1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1982, 1986, 1990, 1993 5 * The Regents of the University of California. 6 * Copyright (c) 2010-2011 Juniper Networks, Inc. 7 * All rights reserved. 8 * 9 * Portions of this software were developed by Robert N. M. Watson under 10 * contract to Juniper Networks, Inc. 11 * 12 * Redistribution and use in source and binary forms, with or without 13 * modification, are permitted provided that the following conditions 14 * are met: 15 * 1. Redistributions of source code must retain the above copyright 16 * notice, this list of conditions and the following disclaimer. 17 * 2. Redistributions in binary form must reproduce the above copyright 18 * notice, this list of conditions and the following disclaimer in the 19 * documentation and/or other materials provided with the distribution. 20 * 3. Neither the name of the University nor the names of its contributors 21 * may be used to endorse or promote products derived from this software 22 * without specific prior written permission. 23 * 24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 34 * SUCH DAMAGE. 35 * 36 * @(#)in_pcb.h 8.1 (Berkeley) 6/10/93 37 * $FreeBSD$ 38 */ 39 40 #ifndef _NETINET_IN_PCB_H_ 41 #define _NETINET_IN_PCB_H_ 42 43 #include <sys/queue.h> 44 #include <sys/epoch.h> 45 #include <sys/_lock.h> 46 #include <sys/_mutex.h> 47 #include <sys/_rwlock.h> 48 #include <sys/_smr.h> 49 #include <net/route.h> 50 51 #ifdef _KERNEL 52 #include <sys/lock.h> 53 #include <sys/proc.h> 54 #include <sys/rwlock.h> 55 #include <sys/sysctl.h> 56 #include <net/vnet.h> 57 #include <vm/uma.h> 58 #endif 59 #include <sys/ck.h> 60 61 /* 62 * struct inpcb is the common protocol control block structure used in most 63 * IP transport protocols. 64 * 65 * Pointers to local and foreign host table entries, local and foreign socket 66 * numbers, and pointers up (to a socket structure) and down (to a 67 * protocol-specific control block) are stored here. 68 */ 69 CK_LIST_HEAD(inpcbhead, inpcb); 70 CK_LIST_HEAD(inpcbporthead, inpcbport); 71 CK_LIST_HEAD(inpcblbgrouphead, inpcblbgroup); 72 typedef uint64_t inp_gen_t; 73 74 /* 75 * PCB with AF_INET6 null bind'ed laddr can receive AF_INET input packet. 76 * So, AF_INET6 null laddr is also used as AF_INET null laddr, by utilizing 77 * the following structure. This requires padding always be zeroed out, 78 * which is done right after inpcb allocation and stays through its lifetime. 79 */ 80 struct in_addr_4in6 { 81 u_int32_t ia46_pad32[3]; 82 struct in_addr ia46_addr4; 83 }; 84 85 union in_dependaddr { 86 struct in_addr_4in6 id46_addr; 87 struct in6_addr id6_addr; 88 }; 89 90 /* 91 * NOTE: ipv6 addrs should be 64-bit aligned, per RFC 2553. in_conninfo has 92 * some extra padding to accomplish this. 93 * NOTE 2: tcp_syncache.c uses first 5 32-bit words, which identify fport, 94 * lport, faddr to generate hash, so these fields shouldn't be moved. 95 */ 96 struct in_endpoints { 97 u_int16_t ie_fport; /* foreign port */ 98 u_int16_t ie_lport; /* local port */ 99 /* protocol dependent part, local and foreign addr */ 100 union in_dependaddr ie_dependfaddr; /* foreign host table entry */ 101 union in_dependaddr ie_dependladdr; /* local host table entry */ 102 #define ie_faddr ie_dependfaddr.id46_addr.ia46_addr4 103 #define ie_laddr ie_dependladdr.id46_addr.ia46_addr4 104 #define ie6_faddr ie_dependfaddr.id6_addr 105 #define ie6_laddr ie_dependladdr.id6_addr 106 u_int32_t ie6_zoneid; /* scope zone id */ 107 }; 108 109 /* 110 * XXX The defines for inc_* are hacks and should be changed to direct 111 * references. 