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 <net/route.h> 49 50 #ifdef _KERNEL 51 #include <sys/lock.h> 52 #include <sys/proc.h> 53 #include <sys/rwlock.h> 54 #include <sys/smr.h> 55 #include <net/vnet.h> 56 #include <vm/uma.h> 57 #endif 58 #include <sys/ck.h> 59 60 /* 61 * struct inpcb is the common protocol control block structure used in most 62 * IP transport protocols. 63 * 64 * Pointers to local and foreign host table entries, local and foreign socket 65 * numbers, and pointers up (to a socket structure) and down (to a 66 * protocol-specific control block) are stored here. 67 */ 68 CK_LIST_HEAD(inpcbhead, inpcb); 69 CK_LIST_HEAD(inpcbporthead, inpcbport); 70 CK_LIST_HEAD(inpcblbgrouphead, inpcblbgroup); 71 typedef uint64_t inp_gen_t; 72 73 /* 74 * PCB with AF_INET6 null bind'ed laddr can receive AF_INET input packet. 75 * So, AF_INET6 null laddr is also used as AF_INET null laddr, by utilizing 76 * the following structure. 77 */ 78 struct in_addr_4in6 { 79 u_int32_t ia46_pad32[3]; 80 struct in_addr ia46_addr4; 81 }; 82 83 union in_dependaddr { 84 struct in_addr_4in6 id46_addr; 85 struct in6_addr id6_addr; 86 }; 87 88 /* 89 * NOTE: ipv6 addrs should be 64-bit aligned, per RFC 2553. in_conninfo has 90 * some extra padding to accomplish this. 91 * NOTE 2: tcp_syncache.c uses first 5 32-bit words, which identify fport, 92 * lport, faddr to generate hash, so these fields shouldn't be moved. 93 */ 94 struct in_endpoints { 95 u_int16_t ie_fport; /* foreign port */ 96 u_int16_t ie_lport; /* local port */ 97 /* protocol dependent part, local and foreign addr */ 98 union in_dependaddr ie_dependfaddr; /* foreign host table entry */ 99 union in_dependaddr ie_dependladdr; /* local host table entry */ 100 #define ie_faddr ie_dependfaddr.id46_addr.ia46_addr4 101 #define ie_laddr ie_dependladdr.id46_addr.ia46_addr4 102 #define ie6_faddr ie_dependfaddr.id6_addr 103 #define ie6_laddr ie_dependladdr.id6_addr 104 u_int32_t ie6_zoneid; /* scope zone id */ 105 }; 106 107 /* 108 * XXX The defines for inc_* are hacks and should be changed to direct 109 * references. 110 */ 111 struct in_conninfo { 112 u_int8_t inc_flags; 113 u_int8_t inc_len; 114 u_int16_t inc_fibnum; /* XXX was pad, 16 bits is plenty */ 115 /* protocol dependent part */ 116 struct in_endpoints inc_ie; 117 }; 118 119 /* 120 * Flags for inc_flags. 121 */ 122 #define INC_ISIPV6 0x01 123 #define INC_IPV6MINMTU 0x02 124 125 #define inc_fport inc_ie.ie_fport 126 #define inc_lport inc_ie.ie_lport 127 #define inc_faddr inc_ie.ie_faddr 128 #define inc_laddr inc_ie.ie_laddr 129 #define inc6_faddr inc_ie.ie6_faddr 130 #define inc6_laddr inc_ie.ie6_laddr 131 #define inc6_zoneid inc_ie.ie6_zoneid 132 133 #if defined(_KERNEL) || defined(_WANT_INPCB) 134 /* 135 * struct inpcb captures the network layer state for TCP, UDP, and raw IPv4 and 136 * IPv6 sockets. In the case of TCP and UDP, further per-connection state is 137 * hung off of inp_ppcb most of the time. Almost all fields of struct inpcb 138 * are static after creation or protected by a per-inpcb rwlock, inp_lock. 139 * 140 * A inpcb database is indexed by addresses/ports hash as well as list of 141 * all pcbs that belong to a certain proto. Database lookups or list traversals 142 * are be performed inside SMR section. Once desired PCB is found its own 143 * lock is to be obtained and SMR section exited. 144 * 145 * Key: 146 * (b) - Protected by the hpts lock. 147 * (c) - Constant after initialization 148 * (e) - Protected by the SMR section 149 * (i) - Protected by the inpcb lock 150 * (p) - Protected by the pcbinfo lock for the inpcb 151 * (h) - Protected by the pcbhash lock for the inpcb 152 * (s) - Protected by another subsystem's locks 153 * (x) - Undefined locking 154 * 155 * Notes on the tcp_hpts: 156 * 157 * First Hpts lock order is 158 * 1) INP_WLOCK() 159 * 2) HPTS_LOCK() i.e. hpts->pmtx 160 * 161 * To insert a TCB on the hpts you *must* be holding the INP_WLOCK(). 