1930 vxlan_sockaddr_in_copy(&sc->vxl_dst_addr, &vxlsa->sa); 1931 error = 0; 1932 } else 1933 error = EBUSY; 1934 VXLAN_WUNLOCK(sc); 1935 1936 return (error); 1937 } 1938 1939 static int 1940 vxlan_ctrl_set_local_port(struct vxlan_softc *sc, void *arg) 1941 { 1942 struct ifvxlancmd *cmd; 1943 int error; 1944 1945 cmd = arg; 1946 1947 if (cmd->vxlcmd_port == 0) 1948 return (EINVAL); 1949 1950 VXLAN_WLOCK(sc); 1951 if (vxlan_can_change_config(sc)) { 1952 sc->vxl_src_addr.in4.sin_port = htons(cmd->vxlcmd_port); 1953 error = 0; 1954 } else 1955 error = EBUSY; 1956 VXLAN_WUNLOCK(sc); 1957 1958 return (error); 1959 } 1960 1961 static int 1962 vxlan_ctrl_set_remote_port(struct vxlan_softc *sc, void *arg) 1963 { 1964 struct ifvxlancmd *cmd; 1965 int error; 1966 1967 cmd = arg; 1968 1969 if (cmd->vxlcmd_port == 0) 1970 return (EINVAL); 1971 1972 VXLAN_WLOCK(sc); 1973 if (vxlan_can_change_config(sc)) { 1974 sc->vxl_dst_addr.in4.sin_port = htons(cmd->vxlcmd_port); 1975 error = 0; 1976 } else 1977 error = EBUSY; 1978 VXLAN_WUNLOCK(sc); 1979 1980 return (error); 1981 } 1982 1983 static int 1984 vxlan_ctrl_set_port_range(struct vxlan_softc *sc, void *arg) 1985 { 1986 struct ifvxlancmd *cmd; 1987 uint16_t min, max; 1988 int error; 1989 1990 cmd = arg; 1991 min = cmd->vxlcmd_port_min; 1992 max = cmd->vxlcmd_port_max; 1993 1994 if (max < min) 1995 return (EINVAL); 1996 1997 VXLAN_WLOCK(sc); 1998 if (vxlan_can_change_config(sc)) { 1999 sc->vxl_min_port = min; 2000 sc->vxl_max_port = max; 2001 error = 0; 2002 } else 2003 error = EBUSY; 2004 VXLAN_WUNLOCK(sc); 2005 2006 return (error); 2007 } 2008 2009 static int 2010 vxlan_ctrl_set_ftable_timeout(struct vxlan_softc *sc, void *arg) 2011 { 2012 struct ifvxlancmd *cmd; 2013 int error; 2014 2015 cmd = arg; 2016 2017 VXLAN_WLOCK(sc); 2018 if (vxlan_check_ftable_timeout(cmd->vxlcmd_ftable_timeout) == 0) { 2019 sc->vxl_ftable_timeout = cmd->vxlcmd_ftable_timeout; 2020 error = 0; 2021 } else 2022 error = EINVAL; 2023 VXLAN_WUNLOCK(sc); 2024 2025 return (error); 2026 } 2027 2028 static int 2029 vxlan_ctrl_set_ftable_max(struct vxlan_softc *sc, void *arg) 2030 { 2031 struct ifvxlancmd *cmd; 2032 int error; 2033 2034 cmd = arg; 2035 2036 VXLAN_WLOCK(sc); 2037 if (vxlan_check_ftable_max(cmd->vxlcmd_ftable_max) == 0) { 2038 sc->vxl_ftable_max = cmd->vxlcmd_ftable_max; 2039 error = 0; 2040 } else 2041 error = EINVAL; 2042 VXLAN_WUNLOCK(sc); 2043 2044 return (error); 2045 } 2046 2047 static int 2048 vxlan_ctrl_set_multicast_if(struct vxlan_softc * sc, void *arg) 2049 { 2050 struct ifvxlancmd *cmd; 2051 int error; 2052 2053 cmd = arg; 2054 2055 VXLAN_WLOCK(sc); 2056 if (vxlan_can_change_config(sc)) { 2057 strlcpy(sc->vxl_mc_ifname, cmd->vxlcmd_ifname, IFNAMSIZ); 2058 error = 0; 2059 } else 2060 error = EBUSY; 2061 VXLAN_WUNLOCK(sc); 2062 2063 return (error); 2064 } 2065 2066 static int 2067 vxlan_ctrl_set_ttl(struct vxlan_softc *sc, void *arg) 2068 { 2069 struct ifvxlancmd *cmd; 2070 int error; 2071 2072 cmd = arg; 2073 2074 VXLAN_WLOCK(sc); 2075 if (vxlan_check_ttl(cmd->vxlcmd_ttl) == 0) { 2076 sc->vxl_ttl = cmd->vxlcmd_ttl; 2077 if (sc->vxl_im4o != NULL) 2078 sc->vxl_im4o->imo_multicast_ttl = sc->vxl_ttl; 2079 if (sc->vxl_im6o != NULL) 2080 sc->vxl_im6o->im6o_multicast_hlim = sc->vxl_ttl; 2081 error = 0; 2082 } else 2083 error = EINVAL; 2084 VXLAN_WUNLOCK(sc); 2085 2086 return (error); 2087 } 2088 2089 static int 2090 vxlan_ctrl_set_learn(struct vxlan_softc *sc, void *arg) 2091 { 2092 struct ifvxlancmd *cmd; 2093 2094 cmd = arg; 2095 2096 VXLAN_WLOCK(sc); 2097 if (cmd->vxlcmd_flags & VXLAN_CMD_FLAG_LEARN) 2098 sc->vxl_flags |= VXLAN_FLAG_LEARN; 2099 else 2100 sc->vxl_flags &= ~VXLAN_FLAG_LEARN; 2101 VXLAN_WUNLOCK(sc); 2102 2103 return (0); 2104 } 2105 2106 static int 2107 vxlan_ctrl_ftable_entry_add(struct vxlan_softc *sc, void *arg) 2108 { 2109 union vxlan_sockaddr vxlsa; 2110 struct ifvxlancmd *cmd; 2111 struct vxlan_ftable_entry *fe; 2112 int error; 2113 2114 cmd = arg; 2115 vxlsa = cmd->vxlcmd_sa; 2116 2117 if (!VXLAN_SOCKADDR_IS_IPV46(&vxlsa)) 2118 return (EINVAL); 2119 if (vxlan_sockaddr_in_any(&vxlsa) != 0) 2120 return (EINVAL); 2121 if (vxlan_sockaddr_in_multicast(&vxlsa) != 