112 */ 113 struct in_conninfo { 114 u_int8_t inc_flags; 115 u_int8_t inc_len; 116 u_int16_t inc_fibnum; /* XXX was pad, 16 bits is plenty */ 117 /* protocol dependent part */ 118 struct in_endpoints inc_ie; 119 }; 120 121 /* 122 * Flags for inc_flags. 123 */ 124 #define INC_ISIPV6 0x01 125 #define INC_IPV6MINMTU 0x02 126 127 #define inc_fport inc_ie.ie_fport 128 #define inc_lport inc_ie.ie_lport 129 #define inc_faddr inc_ie.ie_faddr 130 #define inc_laddr inc_ie.ie_laddr 131 #define inc6_faddr inc_ie.ie6_faddr 132 #define inc6_laddr inc_ie.ie6_laddr 133 #define inc6_zoneid inc_ie.ie6_zoneid 134 135 #if defined(_KERNEL) || defined(_WANT_INPCB) 136 /* 137 * struct inpcb captures the network layer state for TCP, UDP, and raw IPv4 and 138 * IPv6 sockets. In the case of TCP and UDP, further per-connection state is 139 * hung off of inp_ppcb most of the time. Almost all fields of struct inpcb 140 * are static after creation or protected by a per-inpcb rwlock, inp_lock. 141 * 142 * A inpcb database is indexed by addresses/ports hash as well as list of 143 * all pcbs that belong to a certain proto. Database lookups or list traversals 144 * are be performed inside SMR section. Once desired PCB is found its own 145 * lock is to be obtained and SMR section exited. 146 * 147 * Key: 148 * (c) - Constant after initialization 149 * (e) - Protected by the SMR section 150 * (i) - Protected by the inpcb lock 151 * (p) - Protected by the pcbinfo lock for the inpcb 152 * (h) - Protected by the pcbhash lock for the inpcb 153 * (s) - Protected by another subsystem's locks 154 * (x) - Undefined locking 155 * 156 * A few other notes: 157 * 158 * When a read lock is held, stability of the field is guaranteed; to write 159 * to a field, a write lock must generally be held. 160 * 161 * netinet/netinet6-layer code should not assume that the inp_socket pointer 162 * is safe to dereference without inp_lock being held, there may be 163 * close(2)-related races. 164 * 165 * The inp_vflag field is overloaded, and would otherwise ideally be (c). 166 */ 167 struct icmp6_filter; 168 struct inpcbpolicy; 169 struct m_snd_tag; 170 struct inpcb { 171 /* Cache line #1 (amd64) */ 172 CK_LIST_ENTRY(inpcb) inp_hash_exact; /* hash table linkage */ 173 CK_LIST_ENTRY(inpcb) inp_hash_wild; /* hash table linkage */ 174 struct rwlock inp_lock; 175 /* Cache line #2 (amd64) */ 176 #define inp_start_zero inp_refcount 177 #define inp_zero_size (sizeof(struct inpcb) - \ 178 offsetof(struct inpcb, inp_start_zero)) 179 u_int inp_refcount; /* (i) refcount */ 180 int inp_flags; /* (i) generic IP/datagram flags */ 181 int inp_flags2; /* (i) generic IP/datagram flags #2*/ 182 uint8_t inp_numa_domain; /* numa domain */ 183 void *inp_ppcb; /* (i) pointer to per-protocol pcb */ 184 struct socket *inp_socket; /* (i) back pointer to socket */ 185 struct inpcbinfo *inp_pcbinfo; /* (c) PCB list info */ 186 struct ucred *inp_cred; /* (c) cache of socket cred */ 187 u_int32_t inp_flow; /* (i) IPv6 flow information */ 188 u_char inp_vflag; /* (i) IP version flag (v4/v6) */ 189 u_char inp_ip_ttl; /* (i) time to live proto */ 190 u_char inp_ip_p; /* (c) protocol proto */ 191 u_char inp_ip_minttl; /* (i) minimum TTL or drop */ 192 uint32_t inp_flowid; /* (x) flow id / queue id */ 193 smr_seq_t inp_smr; /* (i) sequence number at disconnect */ 194 struct m_snd_tag *inp_snd_tag; /* (i) send tag for outgoing mbufs */ 195 uint32_t inp_flowtype; /* (x) M_HASHTYPE value */ 196 197 /* Local and foreign ports, local and foreign addr. */ 198 struct in_conninfo inp_inc; /* (i,h) list for PCB's local port */ 199 200 /* MAC and IPSEC policy information. */ 201 struct label *inp_label; /* (i) MAC label */ 202 struct inpcbpolicy *inp_sp; /* (s) for IPSEC */ 203 204 /* Protocol-dependent part; options. */ 205 struct { 206 u_char inp_ip_tos; /* (i) type of service proto */ 207 struct mbuf *inp_options; /* (i) IP options */ 208 struct ip_moptions *inp_moptions; /* (i) mcast options */ 209 }; 210 struct { 211 /* (i) IP options */ 212 struct mbuf *in6p_options; 213 /* (i) IP6 options for outgoing packets */ 214 struct ip6_pktopts *in6p_outputopts; 215 /* (i) IP multicast options */ 216 struct ip6_moptions *in6p_moptions; 217 /* (i) ICMPv6 code type filter */ 218 struct icmp6_filter *in6p_icmp6filt; 219 /* (i) IPV6_CHECKSUM setsockopt */ 220 int in6p_cksum; 221 short in6p_hops; 222 }; 223 CK_LIST_ENTRY(inpcb) inp_portlist; /* (r:e/w:h) port list */ 224 struct inpcbport *inp_phd; /* (r:e/w:h) head of this list */ 225 inp_gen_t inp_gencnt; /* (c) generation count */ 226 void *spare_ptr; /* Spare pointer. */ 227 rt_gen_t inp_rt_cookie; /* generation for route entry */ 228 union { /* cached L3 information */ 229 struct route inp_route; 230 struct route_in6 inp_route6; 231 }; 232 CK_LIST_ENTRY(inpcb) inp_list; /* (r:e/w:p) all PCBs for proto */ 233 }; 234 #endif /* _KERNEL */ 235 236 #define inp_fport inp_inc.inc_fport 237 #define inp_lport inp_inc.inc_lport 238 #define inp_faddr inp_inc.inc_faddr 239 #define inp_laddr inp_inc.inc_laddr 240 241 #define in6p_faddr inp_inc.inc6_faddr 242 #define in6p_laddr inp_inc.inc6_laddr 243 #define in6p_zoneid inp_inc.inc6_zoneid 244 245 #define inp_vnet inp_pcbinfo->ipi_vnet 246 247 /* 248 * The range of the generation count, as used in this implementation, is 9e19. 249 * We would have to create 300 billion connections per second for this number 250 * to roll over in a year. This seems sufficiently unlikely that we simply 251 * don't concern ourselves with that possibility. 252 */ 253 254 /* 255 * Interface exported to userland by various protocols which use inpcbs. Hack 256 * alert -- only define if struct xsocket is in scope. 257 * Fields prefixed with "xi_" are unique to this structure, and the rest 258 * match fields in the struct inpcb, to ease coding and porting. 259 * 260 * Legend: 261 * (s) - used by userland utilities in src 262 * (p) - used by utilities in ports 263 * (3) - is known to be used by third party software not in ports 264 * (n) - no known usage 265 */ 266 #ifdef _SYS_SOCKETVAR_H_ 267 struct xinpcb { 268 ksize_t xi_len; /* length of this structure */ 269 struct xsocket xi_socket; /* (s,p) */ 270 struct in_conninfo inp_inc; /* (s,p) */ 271 uint64_t inp_gencnt; /* (s,p) */ 272 kvaddr_t inp_ppcb; /* (s) netstat(1) */ 273 int64_t inp_spare64[4]; 274 uint32_t inp_flow; /* (s) */ 275 uint32_t inp_flowid; /* (s) */ 276 uint32_t inp_flowtype; /* (s) */ 277 int32_t inp_flags; /* (s,p) */ 278 int32_t inp_flags2; /* (s) */ 279 uint32_t inp_unused; 280 int32_t in6p_cksum; /* (n) */ 281 int32_t inp_spare32[4]; 282 uint16_t in6p_hops; /* (n) */ 283 uint8_t inp_ip_tos; /* (n) */ 284 int8_t pad8; 285 uint8_t inp_vflag; /* (s,p) */ 286 uint8_t inp_ip_ttl; /* (n) */ 287 uint8_t inp_ip_p; /* (n) */ 288 uint8_t inp_ip_minttl; /* (n) */ 289 int8_t inp_spare8[4]; 290 } __aligned(8); 291 292 struct xinpgen { 293 ksize_t xig_len; /* length of this structure */ 294 u_int xig_count; /* number of PCBs at this time */ 295 uint32_t _xig_spare32; 296 inp_gen_t xig_gen; /* generation count at this time */ 297 so_gen_t xig_sogen; /* socket generation count this time */ 298 uint64_t _xig_spare64[4]; 299 } __aligned(8); 300 301 struct