162 * You may check the inp->inp_in_hpts flag without the hpts lock. 163 * The hpts is the only one that will clear this flag holding 164 * only the hpts lock. This means that in your tcp_output() 165 * routine when you test for the inp_in_hpts flag to be 1 166 * it may be transitioning to 0 (by the hpts). 167 * That's ok since that will just mean an extra call to tcp_output 168 * that most likely will find the call you executed 169 * (when the mis-match occured) will have put the TCB back 170 * on the hpts and it will return. If your 171 * call did not add the inp back to the hpts then you will either 172 * over-send or the cwnd will block you from sending more. 173 * 174 * Note you should also be holding the INP_WLOCK() when you 175 * call the remove from the hpts as well. Though usually 176 * you are either doing this from a timer, where you need and have 177 * the INP_WLOCK() or from destroying your TCB where again 178 * you should already have the INP_WLOCK(). 179 * 180 * The inp_hpts_cpu, inp_hpts_cpu_set, inp_input_cpu and 181 * inp_input_cpu_set fields are controlled completely by 182 * the hpts. Do not ever set these. The inp_hpts_cpu_set 183 * and inp_input_cpu_set fields indicate if the hpts has 184 * setup the respective cpu field. It is advised if this 185 * field is 0, to enqueue the packet with the appropriate 186 * hpts_immediate() call. If the _set field is 1, then 187 * you may compare the inp_*_cpu field to the curcpu and 188 * may want to again insert onto the hpts if these fields 189 * are not equal (i.e. you are not on the expected CPU). 190 * 191 * A note on inp_hpts_calls and inp_input_calls, these 192 * flags are set when the hpts calls either the output 193 * or do_segment routines respectively. If the routine 194 * being called wants to use this, then it needs to 195 * clear the flag before returning. The hpts will not 196 * clear the flag. The flags can be used to tell if 197 * the hpts is the function calling the respective 198 * routine. 199 * 200 * A few other notes: 201 * 202 * When a read lock is held, stability of the field is guaranteed; to write 203 * to a field, a write lock must generally be held. 204 * 205 * netinet/netinet6-layer code should not assume that the inp_socket pointer 206 * is safe to dereference without inp_lock being held, even for protocols 207 * other than TCP (where the inpcb persists during TIMEWAIT even after the 208 * socket has been freed), or there may be close(2)-related races. 209 * 210 * The inp_vflag field is overloaded, and would otherwise ideally be (c). 211 */ 212 struct icmp6_filter; 213 struct inpcbpolicy; 214 struct m_snd_tag; 215 struct inpcb { 216 /* Cache line #1 (amd64) */ 217 CK_LIST_ENTRY(inpcb) inp_hash; /* (w:h/r:e) hash list */ 218 struct rwlock inp_lock; 219 /* Cache line #2 (amd64) */ 220 #define inp_start_zero inp_hpts 221 #define inp_zero_size (sizeof(struct inpcb) - \ 222 offsetof(struct inpcb, inp_start_zero)) 223 TAILQ_ENTRY(inpcb) inp_hpts; /* pacing out queue next lock(b) */ 224 uint32_t inp_hpts_gencnt; /* XXXGL */ 225 uint32_t inp_hpts_request; /* Current hpts request, zero if 226 * fits in the pacing window (i&b). */ 227 /* 228 * Note the next fields are protected by a 229 * different lock (hpts-lock). This means that 230 * they must correspond in size to the smallest 231 * protectable bit field (uint8_t on x86, and 232 * other platfomrs potentially uint32_t?). Also 233 * since CPU switches can occur at different times the two 234 * fields can *not* be collapsed into a signal bit field. 