0) 2122 return (EINVAL); 2123 /* BMV: We could support both IPv4 and IPv6 later. */ 2124 if (vxlsa.sa.sa_family != sc->vxl_dst_addr.sa.sa_family) 2125 return (EAFNOSUPPORT); 2126 2127 if (VXLAN_SOCKADDR_IS_IPV6(&vxlsa)) { 2128 error = vxlan_sockaddr_in6_embedscope(&vxlsa); 2129 if (error) 2130 return (error); 2131 } 2132 2133 fe = vxlan_ftable_entry_alloc(); 2134 if (fe == NULL) 2135 return (ENOMEM); 2136 2137 if (vxlsa.in4.sin_port == 0) 2138 vxlsa.in4.sin_port = sc->vxl_dst_addr.in4.sin_port; 2139 2140 vxlan_ftable_entry_init(sc, fe, cmd->vxlcmd_mac, &vxlsa.sa, 2141 VXLAN_FE_FLAG_STATIC); 2142 2143 VXLAN_WLOCK(sc); 2144 error = vxlan_ftable_entry_insert(sc, fe); 2145 VXLAN_WUNLOCK(sc); 2146 2147 if (error) 2148 vxlan_ftable_entry_free(fe); 2149 2150 return (error); 2151 } 2152 2153 static int 2154 vxlan_ctrl_ftable_entry_rem(struct vxlan_softc *sc, void *arg) 2155 { 2156 struct ifvxlancmd *cmd; 2157 struct vxlan_ftable_entry *fe; 2158 int error; 2159 2160 cmd = arg; 2161 2162 VXLAN_WLOCK(sc); 2163 fe = vxlan_ftable_entry_lookup(sc, cmd->vxlcmd_mac); 2164 if (fe != NULL) { 2165 vxlan_ftable_entry_destroy(sc, fe); 2166 error = 0; 2167 } else 2168 error = ENOENT; 2169 VXLAN_WUNLOCK(sc); 2170 2171 return (error); 2172 } 2173 2174 static int 2175 vxlan_ctrl_flush(struct vxlan_softc *sc, void *arg) 2176 { 2177 struct ifvxlancmd *cmd; 2178 int all; 2179 2180 cmd = arg; 2181 all = cmd->vxlcmd_flags & VXLAN_CMD_FLAG_FLUSH_ALL; 2182 2183 VXLAN_WLOCK(sc); 2184 vxlan_ftable_flush(sc, all); 2185 VXLAN_WUNLOCK(sc); 2186 2187 return (0); 2188 } 2189 2190 static int 2191 vxlan_ioctl_drvspec(struct vxlan_softc *sc, struct ifdrv *ifd, int get) 2192 { 2193 const struct vxlan_control *vc; 2194 union { 2195 struct ifvxlancfg cfg; 2196 struct ifvxlancmd cmd; 2197 } args; 2198 int out, error; 2199 2200 if (ifd->ifd_cmd >= vxlan_control_table_size) 2201 return (EINVAL); 2202 2203 bzero(&args, sizeof(args)); 2204 vc = &vxlan_control_table[ifd->ifd_cmd]; 2205 out = (vc->vxlc_flags & VXLAN_CTRL_FLAG_COPYOUT) != 0; 2206 2207 if ((get != 0 && out == 0) || (get == 0 && out != 0)) 2208 return (EINVAL); 2209 2210 if (vc->vxlc_flags & VXLAN_CTRL_FLAG_SUSER) { 2211 error = priv_check(curthread, PRIV_NET_VXLAN); 2212 if (error) 2213 return (error); 2214 } 2215 2216 if (ifd->ifd_len != vc->vxlc_argsize || 2217 ifd->ifd_len > sizeof(args)) 2218 return (EINVAL); 2219 2220 if (vc->vxlc_flags & VXLAN_CTRL_FLAG_COPYIN) { 2221 error = copyin(ifd->ifd_data, &args, ifd->ifd_len); 2222 if (error) 2223 return (error); 2224 } 2225 2226 error = vc->vxlc_func(sc, &args); 2227 if (error) 2228 return (error); 2229 2230 if (vc->vxlc_flags & VXLAN_CTRL_FLAG_COPYOUT) { 2231 error = copyout(&args, ifd->ifd_data, ifd->ifd_len); 2232 if (error) 2233 return (error); 2234 } 2235 2236 return (0); 2237 } 2238 2239 static int 2240 vxlan_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data) 2241 { 2242 struct vxlan_softc *sc; 2243 struct ifreq *ifr; 2244 struct ifdrv *ifd; 2245 int error; 2246 2247 sc = ifp->if_softc; 2248 ifr = (struct ifreq *) data; 2249 ifd = (struct ifdrv *) data; 2250 2251 switch (cmd) { 2252 case SIOCADDMULTI: 2253 case SIOCDELMULTI: 2254 error = 0; 2255 break; 2256 2257 case SIOCGDRVSPEC: 2258 case SIOCSDRVSPEC: 2259 error = vxlan_ioctl_drvspec(sc, ifd, cmd == SIOCGDRVSPEC); 2260 break; 2261 2262 case SIOCSIFFLAGS: 2263 error = vxlan_ioctl_ifflags(sc); 2264 break; 2265 2266 case SIOCSIFMEDIA: 2267 case SIOCGIFMEDIA: 2268 error = ifmedia_ioctl(ifp, ifr, &sc->vxl_media, cmd); 2269 break; 2270 2271 default: 2272 error = ether_ioctl(ifp, cmd, data); 2273 break; 2274 } 2275 2276 return (error); 2277 } 2278 2279 #if defined(INET) || defined(INET6) 2280 static uint16_t 2281 vxlan_pick_source_port(struct vxlan_softc *sc, struct mbuf *m) 2282 { 2283 int range; 2284 uint32_t hash; 2285 2286 range = sc->vxl_max_port - sc->vxl_min_port + 1; 2287 2288 if (M_HASHTYPE_ISHASH(m)) 2289 hash = m->m_pkthdr.flowid; 2290 else 2291 hash = jenkins_hash(m->m_data, ETHER_HDR_LEN, 2292 sc->vxl_port_hash_key); 2293 2294 return (sc->vxl_min_port + (hash % range)); 2295 } 2296 2297 static void 2298 vxlan_encap_header(struct