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. This requires padding always be zeroed out, 77 * which is done right after inpcb allocation and stays through its lifetime. 78 */ 79 struct in_addr_4in6 { 80 u_int32_t ia46_pad32[3]; 81 struct in_addr ia46_addr4; 82 }; 83 84 union in_dependaddr { 85 struct in_addr_4in6 id46_addr; 86 struct in6_addr id6_addr; 87 }; 88 89 /* 90 * NOTE: ipv6 addrs should be 64-bit aligned, per RFC 2553. in_conninfo has 91 * some extra padding to accomplish this. 92 * NOTE 2: tcp_syncache.c uses first 5 32-bit words, which identify fport, 93 * lport, faddr to generate hash, so these fields shouldn't be moved. 94 */ 95 struct in_endpoints { 96 u_int16_t ie_fport; /* foreign port */ 97 u_int16_t ie_lport; /* local port */ 98 /* protocol dependent part, local and foreign addr */ 99 union in_dependaddr ie_dependfaddr; /* foreign host table entry */ 100 union in_dependaddr ie_dependladdr; /* local host table entry */ 101 #define ie_faddr ie_dependfaddr.id46_addr.ia46_addr4 102 #define ie_laddr ie_dependladdr.id46_addr.ia46_addr4 103 #define ie6_faddr ie_dependfaddr.id6_addr 104 #define ie6_laddr ie_dependladdr.id6_addr 105 u_int32_t ie6_zoneid; /* scope zone id */ 106 }; 107 108 /* 109 * XXX The defines for inc_* are hacks and should be changed to direct 110 * references. 111 */ 112 struct in_conninfo { 113 u_int8_t inc_flags; 114 u_int8_t inc_len; 115 u_int16_t inc_fibnum; /* XXX was pad, 16 bits is plenty */ 116 /* protocol dependent part */ 117 struct in_endpoints inc_ie; 118 }; 119 120 /* 121 * Flags for inc_flags. 122 */ 123 #define INC_ISIPV6 0x01 124 #define INC_IPV6MINMTU 0x02 125 126 #define inc_fport inc_ie.ie_fport 127 #define inc_lport inc_ie.ie_lport 128 #define inc_faddr inc_ie.ie_faddr 129 #define inc_laddr inc_ie.ie_laddr 130 #define inc6_faddr inc_ie.ie6_faddr 131 #define inc6_laddr inc_ie.ie6_laddr 132 #define inc6_zoneid inc_ie.ie6_zoneid 133 134 #if defined(_KERNEL) || defined(_WANT_INPCB) 135 /* 136 * struct inpcb captures the network layer state for TCP, UDP, and raw IPv4 and 137 * IPv6 sockets. In the case of TCP and UDP, further per-connection state is 138 * hung off of inp_ppcb most of the time. Almost all fields of struct inpcb 139 * are static after creation or protected by a per-inpcb rwlock, inp_lock. 140 * 141 * A inpcb database is indexed by addresses/ports hash as well as list of 142 * all pcbs that belong to a certain proto. Database lookups or list traversals 143 * are be performed inside SMR section. Once desired PCB is found its own 144 * lock is to be obtained and SMR section exited. 145 * 146 * Key: 147 * (b) - Protected by the hpts lock. 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 * Notes on the tcp_hpts: 157 * 158 * First Hpts lock order is 159 * 1) INP_WLOCK() 160 * 2) HPTS_LOCK() i.e. hpts->pmtx 161 * 162 * To insert a TCB on the hpts you *must* be holding the INP_WLOCK(). 163 * You may check the inp->inp_in_hpts flag without the hpts lock. 164 * The hpts is the only one that will clear this flag holding 165 * only the hpts lock. This means that in your tcp_output() 166 * routine when you test for the inp_in_hpts flag to be 1 167 * it may be transitioning to 0 (by the hpts). 168 * That's ok since that will just mean an extra call to tcp_output 169 * that most likely will find the call you executed 170 * (when the mis-match occured) will have put the TCB back 171 * on the hpts and it will return. If your 172 * call did not add the inp back to the hpts then you will either 173 * over-send or the cwnd will block you from sending more. 174 * 175 * Note you should also be holding the INP_WLOCK() when you 176 * call the remove from the hpts as well. Though usually 177 * you are either doing this from a timer, where you need and have 178 * the INP_WLOCK() or from destroying your TCB where again 179 * you should already have the INP_WLOCK(). 