235 */ 236 #if defined(__amd64__) || defined(__i386__) 237 uint8_t inp_in_hpts; /* on output hpts (lock b) */ 238 uint8_t inp_in_dropq; /* on input hpts (lock b) */ 239 #else 240 uint32_t inp_in_hpts; /* on output hpts (lock b) */ 241 uint32_t inp_in_dropq; /* on input hpts (lock b) */ 242 #endif 243 volatile uint16_t inp_hpts_cpu; /* Lock (i) */ 244 volatile uint16_t inp_irq_cpu; /* Set by LRO in behalf of or the driver */ 245 u_int inp_refcount; /* (i) refcount */ 246 int inp_flags; /* (i) generic IP/datagram flags */ 247 int inp_flags2; /* (i) generic IP/datagram flags #2*/ 248 uint16_t inp_dropq_cpu; /* Lock (i) */ 249 uint8_t inp_hpts_cpu_set :1, /* on output hpts (i) */ 250 inp_dropq_cpu_set : 1, /* on input hpts (i) */ 251 inp_hpts_calls :1, /* (i) from output hpts */ 252 inp_irq_cpu_set :1, /* (i) from LRO/Driver */ 253 inp_spare_bits2 : 3; 254 uint8_t inp_numa_domain; /* numa domain */ 255 void *inp_ppcb; /* (i) pointer to per-protocol pcb */ 256 struct socket *inp_socket; /* (i) back pointer to socket */ 257 int32_t inp_hptsslot; /* Hpts wheel slot this tcb is Lock(i&b) */ 258 uint32_t inp_hpts_drop_reas; /* reason we are dropping the PCB (lock i&b) */ 259 uint32_t inp_dropq_gencnt; 260 TAILQ_ENTRY(inpcb) inp_dropq; /* hpts drop queue next lock(b) */ 261 struct inpcbinfo *inp_pcbinfo; /* (c) PCB list info */ 262 struct ucred *inp_cred; /* (c) cache of socket cred */ 263 u_int32_t inp_flow; /* (i) IPv6 flow information */ 264 u_char inp_vflag; /* (i) IP version flag (v4/v6) */ 265 u_char inp_ip_ttl; /* (i) time to live proto */ 266 u_char inp_ip_p; /* (c) protocol proto */ 267 u_char inp_ip_minttl; /* (i) minimum TTL or drop */ 268 uint32_t inp_flowid; /* (x) flow id / queue id */ 269 struct m_snd_tag *inp_snd_tag; /* (i) send tag for outgoing mbufs */ 270 uint32_t inp_flowtype; /* (x) M_HASHTYPE value */ 271 uint32_t inp_rss_listen_bucket; /* (x) overridden RSS listen bucket */ 272 273 /* Local and foreign ports, local and foreign addr. */ 274 struct in_conninfo inp_inc; /* (i) list for PCB's local port */ 275 276 /* MAC and IPSEC policy information. */ 277 struct label *inp_label; /* (i) MAC label */ 278 struct inpcbpolicy *inp_sp; /* (s) for IPSEC */ 279 280 /* Protocol-dependent part; options. */ 281 struct { 282 u_char inp_ip_tos; /* (i) type of service proto */ 283 struct mbuf *inp_options; /* (i) IP options */ 284 struct ip_moptions *inp_moptions; /* (i) mcast options */ 285 }; 286 struct { 287 /* (i) IP options */ 288 struct mbuf *in6p_options; 289 /* (i) IP6 options for outgoing packets */ 290 struct ip6_pktopts *in6p_outputopts; 291 /* (i) IP multicast options */ 292 struct ip6_moptions *in6p_moptions; 293 /* (i) ICMPv6 code type filter */ 294 struct icmp6_filter *in6p_icmp6filt; 295 /* (i) IPV6_CHECKSUM setsockopt */ 296 int in6p_cksum; 297 short in6p_hops; 298 }; 299 CK_LIST_ENTRY(inpcb) inp_portlist; /* (r:e/w:h) port list */ 300 struct inpcbport *inp_phd; /* (r:e/w:h) head of this list */ 301 inp_gen_t inp_gencnt; /* (c) generation count */ 302 void *spare_ptr; /* Spare pointer. */ 303 rt_gen_t inp_rt_cookie; /* generation for route entry */ 304 union { /* cached L3 information */ 305 struct route inp_route; 306 struct route_in6 inp_route6; 307 }; 308 CK_LIST_ENTRY(inpcb) inp_list; /* (r:e/w:p) all PCBs for proto */ 309 }; 310 #endif /* _KERNEL */ 311 312 #define inp_fport inp_inc.inc_fport 313 #define inp_lport inp_inc.inc_lport 314 #define inp_faddr inp_inc.inc_faddr 315 #define inp_laddr inp_inc.inc_laddr 316 317 #define in6p_faddr inp_inc.inc6_faddr 318 #define in6p_laddr inp_inc.inc6_laddr 319 #define in6p_zoneid inp_inc.inc6_zoneid 320 321 #define inp_vnet inp_pcbinfo->ipi_vnet 322 323 /* 324 * The range of the generation count, as used in this implementation, is 9e19. 