sockopt_parameters { 302 struct in_conninfo sop_inc; 303 uint64_t sop_id; 304 int sop_level; 305 int sop_optname; 306 char sop_optval[]; 307 }; 308 309 #ifdef _KERNEL 310 int sysctl_setsockopt(SYSCTL_HANDLER_ARGS, struct inpcbinfo *pcbinfo, 311 int (*ctloutput_set)(struct inpcb *, struct sockopt *)); 312 void in_pcbtoxinpcb(const struct inpcb *, struct xinpcb *); 313 #endif 314 #endif /* _SYS_SOCKETVAR_H_ */ 315 316 #ifdef _KERNEL 317 /* 318 * Per-VNET pcb database for each high-level protocol (UDP, TCP, ...) in both 319 * IPv4 and IPv6. 320 * 321 * The pcbs are protected with SMR section and thus all lists in inpcbinfo 322 * are CK-lists. Locking is required to insert a pcb into database. Two 323 * locks are provided: one for the hash and one for the global list of pcbs, 324 * as well as overall count and generation count. 325 * 326 * Locking key: 327 * 328 * (c) Constant or nearly constant after initialisation 329 * (e) Protected by SMR section 330 * (g) Locked by ipi_lock 331 * (h) Locked by ipi_hash_lock 332 */ 333 struct inpcbinfo { 334 /* 335 * Global lock protecting inpcb list modification 336 */ 337 struct mtx ipi_lock; 338 struct inpcbhead ipi_listhead; /* (r:e/w:g) */ 339 u_int ipi_count; /* (g) */ 340 341 /* 342 * Generation count -- incremented each time a connection is allocated 343 * or freed. 344 */ 345 u_quad_t ipi_gencnt; /* (g) */ 346 347 /* 348 * Fields associated with port lookup and allocation. 349 */ 350 u_short ipi_lastport; /* (h) */ 351 u_short ipi_lastlow; /* (h) */ 352 u_short ipi_lasthi; /* (h) */ 353 354 /* 355 * UMA zone from which inpcbs are allocated for this protocol. 356 */ 357 uma_zone_t ipi_zone; /* (c) */ 358 uma_zone_t ipi_portzone; /* (c) */ 359 smr_t ipi_smr; /* (c) */ 360 361 /* 362 * Global hash of inpcbs, hashed by local and foreign addresses and 363 * port numbers. The "exact" hash holds PCBs connected to a foreign 364 * address, and "wild" holds the rest. 365 */ 366 struct mtx ipi_hash_lock; 367 struct inpcbhead *ipi_hash_exact; /* (r:e/w:h) */ 368 struct inpcbhead *ipi_hash_wild; /* (r:e/w:h) */ 369 u_long ipi_hashmask; /* (c) */ 370 371 /* 372 * Global hash of inpcbs, hashed by only local port number. 373 */ 374 struct inpcbporthead *ipi_porthashbase; /* (h) */ 375 u_long ipi_porthashmask; /* (h) */ 376 377 /* 378 * Load balance groups used for the SO_REUSEPORT_LB option, 379 * hashed by local port. 380 */ 381 struct inpcblbgrouphead *ipi_lbgrouphashbase; /* (r:e/w:h) */ 382 u_long ipi_lbgrouphashmask; /* (h) */ 383 384 /* 385 * Pointer to network stack instance 386 */ 387 struct vnet *ipi_vnet; /* (c) */ 388 }; 389 390 /* 391 * Global allocation storage for each high-level protocol (UDP, TCP, ...). 392 * Each corresponding per-VNET inpcbinfo points into this one. 393 */ 394 struct inpcbstorage { 395 uma_zone_t ips_zone; 396 uma_zone_t ips_portzone; 397 uma_init ips_pcbinit; 398 size_t ips_size; 399 const char * ips_zone_name; 400 const char * ips_portzone_name; 401 const char * ips_infolock_name; 402 const char * ips_hashlock_name; 403 }; 404 405 #define INPCBSTORAGE_DEFINE(prot, ppcb, lname, zname, iname, hname) \ 406 static int \ 407 prot##_inpcb_init(void *mem, int size __unused, int flags __unused) \ 408 { \ 409 struct inpcb *inp = mem; \ 410 \ 411 rw_init_flags(&inp->inp_lock, lname, RW_RECURSE | RW_DUPOK); \ 412 return (0); \ 413 } \ 414 static struct inpcbstorage prot = { \ 415 .ips_size = sizeof(struct ppcb), \ 416 .ips_pcbinit = prot##_inpcb_init, \ 417 .ips_zone_name = zname, \ 418 .ips_portzone_name = zname " ports", \ 419 .ips_infolock_name = iname, \ 420 .ips_hashlock_name = hname, \ 421 }; \ 422 SYSINIT(prot##_inpcbstorage_init, SI_SUB_PROTO_DOMAIN, \ 423 SI_ORDER_SECOND, in_pcbstorage_init, &prot); \ 424 SYSUNINIT(prot##_inpcbstorage_uninit, SI_SUB_PROTO_DOMAIN, \ 425 SI_ORDER_SECOND, in_pcbstorage_destroy, &prot) 426 427 /* 428 * Load balance groups used for the SO_REUSEPORT_LB socket option. Each group 429 * (or unique address:port combination) can be re-used at most 430 * INPCBLBGROUP_SIZMAX (256) times. The inpcbs are stored in il_inp which 431 * is dynamically resized as processes bind/unbind to that specific group. 432 */ 433 struct inpcblbgroup { 434 CK_LIST_ENTRY(inpcblbgroup) il_list; 435 struct epoch_context il_epoch_ctx; 436 struct ucred *il_cred; 437 uint16_t il_lport; /* (c) */ 438 u_char il_vflag; /* (c) */ 439 uint8_t il_numa_domain; 440 uint32_t il_pad2; 441 union in_dependaddr il_dependladdr; /* (c) */ 442 #define il_laddr il_dependladdr.id46_addr.ia46_addr4 443 #define il6_laddr il_dependladdr.id6_addr 444 uint32_t il_inpsiz; /* max count in il_inp[] (h) */ 445 uint32_t il_inpcnt; /* cur count in il_inp[] (h) */ 446 struct inpcb *il_inp[]; /* (h) */ 447 }; 448 449 #define INP_LOCK_DESTROY(inp) rw_destroy(&(inp)->inp_lock) 450 #define INP_RLOCK(inp) rw_rlock(&(inp)->inp_lock) 451 #define INP_WLOCK(inp) rw_wlock(&(inp)->inp_lock) 452 #define INP_TRY_RLOCK(inp) rw_try_rlock(&(inp)->inp_lock) 453 #define INP_TRY_WLOCK(inp) rw_try_wlock(&(inp)->inp_lock) 454 #define INP_RUNLOCK(inp) rw_runlock(&(inp)->inp_lock) 455 #define INP_WUNLOCK(inp) rw_wunlock(&(inp)->inp_lock) 456 #define INP_UNLOCK(inp) rw_unlock(&(inp)->inp_lock) 457 #define INP_TRY_UPGRADE(inp) rw_try_upgrade(&(inp)->inp_lock) 458 #define INP_DOWNGRADE(inp) rw_downgrade(&(inp)->inp_lock) 459 #define INP_WLOCKED(inp) rw_wowned(&(inp)->inp_lock) 460 #define INP_LOCK_ASSERT(inp) rw_assert(&(inp)->inp_lock, RA_LOCKED) 461 #define INP_RLOCK_ASSERT(inp) rw_assert(&(inp)->inp_lock, RA_RLOCKED) 462 #define INP_WLOCK_ASSERT(inp) rw_assert(&(inp)->inp_lock, RA_WLOCKED) 463 #define INP_UNLOCK_ASSERT(inp) rw_assert(&(inp)->inp_lock, RA_UNLOCKED) 464 465 /* 466 * These locking functions are for inpcb consumers outside of sys/netinet, 467 * more specifically, they were added for the benefit of TOE drivers. The 468 * macros are reserved for use by the stack. 469 */ 470 void inp_wlock(struct inpcb *); 471 void inp_wunlock(struct inpcb *); 472 void inp_rlock(struct inpcb *); 473 void inp_runlock(struct inpcb *); 474 475 #ifdef INVARIANT_SUPPORT 476 void inp_lock_assert(struct inpcb *); 477 void inp_unlock_assert(struct inpcb *); 478 #else 479 #define inp_lock_assert(inp) do {} while (0) 480 #define inp_unlock_assert(inp) do {} while (0) 481 #endif 482 483 void inp_apply_all(struct inpcbinfo *, void (*func)(struct inpcb *, void *), 484 void *arg); 485 int inp_ip_tos_get(const struct inpcb *inp); 486 void inp_ip_tos_set(struct inpcb *inp, int val); 487 struct socket * 488 inp_inpcbtosocket(struct inpcb *inp); 489 struct tcpcb * 490 inp_inpcbtotcpcb(struct inpcb *inp); 491 void inp_4tuple_get(struct inpcb *inp, uint32_t *laddr, uint16_t *lp, 492 uint32_t *faddr, uint16_t *fp); 493 int inp_so_options(const struct inpcb *inp); 494 495 #endif /* _KERNEL */ 496 497 #define INP_INFO_WLOCK(ipi) mtx_lock(&(ipi)->ipi_lock) 498 #define INP_INFO_WLOCKED(ipi) mtx_owned(&(ipi)->ipi_lock) 499 #define INP_INFO_WUNLOCK(ipi) mtx_unlock(&(ipi)->ipi_lock) 500 #define INP_INFO_LOCK_ASSERT(ipi) MPASS(SMR_ENTERED((ipi)->ipi_smr) || \ 501 mtx_owned(&(ipi)->ipi_lock)) 502 #define INP_INFO_WLOCK_ASSERT(ipi) mtx_assert(&(ipi)->ipi_lock, MA_OWNED) 503 #define