180 * 181 * The inp_hpts_cpu, inp_hpts_cpu_set, inp_input_cpu and 182 * inp_input_cpu_set fields are controlled completely by 183 * the hpts. Do not ever set these. The inp_hpts_cpu_set 184 * and inp_input_cpu_set fields indicate if the hpts has 185 * setup the respective cpu field. It is advised if this 186 * field is 0, to enqueue the packet with the appropriate 187 * hpts_immediate() call. If the _set field is 1, then 188 * you may compare the inp_*_cpu field to the curcpu and 189 * may want to again insert onto the hpts if these fields 190 * are not equal (i.e. you are not on the expected CPU). 191 * 192 * A note on inp_hpts_calls and inp_input_calls, these 193 * flags are set when the hpts calls either the output 194 * or do_segment routines respectively. If the routine 195 * being called wants to use this, then it needs to 196 * clear the flag before returning. The hpts will not 197 * clear the flag. The flags can be used to tell if 198 * the hpts is the function calling the respective 199 * routine. 200 * 201 * A few other notes: 202 * 203 * When a read lock is held, stability of the field is guaranteed; to write 204 * to a field, a write lock must generally be held. 205 * 206 * netinet/netinet6-layer code should not assume that the inp_socket pointer 207 * is safe to dereference without inp_lock being held, even for protocols 208 * other than TCP (where the inpcb persists during TIMEWAIT even after the 209 * socket has been freed), or there may be close(2)-related races. 210 * 211 * The inp_vflag field is overloaded, and would otherwise ideally be (c). 212 */ 213 struct icmp6_filter; 214 struct inpcbpolicy; 215 struct m_snd_tag; 216 struct inpcb { 217 /* Cache line #1 (amd64) */ 218 CK_LIST_ENTRY(inpcb) inp_hash; /* (w:h/r:e) hash list */ 219 struct rwlock inp_lock; 220 /* Cache line #2 (amd64) */ 221 #define inp_start_zero inp_hpts 222 #define inp_zero_size (sizeof(struct inpcb) - \ 223 offsetof(struct inpcb, inp_start_zero)) 224 TAILQ_ENTRY(inpcb) inp_hpts; /* pacing out queue next lock(b) */ 225 uint32_t inp_hpts_gencnt; /* XXXGL */ 226 uint32_t inp_hpts_request; /* Current hpts request, zero if 227 * fits in the pacing window (i&b). */ 228 /* 229 * Note the next fields are protected by a 230 * different lock (hpts-lock). This means that 231 * they must correspond in size to the smallest 232 * protectable bit field (uint8_t on x86, and 233 * other platfomrs potentially uint32_t?). Also 234 * since CPU switches can occur at different times the two 235 * fields can *not* be collapsed into a signal bit field. 236 */ 237 #if defined(__amd64__) || defined(__i386__) 238 uint8_t inp_in_hpts; /* on output hpts (lock b) */ 239 #else 240 uint32_t inp_in_hpts; /* on output hpts (lock b) */ 241 #endif 242 volatile uint16_t inp_hpts_cpu; /* Lock (i) */ 243 volatile uint16_t inp_irq_cpu; /* Set by LRO in behalf of or the driver */ 244 u_int inp_refcount; /* (i) refcount */ 245 int inp_flags; /* (i) generic IP/datagram flags */ 246 int inp_flags2; /* (i) generic IP/datagram flags #2*/ 247 uint8_t inp_hpts_cpu_set :1, /* on output hpts (i) */ 248 inp_hpts_calls :1, /* (i) from output hpts */ 249 inp_irq_cpu_set :1, /* (i) from LRO/Driver */ 250 inp_spare_bits2 : 3; 251 uint8_t inp_numa_domain; /* numa domain */ 252 void *inp_ppcb; /* (i) pointer to per-protocol pcb */ 253 struct socket *inp_socket; /* (i) back pointer to socket */ 254 int32_t inp_hptsslot; /* Hpts wheel slot this tcb is Lock(i&b) */ 255 uint32_t inp_hpts_drop_reas; /* reason we are dropping the PCB (lock i&b) */ 256 struct inpcbinfo *inp_pcbinfo; /* (c) PCB list info */ 257 struct ucred *inp_cred; /* (c) cache of socket cred */ 258 u_int32_t inp_flow; /* (i) IPv6 flow information */ 259 u_char inp_vflag; /* (i) IP version flag (v4/v6) */ 260 u_char inp_ip_ttl; /* (i) time to live proto */ 261 u_char inp_ip_p; /* (c) protocol proto */ 262 u_char inp_ip_minttl; /* (i) minimum TTL or drop */ 263 uint32_t inp_flowid; /* (x) flow id / queue id */ 264 struct m_snd_tag *inp_snd_tag; /* (i) send tag for outgoing mbufs */ 265 uint32_t inp_flowtype; /* (x) M_HASHTYPE value */ 266 uint32_t inp_rss_listen_bucket; /* (x) overridden RSS listen bucket */ 267 268 /* Local and foreign ports, local and foreign addr. */ 269 struct in_conninfo inp_inc; /* (i) list for PCB's local port */ 270 271 /* MAC and IPSEC policy information. */ 272 struct label *inp_label; /* (i) MAC label */ 273 struct inpcbpolicy *inp_sp; /* (s) for IPSEC */ 274 275 /* Protocol-dependent part; options. */ 276 struct { 277 u_char inp_ip_tos; /* (i) type of service proto */ 278 struct mbuf *inp_options; /* (i) IP options */ 279 struct ip_moptions *inp_moptions; /* (i) mcast options */ 280 }; 281 struct { 282 /* (i) IP options */ 283 struct mbuf *in6p_options; 284 /* (i) IP6 options for outgoing packets */ 285 struct ip6_pktopts *in6p_outputopts; 286 /* (i) IP multicast options */ 287 struct ip6_moptions *in6p_moptions; 288 /* (i) ICMPv6 code type filter */ 289 struct icmp6_filter *in6p_icmp6filt; 290 /* (i) IPV6_CHECKSUM setsockopt */ 291 int in6p_cksum; 292 short in6p_hops; 293 }; 294 CK_LIST_ENTRY(inpcb) inp_portlist; /* (r:e/w:h) port list */ 295 struct inpcbport *inp_phd; /* (r:e/w:h) head of this list */ 296 inp_gen_t inp_gencnt; /* (c) generation count */ 297 void *spare_ptr; /* Spare pointer. */ 298 rt_gen_t inp_rt_cookie; /* generation for route entry */ 299 union { /* cached L3 information */ 300 struct route inp_route; 301 struct route_in6 inp_route6; 302 }; 303 CK_LIST_ENTRY(inpcb) inp_list; /* (r:e/w:p) all PCBs for proto */ 304 }; 305 #endif /* _KERNEL */ 306 307 #define inp_fport inp_inc.inc_fport 308 #define inp_lport inp_inc.inc_lport 309 #define inp_faddr inp_inc.inc_faddr 310 #define inp_laddr inp_inc.inc_laddr 311 312 #define in6p_faddr inp_inc.inc6_faddr 313 #define in6p_laddr inp_inc.inc6_laddr 314 #define in6p_zoneid inp_inc.inc6_zoneid 315 316 #define inp_vnet inp_pcbinfo->ipi_vnet 317 318 /* 319 * The range of the generation count, as used in this implementation, is 9e19. 320 * We would have to create 300 billion connections per second for this number 321 * to roll over in a year. This seems sufficiently unlikely that we simply 322 * don't concern ourselves with that possibility. 323 */ 324 325 /* 326 * Interface exported to userland by various protocols which use inpcbs. Hack 327 * alert -- only define if struct xsocket is in scope. 328 * Fields prefixed with "xi_" are unique to this structure, and the rest 329 * match fields in the struct inpcb, to ease coding and porting. 