325 * We would have to create 300 billion connections per second for this number 326 * to roll over in a year. This seems sufficiently unlikely that we simply 327 * don't concern ourselves with that possibility. 328 */ 329 330 /* 331 * Interface exported to userland by various protocols which use inpcbs. Hack 332 * alert -- only define if struct xsocket is in scope. 333 * Fields prefixed with "xi_" are unique to this structure, and the rest 334 * match fields in the struct inpcb, to ease coding and porting. 335 * 336 * Legend: 337 * (s) - used by userland utilities in src 338 * (p) - used by utilities in ports 339 * (3) - is known to be used by third party software not in ports 340 * (n) - no known usage 341 */ 342 #ifdef _SYS_SOCKETVAR_H_ 343 struct xinpcb { 344 ksize_t xi_len; /* length of this structure */ 345 struct xsocket xi_socket; /* (s,p) */ 346 struct in_conninfo inp_inc; /* (s,p) */ 347 uint64_t inp_gencnt; /* (s,p) */ 348 kvaddr_t inp_ppcb; /* (s) netstat(1) */ 349 int64_t inp_spare64[4]; 350 uint32_t inp_flow; /* (s) */ 351 uint32_t inp_flowid; /* (s) */ 352 uint32_t inp_flowtype; /* (s) */ 353 int32_t inp_flags; /* (s,p) */ 354 int32_t inp_flags2; /* (s) */ 355 int32_t inp_rss_listen_bucket; /* (n) */ 356 int32_t in6p_cksum; /* (n) */ 357 int32_t inp_spare32[4]; 358 uint16_t in6p_hops; /* (n) */ 359 uint8_t inp_ip_tos; /* (n) */ 360 int8_t pad8; 361 uint8_t inp_vflag; /* (s,p) */ 362 uint8_t inp_ip_ttl; /* (n) */ 363 uint8_t inp_ip_p; /* (n) */ 364 uint8_t inp_ip_minttl; /* (n) */ 365 int8_t inp_spare8[4]; 366 } __aligned(8); 367 368 struct xinpgen { 369 ksize_t xig_len; /* length of this structure */ 370 u_int xig_count; /* number of PCBs at this time */ 371 uint32_t _xig_spare32; 372 inp_gen_t xig_gen; /* generation count at this time */ 373 so_gen_t xig_sogen; /* socket generation count this time */ 374 uint64_t _xig_spare64[4]; 375 } __aligned(8); 376 #ifdef _KERNEL 377 void in_pcbtoxinpcb(const struct inpcb *, struct xinpcb *); 378 #endif 379 #endif /* _SYS_SOCKETVAR_H_ */ 380 381 #ifdef _KERNEL 382 /* 383 * Global data structure for each high-level protocol (UDP, TCP, ...) in both 384 * IPv4 and IPv6. Holds inpcb lists and information for managing them. 385 * 386 * The pcbs are protected with SMR section and thus all lists in inpcbinfo 387 * are CK-lists. Locking is required to insert a pcb into database. Two 388 * locks are provided: one for the hash and one for the global list of pcbs, 389 * as well as overall count and generation count. 390 * 391 * Locking key: 392 * 393 * (c) Constant or nearly constant after initialisation 394 * (e) Protected by SMR section 395 * (g) Locked by ipi_lock 396 * (h) Locked by ipi_hash_lock 397 */ 398 struct inpcbinfo { 399 /* 400 * Global lock protecting inpcb list modification 401 */ 402 struct mtx ipi_lock; 403 struct inpcbhead ipi_listhead; /* (r:e/w:g) */ 404 u_int ipi_count; /* (g) */ 405 406 /* 407 * Generation count -- incremented each time a connection is allocated 408 * or freed. 409 */ 410 u_quad_t ipi_gencnt; /* (g) */ 411 412 /* 413 * Fields associated with port lookup and allocation. 414 */ 415 u_short ipi_lastport; /* (h) */ 416 u_short ipi_lastlow; /* (h) */ 417 u_short ipi_lasthi; /* (h) */ 418 419 /* 420 * UMA zone from which inpcbs are allocated for this protocol. 421 */ 422 uma_zone_t ipi_zone; /* (c) */ 423 uma_zone_t ipi_portzone; /* (c) */ 424 smr_t ipi_smr; /* (c) */ 425 426 /* 427 * Global hash of inpcbs, hashed by local and foreign addresses and 428 * port numbers. 