vxlan_softc *sc, struct mbuf *m, int ipoff, 2299 uint16_t srcport, uint16_t dstport) 2300 { 2301 struct vxlanudphdr *hdr; 2302 struct udphdr *udph; 2303 struct vxlan_header *vxh; 2304 int len; 2305 2306 len = m->m_pkthdr.len - ipoff; 2307 MPASS(len >= sizeof(struct vxlanudphdr)); 2308 hdr = mtodo(m, ipoff); 2309 2310 udph = &hdr->vxlh_udp; 2311 udph->uh_sport = srcport; 2312 udph->uh_dport = dstport; 2313 udph->uh_ulen = htons(len); 2314 udph->uh_sum = 0; 2315 2316 vxh = &hdr->vxlh_hdr; 2317 vxh->vxlh_flags = htonl(VXLAN_HDR_FLAGS_VALID_VNI); 2318 vxh->vxlh_vni = htonl(sc->vxl_vni << VXLAN_HDR_VNI_SHIFT); 2319 } 2320 #endif 2321 2322 static int 2323 vxlan_encap4(struct vxlan_softc *sc, const union vxlan_sockaddr *fvxlsa, 2324 struct mbuf *m) 2325 { 2326 #ifdef INET 2327 struct ifnet *ifp; 2328 struct ip *ip; 2329 struct in_addr srcaddr, dstaddr; 2330 uint16_t srcport, dstport; 2331 int len, mcast, error; 2332 2333 ifp = sc->vxl_ifp; 2334 srcaddr = sc->vxl_src_addr.in4.sin_addr; 2335 srcport = vxlan_pick_source_port(sc, m); 2336 dstaddr = fvxlsa->in4.sin_addr; 2337 dstport = fvxlsa->in4.sin_port; 2338 2339 M_PREPEND(m, sizeof(struct ip) + sizeof(struct vxlanudphdr), 2340 M_NOWAIT); 2341 if (m == NULL) { 2342 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1); 2343 return (ENOBUFS); 2344 } 2345 2346 len = m->m_pkthdr.len; 2347 2348 ip = mtod(m, struct ip *); 2349 ip->ip_tos = 0; 2350 ip->ip_len = htons(len); 2351 ip->ip_off = 0; 2352 ip->ip_ttl = sc->vxl_ttl; 2353 ip->ip_p = IPPROTO_UDP; 2354 ip->ip_sum = 0; 2355 ip->ip_src = srcaddr; 2356 ip->ip_dst = dstaddr; 2357 2358 vxlan_encap_header(sc, m, sizeof(struct ip), srcport, dstport); 2359 2360 mcast = (m->m_flags & (M_MCAST | M_BCAST)) ? 1 : 0; 2361 m->m_flags &= ~(M_MCAST | M_BCAST); 2362 2363 error = ip_output(m, NULL, NULL, 0, sc->vxl_im4o, NULL); 2364 if (error == 0) { 2365 if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1); 2366 if_inc_counter(ifp, IFCOUNTER_OBYTES, len); 2367 if (mcast != 0) 2368 if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1); 2369 } else 2370 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1); 2371 2372 return (error); 2373 #else 2374 m_freem(m); 2375 return (ENOTSUP); 2376 #endif 2377 } 2378 2379 static int 2380 vxlan_encap6(struct vxlan_softc *sc, const union vxlan_sockaddr *fvxlsa, 2381 struct mbuf *m) 2382 { 2383 #ifdef INET6 2384 struct ifnet *ifp; 2385 struct ip6_hdr *ip6; 2386 const struct in6_addr *srcaddr, *dstaddr; 2387 uint16_t srcport, dstport; 2388 int len, mcast, error; 2389 2390 ifp = sc->vxl_ifp; 2391 srcaddr = &sc->vxl_src_addr.in6.sin6_addr; 2392 srcport = vxlan_pick_source_port(sc, m); 2393 dstaddr = &fvxlsa->in6.sin6_addr; 2394 dstport = fvxlsa->in6.sin6_port; 2395 2396 M_PREPEND(m, sizeof(struct ip6_hdr) + sizeof(struct vxlanudphdr), 2397 M_NOWAIT); 2398 if (m == NULL) { 2399 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1); 2400 return (ENOBUFS); 2401 } 2402 2403 len = m->m_pkthdr.len; 2404 2405 ip6 = mtod(m, struct ip6_hdr *); 2406 ip6->ip6_flow = 0; /* BMV: Keep in forwarding entry? */ 2407 ip6->ip6_vfc = IPV6_VERSION; 2408 ip6->ip6_plen = 0; 2409 ip6->ip6_nxt = IPPROTO_UDP; 2410 ip6->ip6_hlim = sc->vxl_ttl; 2411 ip6->ip6_src = *srcaddr; 2412 ip6->ip6_dst = *dstaddr; 2413 2414 vxlan_encap_header(sc, m, sizeof(struct ip6_hdr), srcport, dstport); 2415 2416 /* 2417 * XXX BMV We need support for RFC6935 before we can send and 2418 * receive IPv6 UDP packets with a zero checksum. 2419 */ 2420 { 2421 struct udphdr *hdr = mtodo(m, sizeof(struct ip6_hdr)); 2422 hdr->uh_sum = in6_cksum_pseudo(ip6, 2423 m->m_pkthdr.len - sizeof(struct ip6_hdr), IPPROTO_UDP, 0); 2424 m->m_pkthdr.csum_flags = CSUM_UDP_IPV6; 2425 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum); 2426 } 2427 2428 mcast = (m->m_flags & (M_MCAST | M_BCAST)) ? 