330 * 331 * Legend: 332 * (s) - used by userland utilities in src 333 * (p) - used by utilities in ports 334 * (3) - is known to be used by third party software not in ports 335 * (n) - no known usage 336 */ 337 #ifdef _SYS_SOCKETVAR_H_ 338 struct xinpcb { 339 ksize_t xi_len; /* length of this structure */ 340 struct xsocket xi_socket; /* (s,p) */ 341 struct in_conninfo inp_inc; /* (s,p) */ 342 uint64_t inp_gencnt; /* (s,p) */ 343 kvaddr_t inp_ppcb; /* (s) netstat(1) */ 344 int64_t inp_spare64[4]; 345 uint32_t inp_flow; /* (s) */ 346 uint32_t inp_flowid; /* (s) */ 347 uint32_t inp_flowtype; /* (s) */ 348 int32_t inp_flags; /* (s,p) */ 349 int32_t inp_flags2; /* (s) */ 350 int32_t inp_rss_listen_bucket; /* (n) */ 351 int32_t in6p_cksum; /* (n) */ 352 int32_t inp_spare32[4]; 353 uint16_t in6p_hops; /* (n) */ 354 uint8_t inp_ip_tos; /* (n) */ 355 int8_t pad8; 356 uint8_t inp_vflag; /* (s,p) */ 357 uint8_t inp_ip_ttl; /* (n) */ 358 uint8_t inp_ip_p; /* (n) */ 359 uint8_t inp_ip_minttl; /* (n) */ 360 int8_t inp_spare8[4]; 361 } __aligned(8); 362 363 struct xinpgen { 364 ksize_t xig_len; /* length of this structure */ 365 u_int xig_count; /* number of PCBs at this time */ 366 uint32_t _xig_spare32; 367 inp_gen_t xig_gen; /* generation count at this time */ 368 so_gen_t xig_sogen; /* socket generation count this time */ 369 uint64_t _xig_spare64[4]; 370 } __aligned(8); 371 #ifdef _KERNEL 372 void in_pcbtoxinpcb(const struct inpcb *, struct xinpcb *); 373 #endif 374 #endif /* _SYS_SOCKETVAR_H_ */ 375 376 #ifdef _KERNEL 377 /* 378 * Global data structure for each high-level protocol (UDP, TCP, ...) in both 379 * IPv4 and IPv6. Holds inpcb lists and information for managing them. 380 * 381 * The pcbs are protected with SMR section and thus all lists in inpcbinfo 382 * are CK-lists. Locking is required to insert a pcb into database. Two 383 * locks are provided: one for the hash and one for the global list of pcbs, 384 * as well as overall count and generation count. 385 * 386 * Locking key: 387 * 388 * (c) Constant or nearly constant after initialisation 389 * (e) Protected by SMR section 390 * (g) Locked by ipi_lock 391 * (h) Locked by ipi_hash_lock 392 */ 393 struct inpcbinfo { 394 /* 395 * Global lock protecting inpcb list modification 396 */ 397 struct mtx ipi_lock; 398 struct inpcbhead ipi_listhead; /* (r:e/w:g) */ 399 u_int ipi_count; /* (g) */ 400 401 /* 402 * Generation count -- incremented each time a connection is allocated 403 * or freed. 404 */ 405 u_quad_t ipi_gencnt; /* (g) */ 406 407 /* 408 * Fields associated with port lookup and allocation. 409 */ 410 u_short ipi_lastport; /* (h) */ 411 u_short ipi_lastlow; /* (h) */ 412 u_short ipi_lasthi; /* (h) */ 413 414 /* 415 * UMA zone from which inpcbs are allocated for this protocol. 416 */ 417 uma_zone_t ipi_zone; /* (c) */ 418 uma_zone_t ipi_portzone; /* (c) */ 419 smr_t ipi_smr; /* (c) */ 420 421 /* 422 * Global hash of inpcbs, hashed by local and foreign addresses and 423 * port numbers. 424 */ 425 struct mtx ipi_hash_lock; 426 struct inpcbhead *ipi_hashbase; /* (r:e/w:h) */ 427 u_long ipi_hashmask; /* (c) */ 428 429 /* 430 * Global hash of inpcbs, hashed by only local port number. 431 */ 432 struct inpcbporthead *ipi_porthashbase; /* (h) */ 433 u_long ipi_porthashmask; /* (h) */ 434 435 /* 436 * Load balance groups used for the SO_REUSEPORT_LB option, 437 * hashed by local port. 438 */ 439 struct inpcblbgrouphead *ipi_lbgrouphashbase; /* (r:e/w:h) */ 440 u_long ipi_lbgrouphashmask; /* (h) */ 441 442 /* 443 * Pointer to network stack instance 444 */ 445 struct vnet *ipi_vnet; /* (c) */ 446 }; 447 448 /* 449 * Load balance groups used for the SO_REUSEPORT_LB socket option. Each group 450 * (or unique address:port combination) can be re-used at most 451 * INPCBLBGROUP_SIZMAX (256) times. The inpcbs are stored in il_inp which 452 * is dynamically resized as processes bind/unbind to that specific group. 