INP_INFO_WUNLOCK_ASSERT(ipi) \ 504 mtx_assert(&(ipi)->ipi_lock, MA_NOTOWNED) 505 506 #define INP_HASH_WLOCK(ipi) mtx_lock(&(ipi)->ipi_hash_lock) 507 #define INP_HASH_WUNLOCK(ipi) mtx_unlock(&(ipi)->ipi_hash_lock) 508 #define INP_HASH_LOCK_ASSERT(ipi) MPASS(SMR_ENTERED((ipi)->ipi_smr) || \ 509 mtx_owned(&(ipi)->ipi_hash_lock)) 510 #define INP_HASH_WLOCK_ASSERT(ipi) mtx_assert(&(ipi)->ipi_hash_lock, \ 511 MA_OWNED) 512 513 /* 514 * Wildcard matching hash is not just a microoptimisation! The hash for 515 * wildcard IPv4 and wildcard IPv6 must be the same, otherwise AF_INET6 516 * wildcard bound pcb won't be able to receive AF_INET connections, while: 517 * jenkins_hash(&zeroes, 1, s) != jenkins_hash(&zeroes, 4, s) 518 * See also comment above struct in_addr_4in6. 519 */ 520 #define IN_ADDR_JHASH32(addr) \ 521 ((addr)->s_addr == INADDR_ANY ? V_in_pcbhashseed : \ 522 jenkins_hash32((&(addr)->s_addr), 1, V_in_pcbhashseed)) 523 #define IN6_ADDR_JHASH32(addr) \ 524 (memcmp((addr), &in6addr_any, sizeof(in6addr_any)) == 0 ? \ 525 V_in_pcbhashseed : \ 526 jenkins_hash32((addr)->__u6_addr.__u6_addr32, \ 527 nitems((addr)->__u6_addr.__u6_addr32), V_in_pcbhashseed)) 528 529 #define INP_PCBHASH(faddr, lport, fport, mask) \ 530 ((IN_ADDR_JHASH32(faddr) ^ ntohs((lport) ^ (fport))) & (mask)) 531 #define INP6_PCBHASH(faddr, lport, fport, mask) \ 532 ((IN6_ADDR_JHASH32(faddr) ^ ntohs((lport) ^ (fport))) & (mask)) 533 534 #define INP_PCBHASH_WILD(lport, mask) \ 535 ((V_in_pcbhashseed ^ ntohs(lport)) & (mask)) 536 537 #define INP_PCBLBGROUP_PKTHASH(faddr, lport, fport) \ 538 (IN_ADDR_JHASH32(faddr) ^ ntohs((lport) ^ (fport))) 539 #define INP6_PCBLBGROUP_PKTHASH(faddr, lport, fport) \ 540 (IN6_ADDR_JHASH32(faddr) ^ ntohs((lport) ^ (fport))) 541 542 #define INP_PCBPORTHASH(lport, mask) (ntohs((lport)) & (mask)) 543 544 /* 545 * Flags for inp_vflags -- historically version flags only 546 */ 547 #define INP_IPV4 0x1 548 #define INP_IPV6 0x2 549 #define INP_IPV6PROTO 0x4 /* opened under IPv6 protocol */ 550 551 /* 552 * Flags for inp_flags. 553 */ 554 #define INP_RECVOPTS 0x00000001 /* receive incoming IP options */ 555 #define INP_RECVRETOPTS 0x00000002 /* receive IP options for reply */ 556 #define INP_RECVDSTADDR 0x00000004 /* receive IP dst address */ 557 #define INP_HDRINCL 0x00000008 /* user supplies entire IP header */ 558 #define INP_HIGHPORT 0x00000010 /* user wants "high" port binding */ 559 #define INP_LOWPORT 0x00000020 /* user wants "low" port binding */ 560 #define INP_ANONPORT 0x00000040 /* read by netstat(1) */ 561 #define INP_RECVIF 0x00000080 /* receive incoming interface */ 562 #define INP_MTUDISC 0x00000100 /* user can do MTU discovery */ 563 /* INP_FREED 0x00000200 private to in_pcb.c */ 564 #define INP_RECVTTL 0x00000400 /* receive incoming IP TTL */ 565 #define INP_DONTFRAG 0x00000800 /* don't fragment packet */ 566 #define INP_BINDANY 0x00001000 /* allow bind to any address */ 567 #define INP_INHASHLIST 0x00002000 /* in_pcbinshash() has been called */ 568 #define INP_RECVTOS 0x00004000 /* receive incoming IP TOS */ 569 #define IN6P_IPV6_V6ONLY 0x00008000 /* restrict AF_INET6 socket for v6 */ 570 #define IN6P_PKTINFO 0x00010000 /* receive IP6 dst and I/F */ 571 #define IN6P_HOPLIMIT 0x00020000 /* receive hoplimit */ 572 #define IN6P_HOPOPTS 0x00040000 /* receive hop-by-hop options */ 573 #define IN6P_DSTOPTS 0x00080000 /* receive dst options after rthdr */ 574 #define IN6P_RTHDR 0x00100000 /* receive routing header */ 575 #define IN6P_RTHDRDSTOPTS 0x00200000 /* receive dstoptions before rthdr */ 576 #define IN6P_TCLASS 0x00400000 /* receive