1 : 0; 2429 m->m_flags &= ~(M_MCAST | M_BCAST); 2430 2431 error = ip6_output(m, NULL, NULL, 0, sc->vxl_im6o, NULL, NULL); 2432 if (error == 0) { 2433 if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1); 2434 if_inc_counter(ifp, IFCOUNTER_OBYTES, len); 2435 if (mcast != 0) 2436 if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1); 2437 } else 2438 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1); 2439 2440 return (error); 2441 #else 2442 m_freem(m); 2443 return (ENOTSUP); 2444 #endif 2445 } 2446 2447 static int 2448 vxlan_transmit(struct ifnet *ifp, struct mbuf *m) 2449 { 2450 struct rm_priotracker tracker; 2451 union vxlan_sockaddr vxlsa; 2452 struct vxlan_softc *sc; 2453 struct vxlan_ftable_entry *fe; 2454 struct ifnet *mcifp; 2455 struct ether_header *eh; 2456 int ipv4, error; 2457 2458 sc = ifp->if_softc; 2459 eh = mtod(m, struct ether_header *); 2460 fe = NULL; 2461 mcifp = NULL; 2462 2463 ETHER_BPF_MTAP(ifp, m); 2464 2465 VXLAN_RLOCK(sc, &tracker); 2466 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) { 2467 VXLAN_RUNLOCK(sc, &tracker); 2468 m_freem(m); 2469 return (ENETDOWN); 2470 } 2471 2472 if ((m->m_flags & (M_BCAST | M_MCAST)) == 0) 2473 fe = vxlan_ftable_entry_lookup(sc, eh->ether_dhost); 2474 if (fe == NULL) 2475 fe = &sc->vxl_default_fe; 2476 vxlan_sockaddr_copy(&vxlsa, &fe->vxlfe_raddr.sa); 2477 2478 ipv4 = VXLAN_SOCKADDR_IS_IPV4(&vxlsa) != 0; 2479 if (vxlan_sockaddr_in_multicast(&vxlsa) != 0) 2480 mcifp = vxlan_multicast_if_ref(sc, ipv4); 2481 2482 VXLAN_ACQUIRE(sc); 2483 VXLAN_RUNLOCK(sc, &tracker); 2484 2485 if (ipv4 != 0) 2486 error = vxlan_encap4(sc, &vxlsa, m); 2487 else 2488 error = vxlan_encap6(sc, &vxlsa, m); 2489 2490 vxlan_release(sc); 2491 if (mcifp != NULL) 2492 if_rele(mcifp); 2493 2494 return (error); 2495 } 2496 2497 static void 2498 vxlan_qflush(struct ifnet *ifp __unused) 2499 { 2500 } 2501 2502 static void 2503 vxlan_rcv_udp_packet(struct mbuf *m, int offset, struct inpcb *inpcb, 2504 const struct sockaddr *srcsa, void *xvso) 2505 { 2506 struct vxlan_socket *vso; 2507 struct vxlan_header *vxh, vxlanhdr; 2508 uint32_t vni; 2509 int error __unused; 2510 2511 M_ASSERTPKTHDR(m); 2512 vso = xvso; 2513 offset += sizeof(struct udphdr); 2514 2515 if (m->m_pkthdr.len < offset + sizeof(struct vxlan_header)) 2516 goto out; 2517 2518 if (__predict_false(m->m_len < offset + sizeof(struct vxlan_header))) { 2519 m_copydata(m, offset, sizeof(struct vxlan_header), 2520 (caddr_t) &vxlanhdr); 2521 vxh = &vxlanhdr; 2522 } else 2523 vxh = mtodo(m, offset); 2524 2525 /* 2526 * Drop if there is a reserved bit set in either the flags or VNI 2527 * fields of the header. This goes against the specification, but 2528 * a bit set may indicate an unsupported new feature. This matches 2529 * the behavior of the Linux implementation. 2530 */ 2531 if (vxh->vxlh_flags != htonl(VXLAN_HDR_FLAGS_VALID_VNI) || 2532 vxh->vxlh_vni & ~htonl(VXLAN_VNI_MASK)) 2533 goto out; 2534 2535 vni = ntohl(vxh->vxlh_vni) >> VXLAN_HDR_VNI_SHIFT; 2536 /* Adjust to the start of the inner Ethernet frame. */ 2537 m_adj(m, offset + sizeof(struct vxlan_header)); 2538 2539 error = vxlan_input(vso, vni, &m, srcsa); 2540 MPASS(error != 0 || m == NULL); 2541 2542 out: 2543 if (m != NULL) 2544 m_freem(m); 2545 } 2546 2547 static int 2548 vxlan_input(struct vxlan_socket *vso, uint32_t vni, struct mbuf **m0, 2549 const struct sockaddr *sa) 2550 { 2551 struct vxlan_softc *sc; 2552 struct ifnet *ifp; 2553 struct mbuf *m; 2554 struct ether_header *eh; 2555 int error; 2556 2557 sc = vxlan_socket_lookup_softc(vso, vni); 2558 if (sc == NULL) 2559 return (ENOENT); 2560 2561 ifp = sc->vxl_ifp; 2562 m = *m0; 2563 eh = mtod(m, struct ether_header *); 2564 2565 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) { 2566 error = ENETDOWN; 2567 goto out; 2568 } else if (ifp == m->m_pkthdr.rcvif) { 2569 /* XXX Does not catch more complex loops. */ 2570 error = EDEADLK; 2571 goto out; 2572 } 2573 2574 if (sc->vxl_flags & VXLAN_FLAG_LEARN) 2575 vxlan_ftable_learn(sc, sa, eh->ether_shost); 2576 2577 m_clrprotoflags(m); 2578 m->m_pkthdr.rcvif = ifp; 2579 M_SETFIB(m, ifp->if_fib); 2580 2581 error = netisr_queue_src(NETISR_ETHER, 0, m); 2582 *m0 = NULL; 2583 2584 out: 2585 vxlan_release(sc); 2586 return (error); 2587 } 2588 2589 static void 2590 vxlan_set_default_config(struct vxlan_softc *sc) 2591 { 2592 2593 sc->vxl_flags |= VXLAN_FLAG_LEARN; 2594 2595 sc->vxl_vni = VXLAN_VNI_MAX; 2596 sc->vxl_ttl = IPDEFTTL; 2597 2598 if (!vxlan_tunable_int(sc, "legacy_port", vxlan_legacy_port)) { 2599 sc->vxl_src_addr.in4.sin_port = htons(VXLAN_PORT); 2600 sc->vxl_dst_addr.in4.sin_port = htons(VXLAN_PORT); 2601 } else { 2602 sc->vxl_src_addr.in4.sin_port = htons(VXLAN_LEGACY_PORT); 2603 sc->vxl_dst_addr.in4.sin_port = htons(VXLAN_LEGACY_PORT); 2604 } 2605 2606 sc->vxl_min_port = V_ipport_firstauto; 2607 sc->vxl_max_port = V_ipport_lastauto; 2608 2609 sc->vxl_ftable_max = VXLAN_FTABLE_MAX; 2610 sc->vxl_ftable_timeout = VXLAN_FTABLE_TIMEOUT; 2611 } 2612 2613 static int 2614 vxlan_set_user_config(struct vxlan_softc *sc, struct ifvxlanparam *vxlp) 2615 { 2616 2617 #ifndef INET 2618 if (vxlp->vxlp_with & (VXLAN_PARAM_WITH_LOCAL_ADDR4 | 2619 VXLAN_PARAM_WITH_REMOTE_ADDR4)) 2620 return (EAFNOSUPPORT); 2621 #endif 2622 2623 #ifndef INET6 2624 if (vxlp->vxlp_with & (VXLAN_PARAM_WITH_LOCAL_ADDR6 | 2625 VXLAN_PARAM_WITH_REMOTE_ADDR6)) 2626 return (EAFNOSUPPORT); 2627 #else 2628 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_LOCAL_ADDR6) { 2629 int error = vxlan_sockaddr_in6_embedscope(&vxlp->vxlp_local_sa); 2630 if (error) 2631 return (error); 2632 } 2633 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_REMOTE_ADDR6) { 2634 int error = vxlan_sockaddr_in6_embedscope( 2635 &vxlp->vxlp_remote_sa); 2636 if (error) 2637 return (error); 2638 } 2639 #endif 2640 2641 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_VNI) { 2642 if (vxlan_check_vni(vxlp->vxlp_vni) == 0) 2643 sc->vxl_vni = vxlp->vxlp_vni; 2644 } 2645 2646 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_LOCAL_ADDR4) { 2647 sc->vxl_src_addr.in4.sin_len = sizeof(struct sockaddr_in); 2648 sc->vxl_src_addr.in4.sin_family = AF_INET; 2649 sc->vxl_src_addr.in4.sin_addr = 2650 vxlp->vxlp_local_sa.in4.sin_addr; 2651 } else if (vxlp->vxlp_with & VXLAN_PARAM_WITH_LOCAL_ADDR6) { 2652 sc->vxl_src_addr.in6.sin6_len = sizeof(struct sockaddr_in6); 2653 sc->vxl_src_addr.in6.sin6_family = AF_INET6; 2654 sc->vxl_src_addr.in6.sin6_addr = 2655 vxlp->vxlp_local_sa.in6.sin6_addr; 2656 } 2657 2658 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_REMOTE_ADDR4) { 2659 sc->vxl_dst_addr.in4.sin_len = sizeof(struct sockaddr_in); 2660 sc->vxl_dst_addr.in4.sin_family = AF_INET; 2661 sc->vxl_dst_addr.in4.sin_addr = 2662 vxlp->vxlp_remote_sa.in4.sin_addr; 2663 } else if (vxlp->vxlp_with & VXLAN_PARAM_WITH_REMOTE_ADDR6) { 2664 sc->vxl_dst_addr.in6.sin6_len = sizeof(struct sockaddr_in6); 2665 sc->vxl_dst_addr.in6.sin6_family = AF_INET6; 2666 sc->vxl_dst_addr.in6.sin6_addr = 2667 vxlp->vxlp_remote_sa.in6.sin6_addr; 2668 } 2669 2670 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_LOCAL_PORT) 2671 sc->vxl_src_addr.in4.sin_port = htons(vxlp->vxlp_local_port); 2672 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_REMOTE_PORT) 2673 sc->vxl_dst_addr.in4.sin_port = htons(vxlp->vxlp_remote_port); 2674 2675 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_PORT_RANGE) { 2676 if (vxlp->vxlp_min_port <= vxlp->vxlp_max_port) { 2677 sc->vxl_min_port = vxlp->vxlp_min_port; 2678 sc->vxl_max_port = vxlp->vxlp_max_port; 2679 } 2680 } 2681 2682 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_MULTICAST_IF) 2683 strlcpy(sc->vxl_mc_ifname, vxlp->vxlp_mc_ifname, IFNAMSIZ); 2684 2685 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_FTABLE_TIMEOUT) { 2686 if (vxlan_check_ftable_timeout(vxlp->vxlp_ftable_timeout) == 0) 2687 sc->vxl_ftable_timeout = vxlp->vxlp_ftable_timeout; 2688 } 2689 2690 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_FTABLE_MAX) { 2691 if (vxlan_check_ftable_max(vxlp->vxlp_ftable_max) == 0) 2692 sc->vxl_ftable_max = vxlp->vxlp_ftable_max; 2693 } 2694 2695 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_TTL) { 2696 if (vxlan_check_ttl(vxlp->vxlp_ttl) == 0) 2697 sc->vxl_ttl = vxlp->vxlp_ttl; 2698 } 2699 2700 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_LEARN) { 2701 if (vxlp->vxlp_learn == 0) 2702 sc->vxl_flags &= ~VXLAN_FLAG_LEARN; 2703 } 2704 2705 return (0); 2706 } 2707 2708 static int 2709 vxlan_clone_create(struct if_clone *ifc, int unit, caddr_t params) 2710 { 2711 struct vxlan_softc *sc; 2712 struct ifnet *ifp; 2713 struct ifvxlanparam vxlp; 