453 */ 454 struct inpcblbgroup { 455 CK_LIST_ENTRY(inpcblbgroup) il_list; 456 struct epoch_context il_epoch_ctx; 457 uint16_t il_lport; /* (c) */ 458 u_char il_vflag; /* (c) */ 459 u_int8_t il_numa_domain; 460 uint32_t il_pad2; 461 union in_dependaddr il_dependladdr; /* (c) */ 462 #define il_laddr il_dependladdr.id46_addr.ia46_addr4 463 #define il6_laddr il_dependladdr.id6_addr 464 uint32_t il_inpsiz; /* max count in il_inp[] (h) */ 465 uint32_t il_inpcnt; /* cur count in il_inp[] (h) */ 466 struct inpcb *il_inp[]; /* (h) */ 467 }; 468 469 #define INP_LOCK_INIT(inp, d, t) \ 470 rw_init_flags(&(inp)->inp_lock, (t), RW_RECURSE | RW_DUPOK) 471 #define INP_LOCK_DESTROY(inp) rw_destroy(&(inp)->inp_lock) 472 #define INP_RLOCK(inp) rw_rlock(&(inp)->inp_lock) 473 #define INP_WLOCK(inp) rw_wlock(&(inp)->inp_lock) 474 #define INP_TRY_RLOCK(inp) rw_try_rlock(&(inp)->inp_lock) 475 #define INP_TRY_WLOCK(inp) rw_try_wlock(&(inp)->inp_lock) 476 #define INP_RUNLOCK(inp) rw_runlock(&(inp)->inp_lock) 477 #define INP_WUNLOCK(inp) rw_wunlock(&(inp)->inp_lock) 478 #define INP_UNLOCK(inp) rw_unlock(&(inp)->inp_lock) 479 #define INP_TRY_UPGRADE(inp) rw_try_upgrade(&(inp)->inp_lock) 480 #define INP_DOWNGRADE(inp) rw_downgrade(&(inp)->inp_lock) 481 #define INP_WLOCKED(inp) rw_wowned(&(inp)->inp_lock) 482 #define INP_LOCK_ASSERT(inp) rw_assert(&(inp)->inp_lock, RA_LOCKED) 483 #define INP_RLOCK_ASSERT(inp) rw_assert(&(inp)->inp_lock, RA_RLOCKED) 484 #define INP_WLOCK_ASSERT(inp) rw_assert(&(inp)->inp_lock, RA_WLOCKED) 485 #define INP_UNLOCK_ASSERT(inp) rw_assert(&(inp)->inp_lock, RA_UNLOCKED) 486 487 /* 488 * These locking functions are for inpcb consumers outside of sys/netinet, 489 * more specifically, they were added for the benefit of TOE drivers. The 490 * macros are reserved for use by the stack. 491 */ 492 void inp_wlock(struct inpcb *); 493 void inp_wunlock(struct inpcb *); 494 void inp_rlock(struct inpcb *); 495 void inp_runlock(struct inpcb *); 496 497 #ifdef INVARIANT_SUPPORT 498 void inp_lock_assert(struct inpcb *); 499 void inp_unlock_assert(struct inpcb *); 500 #else 501 #define inp_lock_assert(inp) do {} while (0) 502 #define inp_unlock_assert(inp) do {} while (0) 503 #endif 504 505 void inp_apply_all(void (*func)(struct inpcb *, void *), void *arg); 506 int inp_ip_tos_get(const struct inpcb *inp); 507 void inp_ip_tos_set(struct inpcb *inp, int val); 508 struct socket * 509 inp_inpcbtosocket(struct inpcb *inp); 510 struct tcpcb * 511 inp_inpcbtotcpcb(struct inpcb *inp); 512 void inp_4tuple_get(struct inpcb *inp, uint32_t *laddr, uint16_t *lp, 513 uint32_t *faddr, uint16_t *fp); 514 int inp_so_options(const struct inpcb *inp); 515 516 #endif /* _KERNEL */ 517 518 #define INP_INFO_WLOCK(ipi) mtx_lock(&(ipi)->ipi_lock) 519 #define INP_INFO_WLOCKED(ipi) mtx_owned(&(ipi)->ipi_lock) 520 #define INP_INFO_WUNLOCK(ipi) mtx_unlock(&(ipi)->ipi_lock) 521 #define INP_INFO_LOCK_ASSERT(ipi) MPASS(SMR_ENTERED((ipi)->ipi_smr) || \ 522 mtx_owned(&(ipi)->ipi_lock)) 523 #define INP_INFO_WLOCK_ASSERT(ipi) mtx_assert(&(ipi)->ipi_lock, MA_OWNED) 524 #define INP_INFO_WUNLOCK_ASSERT(ipi) \ 525 mtx_assert(&(ipi)->ipi_lock, MA_NOTOWNED) 526 527 #define INP_HASH_WLOCK(ipi) mtx_lock(&(ipi)->ipi_hash_lock) 528 #define INP_HASH_WUNLOCK(ipi) mtx_unlock(&(ipi)->ipi_hash_lock) 529 #define INP_HASH_LOCK_ASSERT(ipi) MPASS(SMR_ENTERED((ipi)->ipi_smr) || \ 530 mtx_owned(&(ipi)->ipi_hash_lock)) 531 #define INP_HASH_WLOCK_ASSERT(ipi) mtx_assert(&(ipi)->ipi_hash_lock, \ 532 MA_OWNED) 533 534 /* 535 * Wildcard matching hash is not just a microoptimisation! The hash for 536 * wildcard IPv4 and wildcard IPv6 must be the same, otherwise AF_INET6 537 * wildcard bound pcb won't be able to receive AF_INET connections, while: 538 * jenkins_hash(&zeroes, 1, s) != jenkins_hash(&zeroes, 4, s) 539 * See also comment above struct in_addr_4in6. 