429 */ 430 struct mtx ipi_hash_lock; 431 struct inpcbhead *ipi_hashbase; /* (r:e/w:h) */ 432 u_long ipi_hashmask; /* (c) */ 433 434 /* 435 * Global hash of inpcbs, hashed by only local port number. 436 */ 437 struct inpcbporthead *ipi_porthashbase; /* (h) */ 438 u_long ipi_porthashmask; /* (h) */ 439 440 /* 441 * Load balance groups used for the SO_REUSEPORT_LB option, 442 * hashed by local port. 443 */ 444 struct inpcblbgrouphead *ipi_lbgrouphashbase; /* (r:e/w:h) */ 445 u_long ipi_lbgrouphashmask; /* (h) */ 446 447 /* 448 * Pointer to network stack instance 449 */ 450 struct vnet *ipi_vnet; /* (c) */ 451 }; 452 453 /* 454 * Load balance groups used for the SO_REUSEPORT_LB socket option. Each group 455 * (or unique address:port combination) can be re-used at most 456 * INPCBLBGROUP_SIZMAX (256) times. The inpcbs are stored in il_inp which 457 * is dynamically resized as processes bind/unbind to that specific group. 458 */ 459 struct inpcblbgroup { 460 CK_LIST_ENTRY(inpcblbgroup) il_list; 461 struct epoch_context il_epoch_ctx; 462 uint16_t il_lport; /* (c) */ 463 u_char il_vflag; /* (c) */ 464 u_int8_t il_numa_domain; 465 uint32_t il_pad2; 466 union in_dependaddr il_dependladdr; /* (c) */ 467 #define il_laddr il_dependladdr.id46_addr.ia46_addr4 468 #define il6_laddr il_dependladdr.id6_addr 469 uint32_t il_inpsiz; /* max count in il_inp[] (h) */ 470 uint32_t il_inpcnt; /* cur count in il_inp[] (h) */ 471 struct inpcb *il_inp[]; /* (h) */ 472 }; 473 474 #define INP_LOCK_INIT(inp, d, t) \ 475 rw_init_flags(&(inp)->inp_lock, (t), RW_RECURSE | RW_DUPOK) 476 #define INP_LOCK_DESTROY(inp) rw_destroy(&(inp)->inp_lock) 477 #define INP_RLOCK(inp) rw_rlock(&(inp)->inp_lock) 478 #define INP_WLOCK(inp) rw_wlock(&(inp)->inp_lock) 479 #define INP_TRY_RLOCK(inp) rw_try_rlock(&(inp)->inp_lock) 480 #define INP_TRY_WLOCK(inp) rw_try_wlock(&(inp)->inp_lock) 481 #define INP_RUNLOCK(inp) rw_runlock(&(inp)->inp_lock) 482 #define INP_WUNLOCK(inp) rw_wunlock(&(inp)->inp_lock) 483 #define INP_UNLOCK(inp) rw_unlock(&(inp)->inp_lock) 484 #define INP_TRY_UPGRADE(inp) rw_try_upgrade(&(inp)->inp_lock) 485 #define INP_DOWNGRADE(inp) rw_downgrade(&(inp)->inp_lock) 486 #define INP_WLOCKED(inp) rw_wowned(&(inp)->inp_lock) 487 #define INP_LOCK_ASSERT(inp) rw_assert(&(inp)->inp_lock, RA_LOCKED) 488 #define INP_RLOCK_ASSERT(inp) rw_assert(&(inp)->inp_lock, RA_RLOCKED) 489 #define INP_WLOCK_ASSERT(inp) rw_assert(&(inp)->inp_lock, RA_WLOCKED) 490 #define INP_UNLOCK_ASSERT(inp) rw_assert(&(inp)->inp_lock, RA_UNLOCKED) 491 492 /* 493 * These locking functions are for inpcb consumers outside of sys/netinet, 494 * more specifically, they were added for the benefit of TOE drivers. The 495 * macros are reserved for use by the stack. 496 */ 497 void inp_wlock(struct inpcb *); 498 void inp_wunlock(struct inpcb *); 499 void inp_rlock(struct inpcb *); 500 void inp_runlock(struct inpcb *); 501 502 #ifdef INVARIANT_SUPPORT 503 void inp_lock_assert(struct inpcb *); 504 void inp_unlock_assert(struct inpcb *); 505 #else 506 #define inp_lock_assert(inp) do {} while (0) 507 #define inp_unlock_assert(inp) do {} while (0) 508 #endif 509 510 void inp_apply_all(void (*func)(struct inpcb *, void *), void *arg); 511 int inp_ip_tos_get(const struct inpcb *inp); 512 void inp_ip_tos_set(struct inpcb *inp, int val); 513 struct socket * 514 inp_inpcbtosocket(struct inpcb *inp); 515 struct tcpcb * 516 inp_inpcbtotcpcb(struct inpcb *inp); 517 void inp_4tuple_get(struct inpcb *inp, uint32_t *laddr, uint16_t *lp, 518 uint32_t *faddr, uint16_t *fp); 519 int inp_so_options(const struct inpcb *inp); 520 521 #endif /* _KERNEL */ 522 523 #define INP_INFO_WLOCK(ipi) mtx_lock(&(ipi)->ipi_lock) 524 #define INP_INFO_WLOCKED(ipi) mtx_owned(&(ipi)->ipi_lock) 525 #define INP_INFO_WUNLOCK(ipi) mtx_unlock(&(ipi)->ipi_lock) 526 #define INP_INFO_LOCK_ASSERT(ipi) MPASS(SMR_ENTERED((ipi)->ipi_smr) || \ 527 mtx_owned(&(ipi)->ipi_lock)) 528 #define INP_INFO_WLOCK_ASSERT(ipi) mtx_assert(&(ipi)->ipi_lock, MA_OWNED) 529 #define INP_INFO_WUNLOCK_ASSERT(ipi) \ 530 mtx_assert(&(ipi)->ipi_lock, MA_NOTOWNED) 531 532 #define INP_HASH_WLOCK(ipi) mtx_lock(&(ipi)->ipi_hash_lock) 533 #define INP_HASH_WUNLOCK(ipi) mtx_unlock(&(ipi)->ipi_hash_lock) 534 #define INP_HASH_LOCK_ASSERT(ipi) MPASS(SMR_ENTERED((ipi)->ipi_smr) || \ 535 mtx_owned(&(ipi)->ipi_hash_lock)) 536 #define INP_HASH_WLOCK_ASSERT(ipi) mtx_assert(&(ipi)->ipi_hash_lock, \ 537 MA_OWNED) 538 539 #define INP_PCBHASH(faddr, lport, fport, mask) \ 540 (((faddr) ^ ((faddr) >> 16) ^ ntohs((lport) ^ (fport))) & (mask)) 541 #define INP_PCBPORTHASH(lport, mask) \ 542 (ntohs((lport)) & (mask)) 543 #define INP_PCBLBGROUP_PKTHASH(faddr, lport, fport) \ 544 ((faddr) ^ ((faddr) >> 16) ^ ntohs((lport) ^ (fport))) 545 #define INP6_PCBHASHKEY(faddr) ((faddr)->s6_addr32[3]) 546 547 /* 548 * Flags for inp_vflags -- historically version flags only 549 */ 550 #define INP_IPV4 0x1 551 #define INP_IPV6 0x2 552 #define INP_IPV6PROTO 0x4 /* opened under IPv6 protocol */ 553 554 /* 555 * Flags for inp_flags. 556 */ 557 #define INP_RECVOPTS 0x00000001 /* receive incoming IP options */ 558 #define INP_RECVRETOPTS 0x00000002 /* receive IP options for reply */ 559 #define INP_RECVDSTADDR 0x00000004 /* receive IP dst address */ 560 #define INP_HDRINCL 0x00000008 /* user supplies entire IP header */ 561 #define INP_HIGHPORT 0x00000010 /* user wants "high" port binding */ 562 #define INP_LOWPORT 0x00000020 /* user wants "low" port binding */ 563 #define INP_ANONPORT 0x00000040 /* port chosen for user */ 564 #define INP_RECVIF 0x00000080 /* receive incoming interface */ 565 #define INP_MTUDISC 0x00000100 /* user can do MTU discovery */ 566 /* INP_FREED 0x00000200 private to in_pcb.c */ 567 #define INP_RECVTTL 0x00000400 /* receive incoming IP TTL */ 568 #define INP_DONTFRAG 0x00000800 /* don't fragment packet */ 569 #define INP_BINDANY 0x00001000 /* allow bind to any address */ 570 #define INP_INHASHLIST 0x00002000 /* in_pcbinshash() has been called */ 571 #define INP_RECVTOS 0x00004000 /* receive incoming IP TOS */ 572 #define IN6P_IPV6_V6ONLY 0x00008000 /* restrict AF_INET6 socket for v6 */ 573 #define IN6P_PKTINFO 0x00010000 /* receive IP6 dst and I/F */ 574 #define IN6P_HOPLIMIT 0x00020000 /* receive hoplimit */ 575 #define IN6P_HOPOPTS 0x00040000 /* receive hop-by-hop options */ 576 #define IN6P_DSTOPTS 0x00080000 /* receive dst options after rthdr */ 577 #define IN6P_RTHDR 0x00100000 /* receive routing header */ 578 #define IN6P_RTHDRDSTOPTS 0x00200000 /* receive dstoptions before rthdr */ 579 #define IN6P_TCLASS 0x00400000 /* receive traffic class value */ 580 #define IN6P_AUTOFLOWLABEL 0x00800000 /* attach flowlabel automatically */ 581 #define INP_TIMEWAIT 0x01000000 /* in TIMEWAIT, ppcb is tcptw */ 582 #define INP_ONESBCAST 0x02000000 /* send all-ones broadcast */ 583 #define INP_DROPPED 0x04000000 /* protocol drop flag */ 584 #define INP_SOCKREF 0x08000000 /* strong socket reference */ 585 #define INP_RESERVED_0 0x10000000 /* reserved field */ 586 #define INP_RESERVED_1 0x20000000 /* reserved field */ 587 #define IN6P_RFC2292 0x40000000 /* used RFC2292 API on the socket */ 588 #define IN6P_MTU 0x80000000 /* receive path MTU */ 589 590 #define INP_CONTROLOPTS (INP_RECVOPTS|INP_RECVRETOPTS|INP_RECVDSTADDR|\ 591 INP_RECVIF|INP_RECVTTL|INP_RECVTOS|\ 592 IN6P_PKTINFO|IN6P_HOPLIMIT|IN6P_HOPOPTS|\ 593 IN6P_DSTOPTS|IN6P_RTHDR|IN6P_RTHDRDSTOPTS|\ 594 IN6P_TCLASS|IN6P_AUTOFLOWLABEL|IN6P_RFC2292|\ 595 IN6P_MTU) 596 597 /* 598 * Flags for inp_flags2. 