traffic class value */ 577 #define IN6P_AUTOFLOWLABEL 0x00800000 /* attach flowlabel automatically */ 578 #define INP_ONESBCAST 0x02000000 /* send all-ones broadcast */ 579 #define INP_DROPPED 0x04000000 /* protocol drop flag */ 580 #define INP_SOCKREF 0x08000000 /* strong socket reference */ 581 #define INP_RESERVED_0 0x10000000 /* reserved field */ 582 #define INP_RESERVED_1 0x20000000 /* reserved field */ 583 #define IN6P_RFC2292 0x40000000 /* used RFC2292 API on the socket */ 584 #define IN6P_MTU 0x80000000 /* receive path MTU */ 585 586 #define INP_CONTROLOPTS (INP_RECVOPTS|INP_RECVRETOPTS|INP_RECVDSTADDR|\ 587 INP_RECVIF|INP_RECVTTL|INP_RECVTOS|\ 588 IN6P_PKTINFO|IN6P_HOPLIMIT|IN6P_HOPOPTS|\ 589 IN6P_DSTOPTS|IN6P_RTHDR|IN6P_RTHDRDSTOPTS|\ 590 IN6P_TCLASS|IN6P_AUTOFLOWLABEL|IN6P_RFC2292|\ 591 IN6P_MTU) 592 593 /* 594 * Flags for inp_flags2. 595 */ 596 /* 0x00000001 */ 597 /* 0x00000002 */ 598 /* 0x00000004 */ 599 #define INP_REUSEPORT 0x00000008 /* SO_REUSEPORT option is set */ 600 /* 0x00000010 */ 601 #define INP_REUSEADDR 0x00000020 /* SO_REUSEADDR option is set */ 602 /* 0x00000040 */ 603 /* 0x00000080 */ 604 #define INP_RECVFLOWID 0x00000100 /* populate recv datagram with flow info */ 605 #define INP_RECVRSSBUCKETID 0x00000200 /* populate recv datagram with bucket id */ 606 #define INP_RATE_LIMIT_CHANGED 0x00000400 /* rate limit needs attention */ 607 #define INP_ORIGDSTADDR 0x00000800 /* receive IP dst address/port */ 608 /* 0x00001000 */ 609 #define INP_REUSEPORT_LB 0x00002000 /* SO_REUSEPORT_LB option is set */ 610 /* 0x00004000 */ 611 /* 0x00008000 */ 612 /* 0x00010000 */ 613 #define INP_2PCP_SET 0x00020000 /* If the Eth PCP should be set explicitly */ 614 #define INP_2PCP_BIT0 0x00040000 /* Eth PCP Bit 0 */ 615 #define INP_2PCP_BIT1 0x00080000 /* Eth PCP Bit 1 */ 616 #define INP_2PCP_BIT2 0x00100000 /* Eth PCP Bit 2 */ 617 #define INP_2PCP_BASE INP_2PCP_BIT0 618 #define INP_2PCP_MASK (INP_2PCP_BIT0 | INP_2PCP_BIT1 | INP_2PCP_BIT2) 619 #define INP_2PCP_SHIFT 18 /* shift PCP field in/out of inp_flags2 */ 620 621 /* 622 * Flags passed to in_pcblookup*(), inp_smr_lock() and inp_next(). 623 */ 624 typedef enum { 625 INPLOOKUP_WILDCARD = 0x00000001, /* Allow wildcard sockets. */ 626 INPLOOKUP_RLOCKPCB = 0x00000002, /* Return inpcb read-locked. */ 627 INPLOOKUP_WLOCKPCB = 0x00000004, /* Return inpcb write-locked. */ 628 } inp_lookup_t; 629 630 #define INPLOOKUP_MASK (INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB | \ 631 INPLOOKUP_WLOCKPCB) 632 #define INPLOOKUP_LOCKMASK (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB) 633 634 #define sotoinpcb(so) ((struct inpcb *)(so)->so_pcb) 635 636 #define INP_SOCKAF(so) so->so_proto->pr_domain->dom_family 637 638 #define INP_CHECK_SOCKAF(so, af) (INP_SOCKAF(so) == af) 639 640 #ifdef _KERNEL 641 VNET_DECLARE(int, ipport_reservedhigh); 642 VNET_DECLARE(int, ipport_reservedlow); 643 VNET_DECLARE(int, ipport_lowfirstauto); 644 VNET_DECLARE(int, ipport_lowlastauto); 645 VNET_DECLARE(int, ipport_firstauto); 646 VNET_DECLARE(int, ipport_lastauto); 647 VNET_DECLARE(int, ipport_hifirstauto); 648 VNET_DECLARE(int, ipport_hilastauto); 649 VNET_DECLARE(int, ipport_randomized); 650 651 #define V_ipport_reservedhigh VNET(ipport_reservedhigh) 652 #define V_ipport_reservedlow VNET(ipport_reservedlow) 653 #define V_ipport_lowfirstauto VNET(ipport_lowfirstauto) 654 #define V_ipport_lowlastauto VNET(ipport_lowlastauto) 655 #define V_ipport_firstauto VNET(ipport_firstauto) 656 #define V_ipport_lastauto VNET(ipport_lastauto) 657 #define V_ipport_hifirstauto VNET(ipport_hifirstauto) 658 #define V_ipport_hilastauto VNET(ipport_hilastauto) 659 #define V_ipport_randomized VNET(ipport_randomized) 660 661 void in_pcbinfo_init(struct inpcbinfo *, struct inpcbstorage *, 662 u_int, u_int); 663 void in_pcbinfo_destroy(struct inpcbinfo *); 664 void in_pcbstorage_init(void *); 665 void in_pcbstorage_destroy(void *); 666 667 void in_pcbpurgeif0(struct inpcbinfo *, struct ifnet *); 668 int in_pcballoc(struct socket *, struct inpcbinfo *); 669 int in_pcbbind(struct inpcb *, struct sockaddr_in *, struct ucred *); 670 int in_pcbbind_setup(struct inpcb *, struct sockaddr_in *, in_addr_t *, 671 u_short *, struct ucred *); 672 int in_pcbconnect(struct inpcb *, struct sockaddr_in *, struct ucred *, 673 bool); 674 int in_pcbconnect_setup(struct inpcb *, struct sockaddr_in *, in_addr_t *, 675 u_short *, in_addr_t *, u_short *, struct ucred *); 676 void in_pcbdetach(struct inpcb *); 677 void in_pcbdisconnect(struct inpcb *); 678 void in_pcbdrop(struct inpcb *); 679 void in_pcbfree(struct inpcb *); 680 int in_pcbinshash(struct inpcb *); 681 int in_pcbladdr(struct inpcb *, struct in_addr *, struct in_addr *, 682 struct ucred *); 683 int in_pcblbgroup_numa(struct inpcb *, int arg); 684 struct inpcb * 685 in_pcblookup(struct inpcbinfo *, struct in_addr, u_int, 686 struct in_addr, u_int, int, struct ifnet *); 687 struct inpcb * 688 in_pcblookup_mbuf(struct inpcbinfo *, struct in_addr, u_int, 689 struct in_addr, u_int, int, struct ifnet *, struct mbuf *); 690 void in_pcbnotifyall(struct inpcbinfo *pcbinfo, struct in_addr, 691 int, struct inpcb *(*)(struct inpcb *, int)); 692 void in_pcbref(struct inpcb *); 693 void in_pcbrehash(struct inpcb *); 694 void in_pcbremhash_locked(struct inpcb *); 695 bool in_pcbrele(struct inpcb *, inp_lookup_t); 696 bool in_pcbrele_rlocked(struct inpcb *); 697 bool in_pcbrele_wlocked(struct inpcb *); 698 699 typedef bool inp_match_t(const struct inpcb *, void *); 700 struct inpcb_iterator { 701 const struct inpcbinfo *ipi; 702 struct inpcb *inp; 703 inp_match_t *match; 704 void *ctx; 705 int hash; 706 #define INP_ALL_LIST -1 707 const inp_lookup_t lock; 708 }; 709 710 /* Note: sparse initializers guarantee .inp = NULL. */ 711 #define INP_ITERATOR(_ipi, _lock, _match, _ctx) \ 712 { \ 713 .ipi = (_ipi), \ 714 .lock = (_lock), \ 715 .hash = INP_ALL_LIST, \ 716 .match = (_match), \ 717 .ctx = (_ctx), \ 718 } 719 #define INP_ALL_ITERATOR(_ipi, _lock) \ 720 { \ 721 .ipi = (_ipi), \ 722 .lock = (_lock), \ 723 .hash = INP_ALL_LIST, \ 724 } 725 726 struct inpcb *inp_next(struct inpcb_iterator *); 727 void in_losing(struct inpcb *); 728 void in_pcbsetsolabel(struct socket *so); 729 int in_getpeeraddr(struct socket *so, struct sockaddr **nam); 730 int in_getsockaddr(struct socket *so, struct sockaddr **nam); 731 struct sockaddr * 732 in_sockaddr(in_port_t port, struct in_addr *addr); 733 void in_pcbsosetlabel(struct socket *so); 734 #ifdef RATELIMIT 735 int 736 in_pcboutput_txrtlmt_locked(struct inpcb *, struct ifnet *, 737 struct mbuf *, uint32_t); 738 int in_pcbattach_txrtlmt(struct inpcb *, struct ifnet *, uint32_t, uint32_t, 739 uint32_t, struct m_snd_tag **); 740 void in_pcbdetach_txrtlmt(struct inpcb *); 741 void in_pcbdetach_tag(struct m_snd_tag *); 742 int in_pcbmodify_txrtlmt(struct inpcb *, uint32_t); 743 int in_pcbquery_txrtlmt(struct inpcb *, uint32_t *); 744 int in_pcbquery_txrlevel(struct inpcb *, uint32_t *); 745 void in_pcboutput_txrtlmt(struct inpcb *, struct ifnet *, struct mbuf *); 746 void in_pcboutput_eagain(struct inpcb *); 747 #endif 748 #endif /* _KERNEL */ 749 750 #endif /* !_NETINET_IN_PCB_H_ */ 751