540 */ 541 #define IN_ADDR_JHASH32(addr) \ 542 ((addr)->s_addr == INADDR_ANY ? V_in_pcbhashseed : \ 543 jenkins_hash32((&(addr)->s_addr), 1, V_in_pcbhashseed)) 544 #define IN6_ADDR_JHASH32(addr) \ 545 (memcmp((addr), &in6addr_any, sizeof(in6addr_any)) == 0 ? \ 546 V_in_pcbhashseed : \ 547 jenkins_hash32((addr)->__u6_addr.__u6_addr32, \ 548 nitems((addr)->__u6_addr.__u6_addr32), V_in_pcbhashseed)) 549 550 #define INP_PCBHASH(faddr, lport, fport, mask) \ 551 ((IN_ADDR_JHASH32(faddr) ^ ntohs((lport) ^ (fport))) & (mask)) 552 #define INP6_PCBHASH(faddr, lport, fport, mask) \ 553 ((IN6_ADDR_JHASH32(faddr) ^ ntohs((lport) ^ (fport))) & (mask)) 554 555 #define INP_PCBHASH_WILD(lport, mask) \ 556 ((V_in_pcbhashseed ^ ntohs(lport)) & (mask)) 557 558 #define INP_PCBLBGROUP_PKTHASH(faddr, lport, fport) \ 559 (IN_ADDR_JHASH32(faddr) ^ ntohs((lport) ^ (fport))) 560 #define INP6_PCBLBGROUP_PKTHASH(faddr, lport, fport) \ 561 (IN6_ADDR_JHASH32(faddr) ^ ntohs((lport) ^ (fport))) 562 563 #define INP_PCBPORTHASH(lport, mask) (ntohs((lport)) & (mask)) 564 565 /* 566 * Flags for inp_vflags -- historically version flags only 567 */ 568 #define INP_IPV4 0x1 569 #define INP_IPV6 0x2 570 #define INP_IPV6PROTO 0x4 /* opened under IPv6 protocol */ 571 572 /* 573 * Flags for inp_flags. 574 */ 575 #define INP_RECVOPTS 0x00000001 /* receive incoming IP options */ 576 #define INP_RECVRETOPTS 0x00000002 /* receive IP options for reply */ 577 #define INP_RECVDSTADDR 0x00000004 /* receive IP dst address */ 578 #define INP_HDRINCL 0x00000008 /* user supplies entire IP header */ 579 #define INP_HIGHPORT 0x00000010 /* user wants "high" port binding */ 580 #define INP_LOWPORT 0x00000020 /* user wants "low" port binding */ 581 #define INP_ANONPORT 0x00000040 /* port chosen for user */ 582 #define INP_RECVIF 0x00000080 /* receive incoming interface */ 583 #define INP_MTUDISC 0x00000100 /* user can do MTU discovery */ 584 /* INP_FREED 0x00000200 private to in_pcb.c */ 585 #define INP_RECVTTL 0x00000400 /* receive incoming IP TTL */ 586 #define INP_DONTFRAG 0x00000800 /* don't fragment packet */ 587 #define INP_BINDANY 0x00001000 /* allow bind to any address */ 588 #define INP_INHASHLIST 0x00002000 /* in_pcbinshash() has been called */ 589 #define INP_RECVTOS 0x00004000 /* receive incoming IP TOS */ 590 #define IN6P_IPV6_V6ONLY 0x00008000 /* restrict AF_INET6 socket for v6 */ 591 #define IN6P_PKTINFO 0x00010000 /* receive IP6 dst and I/F */ 592 #define IN6P_HOPLIMIT 0x00020000 /* receive hoplimit */ 593 #define IN6P_HOPOPTS 0x00040000 /* receive hop-by-hop options */ 594 #define IN6P_DSTOPTS 0x00080000 /* receive dst options after rthdr */ 595 #define IN6P_RTHDR 0x00100000 /* receive routing header */ 596 #define IN6P_RTHDRDSTOPTS 0x00200000 /* receive dstoptions before rthdr */ 597 #define IN6P_TCLASS 0x00400000 /* receive traffic class value */ 598 #define IN6P_AUTOFLOWLABEL 0x00800000 /* attach flowlabel automatically */ 599 #define INP_TIMEWAIT 0x01000000 /* in TIMEWAIT, ppcb is tcptw */ 600 #define INP_ONESBCAST 0x02000000 /* send all-ones broadcast */ 601 #define INP_DROPPED 0x04000000 /* protocol drop flag */ 602 #define INP_SOCKREF 0x08000000 /* strong socket reference */ 603 #define INP_RESERVED_0 0x10000000 /* reserved field */ 604 #define INP_RESERVED_1 0x20000000 /* reserved field */ 605 #define IN6P_RFC2292 0x40000000 /* used RFC2292 API on the socket */ 606 #define IN6P_MTU 0x80000000 /* receive path MTU */ 607 608 #define INP_CONTROLOPTS (INP_RECVOPTS|INP_RECVRETOPTS|INP_RECVDSTADDR|\ 609 INP_RECVIF|INP_RECVTTL|INP_RECVTOS|\ 610 IN6P_PKTINFO|IN6P_HOPLIMIT|IN6P_HOPOPTS|\ 611 IN6P_DSTOPTS|IN6P_RTHDR|IN6P_RTHDRDSTOPTS|\ 612 IN6P_TCLASS|IN6P_AUTOFLOWLABEL|IN6P_RFC2292|\ 613 IN6P_MTU) 614 615 /* 616 * Flags for inp_flags2. 