2714 int error; 2715 2716 sc = malloc(sizeof(struct vxlan_softc), M_VXLAN, M_WAITOK | M_ZERO); 2717 sc->vxl_unit = unit; 2718 vxlan_set_default_config(sc); 2719 2720 if (params != 0) { 2721 error = copyin(params, &vxlp, sizeof(vxlp)); 2722 if (error) 2723 goto fail; 2724 2725 error = vxlan_set_user_config(sc, &vxlp); 2726 if (error) 2727 goto fail; 2728 } 2729 2730 ifp = if_alloc(IFT_ETHER); 2731 if (ifp == NULL) { 2732 error = ENOSPC; 2733 goto fail; 2734 } 2735 2736 sc->vxl_ifp = ifp; 2737 rm_init(&sc->vxl_lock, "vxlanrm"); 2738 callout_init_rw(&sc->vxl_callout, &sc->vxl_lock, 0); 2739 sc->vxl_port_hash_key = arc4random(); 2740 vxlan_ftable_init(sc); 2741 2742 vxlan_sysctl_setup(sc); 2743 2744 ifp->if_softc = sc; 2745 if_initname(ifp, vxlan_name, unit); 2746 ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST; 2747 ifp->if_init = vxlan_init; 2748 ifp->if_ioctl = vxlan_ioctl; 2749 ifp->if_transmit = vxlan_transmit; 2750 ifp->if_qflush = vxlan_qflush; 2751 ifp->if_capabilities |= IFCAP_LINKSTATE; 2752 ifp->if_capenable |= IFCAP_LINKSTATE; 2753 2754 ifmedia_init(&sc->vxl_media, 0, vxlan_media_change, vxlan_media_status); 2755 ifmedia_add(&sc->vxl_media, IFM_ETHER | IFM_AUTO, 0, NULL); 2756 ifmedia_set(&sc->vxl_media, IFM_ETHER | IFM_AUTO); 2757 2758 vxlan_fakeaddr(sc); 2759 ether_ifattach(ifp, sc->vxl_hwaddr); 2760 2761 ifp->if_baudrate = 0; 2762 ifp->if_hdrlen = 0; 2763 2764 return (0); 2765 2766 fail: 2767 free(sc, M_VXLAN); 2768 return (error); 2769 } 2770 2771 static void 2772 vxlan_clone_destroy(struct ifnet *ifp) 2773 { 2774 struct vxlan_softc *sc; 2775 2776 sc = ifp->if_softc; 2777 2778 vxlan_teardown(sc); 2779 2780 vxlan_ftable_flush(sc, 1); 2781 2782 ether_ifdetach(ifp); 2783 if_free(ifp); 2784 ifmedia_removeall(&sc->vxl_media); 2785 2786 vxlan_ftable_fini(sc); 2787 2788 vxlan_sysctl_destroy(sc); 2789 rm_destroy(&sc->vxl_lock); 2790 free(sc, M_VXLAN); 2791 } 2792 2793 /* BMV: Taken from if_bridge. */ 2794 static uint32_t 2795 vxlan_mac_hash(struct vxlan_softc *sc, const uint8_t *addr) 2796 { 2797 uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->vxl_ftable_hash_key; 2798 2799 b += addr[5] << 8; 2800 b += addr[4]; 2801 a += addr[3] << 24; 2802 a += addr[2] << 16; 2803 a += addr[1] << 8; 2804 a += addr[0]; 2805 2806 /* 2807 * The following hash function is adapted from "Hash Functions" by Bob Jenkins 2808 * ("Algorithm Alley", Dr. Dobbs Journal, September 1997). 2809 */ 2810 #define mix(a, b, c) \ 2811 do { \ 2812 a -= b; a -= c; a ^= (c >> 13); \ 2813 b -= c; b -= a; b ^= (a << 8); \ 2814 c -= a; c -= b; c ^= (b >> 13); \ 2815 a -= b; a -= c; a ^= (c >> 12); \ 2816 b -= c; b -= a; b ^= (a << 16); \ 2817 c -= a; c -= b; c ^= (b >> 5); \ 2818 a -= b; a -= c; a ^= (c >> 3); \ 2819 b -= c; b -= a; b ^= (a << 10); \ 2820 c -= a; c -= b; c ^= (b >> 15); \ 2821 } while (0) 2822 2823 mix(a, b, c); 2824 2825 #undef mix 2826 2827 return (c); 2828 } 2829 2830 static void 2831 vxlan_fakeaddr(struct vxlan_softc *sc) 2832 { 2833 2834 /* 2835 * Generate a non-multicast, locally administered address. 2836 * 2837 * BMV: Should we use the FreeBSD OUI range instead? 2838 */ 2839 arc4rand(sc->vxl_hwaddr, ETHER_ADDR_LEN, 1); 2840 sc->vxl_hwaddr[0] &= ~1; 2841 sc->vxl_hwaddr[0] |= 2; 2842 } 2843 2844 static int 2845 vxlan_media_change(struct ifnet *ifp) 2846 { 2847 2848 /* Ignore. */ 2849 return (0); 2850 } 2851 2852 static void 2853 vxlan_media_status(struct ifnet *ifp, struct ifmediareq *ifmr) 2854 { 2855 2856 ifmr->ifm_status = IFM_ACTIVE | IFM_AVALID; 2857 ifmr->ifm_active = IFM_ETHER | IFM_FDX; 2858 } 2859 2860 static int 2861 vxlan_sockaddr_cmp(const union vxlan_sockaddr *vxladdr, 2862 const struct sockaddr *sa) 2863 { 2864 2865 return (bcmp(&vxladdr->sa, sa, vxladdr->sa.sa_len)); 2866 } 2867 2868 static void 2869 vxlan_sockaddr_copy(union vxlan_sockaddr *vxladdr, 2870 const struct sockaddr *sa) 2871 { 2872 2873 MPASS(sa->sa_family == AF_INET || sa->sa_family == AF_INET6); 2874 bzero(vxladdr, sizeof(*vxladdr)); 2875 2876 if (sa->sa_family == AF_INET) { 2877 vxladdr->in4 = *satoconstsin(sa); 2878 vxladdr->in4.sin_len = sizeof(struct sockaddr_in); 2879 } else if (sa->sa_family == AF_INET6) { 2880 vxladdr->in6 = *satoconstsin6(sa); 2881 vxladdr->in6.sin6_len = sizeof(struct sockaddr_in6); 2882 } 2883 } 2884 2885 static int 2886 vxlan_sockaddr_in_equal(const union vxlan_sockaddr *vxladdr, 2887 const struct sockaddr *sa) 2888 { 2889 int equal; 2890 2891 if (sa->sa_family == AF_INET) { 2892 const struct in_addr *in4 = &satoconstsin(sa)->sin_addr; 2893 equal = in4->s_addr == vxladdr->in4.sin_addr.s_addr; 2894 } else if (sa->sa_family == AF_INET6) { 2895 const struct in6_addr *in6 = &satoconstsin6(sa)->sin6_addr; 2896 equal = IN6_ARE_ADDR_EQUAL(in6, &vxladdr->in6.sin6_addr); 2897 } else 2898 equal = 0; 2899 2900 return (equal); 2901 } 2902 2903 static void 2904 vxlan_sockaddr_in_copy(union vxlan_sockaddr *vxladdr, 2905 const struct sockaddr *sa) 2906 { 2907 2908 MPASS(sa->sa_family == AF_INET || sa->sa_family == AF_INET6); 2909 2910 if (sa->sa_family == AF_INET) { 2911 const struct in_addr *in4 = &satoconstsin(sa)->sin_addr; 2912 vxladdr->in4.sin_family = AF_INET; 2913 vxladdr->in4.sin_len = sizeof(struct sockaddr_in); 2914 vxladdr->in4.sin_addr = *in4; 2915 } else if (sa->sa_family == AF_INET6) { 2916 const struct in6_addr *in6 = &satoconstsin6(sa)->sin6_addr; 2917 vxladdr->in6.sin6_family = AF_INET6; 2918 vxladdr->in6.sin6_len = sizeof(struct sockaddr_in6); 2919 vxladdr->in6.sin6_addr = *in6; 2920 } 2921 } 2922 2923 static int 2924 vxlan_sockaddr_supported(const union vxlan_sockaddr *vxladdr, int unspec) 2925 { 2926 const struct sockaddr *sa; 2927 int supported; 2928 2929 sa = &vxladdr->sa; 2930 supported = 0; 2931 2932 if (sa->sa_family == AF_UNSPEC && unspec != 0) { 2933 supported = 1; 2934 } else if (sa->sa_family == AF_INET) { 2935 #ifdef INET 2936 supported = 1; 2937 #endif 2938 } else if (sa->sa_family == AF_INET6) { 2939 #ifdef INET6 2940 supported = 1; 2941 #endif 2942 } 2943 2944 return (supported); 2945 } 2946 2947 static int 2948 vxlan_sockaddr_in_any(const union vxlan_sockaddr *vxladdr) 2949 { 2950 const struct sockaddr *sa; 2951 int any; 2952 2953 sa = &vxladdr->sa; 2954 2955 if (sa->sa_family == AF_INET) { 2956 const struct in_addr *in4 = &satoconstsin(sa)->sin_addr; 2957 any = in4->s_addr == INADDR_ANY; 2958 } else if (sa->sa_family == AF_INET6) { 2959 const struct in6_addr *in6 = &satoconstsin6(sa)->sin6_addr; 2960 any = IN6_IS_ADDR_UNSPECIFIED(in6); 2961 } else 2962 any = -1; 2963 2964 return (any); 2965 } 2966 2967 static int 2968 vxlan_sockaddr_in_multicast(const union vxlan_sockaddr *vxladdr) 2969 { 2970 const struct sockaddr *sa; 2971 int mc; 2972 2973 sa = &vxladdr->sa; 2974 2975 if (sa->sa_family == AF_INET) { 2976 const struct in_addr *in4 = &satoconstsin(sa)->sin_addr; 2977 mc = IN_MULTICAST(ntohl(in4->s_addr)); 2978 } else if (sa->sa_family == AF_INET6) { 2979 const struct in6_addr *in6 = &satoconstsin6(sa)->sin6_addr; 2980 mc = IN6_IS_ADDR_MULTICAST(in6); 2981 } else 2982 mc = -1; 2983 2984 return (mc); 2985 } 2986 2987 static int 2988 vxlan_sockaddr_in6_embedscope(union vxlan_sockaddr *vxladdr) 2989 { 2990 int error; 2991 2992 MPASS(VXLAN_SOCKADDR_IS_IPV6(vxladdr)); 2993 #ifdef INET6 2994 error = sa6_embedscope(&vxladdr->in6, V_ip6_use_defzone); 2995 #else 2996 error = EAFNOSUPPORT; 2997 #endif 2998 2999 return (error); 3000 } 3001 3002 static int 3003 vxlan_can_change_config(struct vxlan_softc *sc) 3004 { 3005 struct ifnet *ifp; 3006 3007 ifp = sc->vxl_ifp; 3008 VXLAN_LOCK_ASSERT(sc); 3009 3010 if (ifp->if_drv_flags & IFF_DRV_RUNNING) 3011 return (0); 3012 if (sc->vxl_flags & (VXLAN_FLAG_INIT | VXLAN_FLAG_TEARDOWN)) 3013 return (0); 3014 3015 return (1); 3016 } 3017 3018 static int 3019 vxlan_check_vni(uint32_t vni) 3020 { 3021 3022 return (vni >= VXLAN_VNI_MAX); 3023 } 3024 3025 static int 3026 vxlan_check_ttl(int ttl) 3027 { 3028 3029 return (ttl > MAXTTL); 3030 } 3031 3032 static int 3033 vxlan_check_ftable_timeout(uint32_t timeout) 3034 { 3035 3036 return (timeout > VXLAN_FTABLE_MAX_TIMEOUT); 3037 } 3038 3039 static int 3040 vxlan_check_ftable_max(uint32_t max) 3041 { 3042 3043 return (max > VXLAN_FTABLE_MAX); 3044 } 3045 3046 static void 3047 vxlan_sysctl_setup(struct vxlan_softc *sc) 3048 { 3049 struct sysctl_ctx_list *ctx; 