599 */ 600 #define INP_MBUF_L_ACKS 0x00000001 /* We need large mbufs for ack compression */ 601 #define INP_MBUF_ACKCMP 0x00000002 /* TCP mbuf ack compression ok */ 602 /* 0x00000004 */ 603 #define INP_REUSEPORT 0x00000008 /* SO_REUSEPORT option is set */ 604 /* 0x00000010 */ 605 #define INP_REUSEADDR 0x00000020 /* SO_REUSEADDR option is set */ 606 #define INP_BINDMULTI 0x00000040 /* IP_BINDMULTI option is set */ 607 #define INP_RSS_BUCKET_SET 0x00000080 /* IP_RSS_LISTEN_BUCKET is set */ 608 #define INP_RECVFLOWID 0x00000100 /* populate recv datagram with flow info */ 609 #define INP_RECVRSSBUCKETID 0x00000200 /* populate recv datagram with bucket id */ 610 #define INP_RATE_LIMIT_CHANGED 0x00000400 /* rate limit needs attention */ 611 #define INP_ORIGDSTADDR 0x00000800 /* receive IP dst address/port */ 612 #define INP_CANNOT_DO_ECN 0x00001000 /* The stack does not do ECN */ 613 #define INP_REUSEPORT_LB 0x00002000 /* SO_REUSEPORT_LB option is set */ 614 #define INP_SUPPORTS_MBUFQ 0x00004000 /* Supports the mbuf queue method of LRO */ 615 #define INP_MBUF_QUEUE_READY 0x00008000 /* The transport is pacing, inputs can be queued */ 616 #define INP_DONT_SACK_QUEUE 0x00010000 /* If a sack arrives do not wake me */ 617 #define INP_2PCP_SET 0x00020000 /* If the Eth PCP should be set explicitly */ 618 #define INP_2PCP_BIT0 0x00040000 /* Eth PCP Bit 0 */ 619 #define INP_2PCP_BIT1 0x00080000 /* Eth PCP Bit 1 */ 620 #define INP_2PCP_BIT2 0x00100000 /* Eth PCP Bit 2 */ 621 #define INP_2PCP_BASE INP_2PCP_BIT0 622 #define INP_2PCP_MASK (INP_2PCP_BIT0 | INP_2PCP_BIT1 | INP_2PCP_BIT2) 623 #define INP_2PCP_SHIFT 18 /* shift PCP field in/out of inp_flags2 */ 624 625 /* 626 * Flags passed to in_pcblookup*(), inp_smr_lock() and inp_next(). 627 */ 628 typedef enum { 629 INPLOOKUP_WILDCARD = 0x00000001, /* Allow wildcard sockets. */ 630 INPLOOKUP_RLOCKPCB = 0x00000002, /* Return inpcb read-locked. */ 631 INPLOOKUP_WLOCKPCB = 0x00000004, /* Return inpcb write-locked. */ 632 } inp_lookup_t; 633 634 #define INPLOOKUP_MASK (INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB | \ 635 INPLOOKUP_WLOCKPCB) 636 #define INPLOOKUP_LOCKMASK (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB) 637 638 #define sotoinpcb(so) ((struct inpcb *)(so)->so_pcb) 639 640 #define INP_SOCKAF(so) so->so_proto->pr_domain->dom_family 641 642 #define INP_CHECK_SOCKAF(so, af) (INP_SOCKAF(so) == af) 643 644 #ifdef _KERNEL 645 VNET_DECLARE(int, ipport_reservedhigh); 646 VNET_DECLARE(int, ipport_reservedlow); 647 VNET_DECLARE(int, ipport_lowfirstauto); 648 VNET_DECLARE(int, ipport_lowlastauto); 649 VNET_DECLARE(int, ipport_firstauto); 650 VNET_DECLARE(int, ipport_lastauto); 651 VNET_DECLARE(int, ipport_hifirstauto); 652 VNET_DECLARE(int, ipport_hilastauto); 653 VNET_DECLARE(int, ipport_randomized); 654 VNET_DECLARE(int, ipport_randomcps); 655 VNET_DECLARE(int, ipport_randomtime); 656 VNET_DECLARE(int, ipport_stoprandom); 657 VNET_DECLARE(int, ipport_tcpallocs); 658 659 #define V_ipport_reservedhigh VNET(ipport_reservedhigh) 660 #define V_ipport_reservedlow VNET(ipport_reservedlow) 661 #define V_ipport_lowfirstauto VNET(ipport_lowfirstauto) 662 #define V_ipport_lowlastauto VNET(ipport_lowlastauto) 663 #define V_ipport_firstauto VNET(ipport_firstauto) 664 #define V_ipport_lastauto VNET(ipport_lastauto) 665 #define V_ipport_hifirstauto VNET(ipport_hifirstauto) 666 #define V_ipport_hilastauto VNET(ipport_hilastauto) 667 #define