617 */ 618 #define INP_MBUF_L_ACKS 0x00000001 /* We need large mbufs for ack compression */ 619 #define INP_MBUF_ACKCMP 0x00000002 /* TCP mbuf ack compression ok */ 620 /* 0x00000004 */ 621 #define INP_REUSEPORT 0x00000008 /* SO_REUSEPORT option is set */ 622 /* 0x00000010 */ 623 #define INP_REUSEADDR 0x00000020 /* SO_REUSEADDR option is set */ 624 #define INP_BINDMULTI 0x00000040 /* IP_BINDMULTI option is set */ 625 #define INP_RSS_BUCKET_SET 0x00000080 /* IP_RSS_LISTEN_BUCKET is set */ 626 #define INP_RECVFLOWID 0x00000100 /* populate recv datagram with flow info */ 627 #define INP_RECVRSSBUCKETID 0x00000200 /* populate recv datagram with bucket id */ 628 #define INP_RATE_LIMIT_CHANGED 0x00000400 /* rate limit needs attention */ 629 #define INP_ORIGDSTADDR 0x00000800 /* receive IP dst address/port */ 630 #define INP_CANNOT_DO_ECN 0x00001000 /* The stack does not do ECN */ 631 #define INP_REUSEPORT_LB 0x00002000 /* SO_REUSEPORT_LB option is set */ 632 #define INP_SUPPORTS_MBUFQ 0x00004000 /* Supports the mbuf queue method of LRO */ 633 #define INP_MBUF_QUEUE_READY 0x00008000 /* The transport is pacing, inputs can be queued */ 634 #define INP_DONT_SACK_QUEUE 0x00010000 /* If a sack arrives do not wake me */ 635 #define INP_2PCP_SET 0x00020000 /* If the Eth PCP should be set explicitly */ 636 #define INP_2PCP_BIT0 0x00040000 /* Eth PCP Bit 0 */ 637 #define INP_2PCP_BIT1 0x00080000 /* Eth PCP Bit 1 */ 638 #define INP_2PCP_BIT2 0x00100000 /* Eth PCP Bit 2 */ 639 #define INP_2PCP_BASE INP_2PCP_BIT0 640 #define INP_2PCP_MASK (INP_2PCP_BIT0 | INP_2PCP_BIT1 | INP_2PCP_BIT2) 641 #define INP_2PCP_SHIFT 18 /* shift PCP field in/out of inp_flags2 */ 642 643 /* 644 * Flags passed to in_pcblookup*(), inp_smr_lock() and inp_next(). 645 */ 646 typedef enum { 647 INPLOOKUP_WILDCARD = 0x00000001, /* Allow wildcard sockets. */ 648 INPLOOKUP_RLOCKPCB = 0x00000002, /* Return inpcb read-locked. */ 649 INPLOOKUP_WLOCKPCB = 0x00000004, /* Return inpcb write-locked. */ 650 } inp_lookup_t; 651 652 #define INPLOOKUP_MASK (INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB | \ 653 INPLOOKUP_WLOCKPCB) 654 #define INPLOOKUP_LOCKMASK (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB) 655 656 #define sotoinpcb(so) ((struct inpcb *)(so)->so_pcb) 657 658 #define INP_SOCKAF(so) so->so_proto->pr_domain->dom_family 659 660 #define INP_CHECK_SOCKAF(so, af) (INP_SOCKAF(so) == af) 661 662 #ifdef _KERNEL 663 VNET_DECLARE(int, ipport_reservedhigh); 664 VNET_DECLARE(int, ipport_reservedlow); 665 VNET_DECLARE(int, ipport_lowfirstauto); 666 VNET_DECLARE(int, ipport_lowlastauto); 667 VNET_DECLARE(int, ipport_firstauto); 668 VNET_DECLARE(int, ipport_lastauto); 669 VNET_DECLARE(int, ipport_hifirstauto); 670 VNET_DECLARE(int, ipport_hilastauto); 671 VNET_DECLARE(int, ipport_randomized); 672 VNET_DECLARE(int, ipport_randomcps); 673 VNET_DECLARE(int, ipport_randomtime); 674 VNET_DECLARE(int, ipport_stoprandom); 675 VNET_DECLARE(int, ipport_tcpallocs); 676 677 #define V_ipport_reservedhigh VNET(ipport_reservedhigh) 678 #define V_ipport_reservedlow VNET(ipport_reservedlow) 679 #define V_ipport_lowfirstauto VNET(ipport_lowfirstauto) 680 #define V_ipport_lowlastauto VNET(ipport_lowlastauto) 681 #define V_ipport_firstauto VNET(ipport_firstauto) 682 #define V_ipport_lastauto VNET(ipport_lastauto) 683 #define V_ipport_hifirstauto