3050 struct sysctl_oid *node; 3051 struct vxlan_statistics *stats; 3052 char namebuf[8]; 3053 3054 ctx = &sc->vxl_sysctl_ctx; 3055 stats = &sc->vxl_stats; 3056 snprintf(namebuf, sizeof(namebuf), "%d", sc->vxl_unit); 3057 3058 sysctl_ctx_init(ctx); 3059 sc->vxl_sysctl_node = SYSCTL_ADD_NODE(ctx, 3060 SYSCTL_STATIC_CHILDREN(_net_link_vxlan), OID_AUTO, namebuf, 3061 CTLFLAG_RD, NULL, ""); 3062 3063 node = SYSCTL_ADD_NODE(ctx, SYSCTL_CHILDREN(sc->vxl_sysctl_node), 3064 OID_AUTO, "ftable", CTLFLAG_RD, NULL, ""); 3065 SYSCTL_ADD_UINT(ctx, SYSCTL_CHILDREN(node), OID_AUTO, "count", 3066 CTLFLAG_RD, &sc->vxl_ftable_cnt, 0, 3067 "Number of entries in fowarding table"); 3068 SYSCTL_ADD_UINT(ctx, SYSCTL_CHILDREN(node), OID_AUTO, "max", 3069 CTLFLAG_RD, &sc->vxl_ftable_max, 0, 3070 "Maximum number of entries allowed in fowarding table"); 3071 SYSCTL_ADD_UINT(ctx, SYSCTL_CHILDREN(node), OID_AUTO, "timeout", 3072 CTLFLAG_RD, &sc->vxl_ftable_timeout, 0, 3073 "Number of seconds between prunes of the forwarding table"); 3074 SYSCTL_ADD_PROC(ctx, SYSCTL_CHILDREN(node), OID_AUTO, "dump", 3075 CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_MPSAFE | CTLFLAG_SKIP, 3076 sc, 0, vxlan_ftable_sysctl_dump, "A", 3077 "Dump the forwarding table entries"); 3078 3079 node = SYSCTL_ADD_NODE(ctx, SYSCTL_CHILDREN(sc->vxl_sysctl_node), 3080 OID_AUTO, "stats", CTLFLAG_RD, NULL, ""); 3081 SYSCTL_ADD_UINT(ctx, SYSCTL_CHILDREN(node), OID_AUTO, 3082 "ftable_nospace", CTLFLAG_RD, &stats->ftable_nospace, 0, 3083 "Fowarding table reached maximum entries"); 3084 SYSCTL_ADD_UINT(ctx, SYSCTL_CHILDREN(node), OID_AUTO, 3085 "ftable_lock_upgrade_failed", CTLFLAG_RD, 3086 &stats->ftable_lock_upgrade_failed, 0, 3087 "Forwarding table update required lock upgrade"); 3088 } 3089 3090 static void 3091 vxlan_sysctl_destroy(struct vxlan_softc *sc) 3092 { 3093 3094 sysctl_ctx_free(&sc->vxl_sysctl_ctx); 3095 sc->vxl_sysctl_node = NULL; 3096 } 3097 3098 static int 3099 vxlan_tunable_int(struct vxlan_softc *sc, const char *knob, int def) 3100 { 3101 char path[64]; 3102 3103 snprintf(path, sizeof(path), "net.link.vxlan.%d.%s", 3104 sc->vxl_unit, knob); 3105 TUNABLE_INT_FETCH(path, &def); 3106 3107 return (def); 3108 } 3109 3110 static void 3111 vxlan_ifdetach_event(void *arg __unused, struct ifnet *ifp) 3112 { 3113 struct vxlan_softc_head list; 3114 struct vxlan_socket *vso; 3115 struct vxlan_softc *sc, *tsc; 3116 3117 LIST_INIT(&list); 3118 3119 if (ifp->if_flags & IFF_RENAMING) 3120 return; 3121 if ((ifp->if_flags & IFF_MULTICAST) == 0) 3122 return; 3123 3124 VXLAN_LIST_LOCK(); 3125 LIST_FOREACH(vso, &vxlan_socket_list, vxlso_entry) 3126 vxlan_socket_ifdetach(vso, ifp, &list); 3127 VXLAN_LIST_UNLOCK(); 3128 3129 LIST_FOREACH_SAFE(sc, &list, vxl_ifdetach_list, tsc) { 3130 LIST_REMOVE(sc, vxl_ifdetach_list); 3131 3132 VXLAN_WLOCK(sc); 3133 if (sc->vxl_flags & VXLAN_FLAG_INIT) 3134 vxlan_init_wait(sc); 3135 vxlan_teardown_locked(sc); 3136 } 3137 } 3138 3139 static void 3140 vxlan_load(void) 3141 { 3142 3143 mtx_init(&vxlan_list_mtx, "vxlan list", NULL, MTX_DEF); 3144 LIST_INIT(&vxlan_socket_list); 3145 vxlan_ifdetach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event, 3146 vxlan_ifdetach_event, NULL, EVENTHANDLER_PRI_ANY); 3147 vxlan_cloner = if_clone_simple(vxlan_name, vxlan_clone_create, 3148 vxlan_clone_destroy, 0); 3149 } 3150 3151 static void 3152 vxlan_unload(void) 3153 { 3154 3155 EVENTHANDLER_DEREGISTER(ifnet_departure_event, 3156 vxlan_ifdetach_event_tag); 3157 if_clone_detach(vxlan_cloner); 3158 mtx_destroy(&vxlan_list_mtx); 3159 MPASS(LIST_EMPTY(&vxlan_socket_list)); 3160 } 3161 3162 static int 3163 vxlan_modevent(module_t mod, int type, void *unused) 3164 { 3165 int error; 3166 3167 error = 0; 3168 3169 switch (type) { 3170 case MOD_LOAD: 3171 vxlan_load(); 3172 break; 3173 case MOD_UNLOAD: 3174 vxlan_unload(); 3175 break; 3176 default: 3177 error = ENOTSUP; 3178 break; 3179 } 3180 3181 return (error); 3182 } 3183 3184 static moduledata_t vxlan_mod = { 3185 "if_vxlan", 3186 vxlan_modevent, 3187 0 3188 }; 3189 3190 DECLARE_MODULE(if_vxlan, vxlan_mod, SI_SUB_PSEUDO, SI_ORDER_ANY); 3191 MODULE_VERSION(if_vxlan, 1); 3192