V_ipport_randomized VNET(ipport_randomized) 668 #define V_ipport_randomcps VNET(ipport_randomcps) 669 #define V_ipport_randomtime VNET(ipport_randomtime) 670 #define V_ipport_stoprandom VNET(ipport_stoprandom) 671 #define V_ipport_tcpallocs VNET(ipport_tcpallocs) 672 673 void in_pcbinfo_destroy(struct inpcbinfo *); 674 void in_pcbinfo_init(struct inpcbinfo *, const char *, u_int, int, char *, 675 uma_init); 676 677 int in_pcbbind_check_bindmulti(const struct inpcb *ni, 678 const struct inpcb *oi); 679 680 void in_pcbpurgeif0(struct inpcbinfo *, struct ifnet *); 681 int in_pcballoc(struct socket *, struct inpcbinfo *); 682 int in_pcbbind(struct inpcb *, struct sockaddr *, struct ucred *); 683 int in_pcbbind_setup(struct inpcb *, struct sockaddr *, in_addr_t *, 684 u_short *, struct ucred *); 685 int in_pcbconnect(struct inpcb *, struct sockaddr *, struct ucred *, bool); 686 int in_pcbconnect_setup(struct inpcb *, struct sockaddr *, in_addr_t *, 687 u_short *, in_addr_t *, u_short *, struct inpcb **, 688 struct ucred *); 689 void in_pcbdetach(struct inpcb *); 690 void in_pcbdisconnect(struct inpcb *); 691 void in_pcbdrop(struct inpcb *); 692 void in_pcbfree(struct inpcb *); 693 int in_pcbinshash(struct inpcb *); 694 int in_pcbladdr(struct inpcb *, struct in_addr *, struct in_addr *, 695 struct ucred *); 696 int in_pcblbgroup_numa(struct inpcb *, int arg); 697 struct inpcb * 698 in_pcblookup(struct inpcbinfo *, struct in_addr, u_int, 699 struct in_addr, u_int, int, struct ifnet *); 700 struct inpcb * 701 in_pcblookup_mbuf(struct inpcbinfo *, struct in_addr, u_int, 702 struct in_addr, u_int, int, struct ifnet *, struct mbuf *); 703 void in_pcbnotifyall(struct inpcbinfo *pcbinfo, struct in_addr, 704 int, struct inpcb *(*)(struct inpcb *, int)); 705 void in_pcbref(struct inpcb *); 706 void in_pcbrehash(struct inpcb *); 707 bool in_pcbrele_rlocked(struct inpcb *); 708 bool in_pcbrele_wlocked(struct inpcb *); 709 710 typedef bool inp_match_t(const struct inpcb *, void *); 711 struct inpcb_iterator { 712 const struct inpcbinfo *ipi; 713 struct inpcb *inp; 714 inp_match_t *match; 715 void *ctx; 716 int hash; 717 #define INP_ALL_LIST -1 718 const inp_lookup_t lock; 719 }; 720 721 /* Note: sparse initializers guarantee .inp = NULL. */ 722 #define INP_ITERATOR(_ipi, _lock, _match, _ctx) \ 723 { \ 724 .ipi = (_ipi), \ 725 .lock = (_lock), \ 726 .hash = INP_ALL_LIST, \ 727 .match = (_match), \ 728 .ctx = (_ctx), \ 729 } 730 #define INP_ALL_ITERATOR(_ipi, _lock) \ 731 { \ 732 .ipi = (_ipi), \ 733 .lock = (_lock), \ 734 .hash = INP_ALL_LIST, \ 735 } 736 737 struct inpcb *inp_next(struct inpcb_iterator *); 738 void in_losing(struct inpcb *); 739 void in_pcbsetsolabel(struct socket *so); 740 int in_getpeeraddr(struct socket *so, struct sockaddr **nam); 741 int in_getsockaddr(struct socket *so, struct sockaddr **nam); 742 struct sockaddr * 743 in_sockaddr(in_port_t port, struct in_addr *addr); 744 void in_pcbsosetlabel(struct socket *so); 745 #ifdef RATELIMIT 746 int 747 in_pcboutput_txrtlmt_locked(struct inpcb *, struct ifnet *, 748 struct mbuf *, uint32_t); 749 int in_pcbattach_txrtlmt(struct inpcb *, struct ifnet *, uint32_t, uint32_t, 750 uint32_t, struct m_snd_tag **); 751 void in_pcbdetach_txrtlmt(struct inpcb *); 752 void in_pcbdetach_tag(struct m_snd_tag *); 753 int in_pcbmodify_txrtlmt(struct inpcb *, uint32_t); 754 int in_pcbquery_txrtlmt(struct inpcb *, uint32_t *); 755 int in_pcbquery_txrlevel(struct inpcb *, uint32_t *); 756 void in_pcboutput_txrtlmt(struct inpcb *, struct ifnet *, struct mbuf *); 757 void in_pcboutput_eagain(struct inpcb *); 758 #endif 759 #endif /* _KERNEL */ 760 761 #endif /* !_NETINET_IN_PCB_H_ */ 762