VNET(ipport_hifirstauto) 684 #define V_ipport_hilastauto VNET(ipport_hilastauto) 685 #define V_ipport_randomized VNET(ipport_randomized) 686 #define V_ipport_randomcps VNET(ipport_randomcps) 687 #define V_ipport_randomtime VNET(ipport_randomtime) 688 #define V_ipport_stoprandom VNET(ipport_stoprandom) 689 #define V_ipport_tcpallocs VNET(ipport_tcpallocs) 690 691 void in_pcbinfo_destroy(struct inpcbinfo *); 692 void in_pcbinfo_init(struct inpcbinfo *, const char *, u_int, int, char *, 693 uma_init); 694 695 int in_pcbbind_check_bindmulti(const struct inpcb *ni, 696 const struct inpcb *oi); 697 698 void in_pcbpurgeif0(struct inpcbinfo *, struct ifnet *); 699 int in_pcballoc(struct socket *, struct inpcbinfo *); 700 int in_pcbbind(struct inpcb *, struct sockaddr *, struct ucred *); 701 int in_pcbbind_setup(struct inpcb *, struct sockaddr *, in_addr_t *, 702 u_short *, struct ucred *); 703 int in_pcbconnect(struct inpcb *, struct sockaddr *, struct ucred *, bool); 704 int in_pcbconnect_setup(struct inpcb *, struct sockaddr *, in_addr_t *, 705 u_short *, in_addr_t *, u_short *, struct inpcb **, 706 struct ucred *); 707 void in_pcbdetach(struct inpcb *); 708 void in_pcbdisconnect(struct inpcb *); 709 void in_pcbdrop(struct inpcb *); 710 void in_pcbfree(struct inpcb *); 711 int in_pcbinshash(struct inpcb *); 712 int in_pcbladdr(struct inpcb *, struct in_addr *, struct in_addr *, 713 struct ucred *); 714 int in_pcblbgroup_numa(struct inpcb *, int arg); 715 struct inpcb * 716 in_pcblookup(struct inpcbinfo *, struct in_addr, u_int, 717 struct in_addr, u_int, int, struct ifnet *); 718 struct inpcb * 719 in_pcblookup_mbuf(struct inpcbinfo *, struct in_addr, u_int, 720 struct in_addr, u_int, int, struct ifnet *, struct mbuf *); 721 void in_pcbnotifyall(struct inpcbinfo *pcbinfo, struct in_addr, 722 int, struct inpcb *(*)(struct inpcb *, int)); 723 void in_pcbref(struct inpcb *); 724 void in_pcbrehash(struct inpcb *); 725 bool in_pcbrele_rlocked(struct inpcb *); 726 bool in_pcbrele_wlocked(struct inpcb *); 727 728 typedef bool inp_match_t(const struct inpcb *, void *); 729 struct inpcb_iterator { 730 const struct inpcbinfo *ipi; 731 struct inpcb *inp; 732 inp_match_t *match; 733 void *ctx; 734 int hash; 735 #define INP_ALL_LIST -1 736 const inp_lookup_t lock; 737 }; 738 739 /* Note: sparse initializers guarantee .inp = NULL. */ 740 #define INP_ITERATOR(_ipi, _lock, _match, _ctx) \ 741 { \ 742 .ipi = (_ipi), \ 743 .lock = (_lock), \ 744 .hash = INP_ALL_LIST, \ 745 .match = (_match), \ 746 .ctx = (_ctx), \ 747 } 748 #define INP_ALL_ITERATOR(_ipi, _lock) \ 749 { \ 750 .ipi = (_ipi), \ 751 .lock = (_lock), \ 752 .hash = INP_ALL_LIST, \ 753 } 754 755 struct inpcb *inp_next(struct inpcb_iterator *); 756 void in_losing(struct inpcb *); 757 void in_pcbsetsolabel(struct socket *so); 758 int in_getpeeraddr(struct socket *so, struct sockaddr **nam); 759 int in_getsockaddr(struct socket *so, struct sockaddr **nam); 760 struct sockaddr * 761 in_sockaddr(in_port_t port, struct in_addr *addr); 762 void in_pcbsosetlabel(struct socket *so); 763 #ifdef RATELIMIT 764 int 765 in_pcboutput_txrtlmt_locked(struct inpcb *, struct ifnet *, 766 struct mbuf *, uint32_t); 767 int in_pcbattach_txrtlmt(struct inpcb *, struct ifnet *, uint32_t, uint32_t, 768 uint32_t, struct m_snd_tag **); 769 void in_pcbdetach_txrtlmt(struct inpcb *); 770 void in_pcbdetach_tag(struct m_snd_tag *); 771 int in_pcbmodify_txrtlmt(struct inpcb *, uint32_t); 772 int in_pcbquery_txrtlmt(struct inpcb *, uint32_t *); 773 int in_pcbquery_txrlevel(struct inpcb *, uint32_t *); 774 void in_pcboutput_txrtlmt(struct inpcb *, struct ifnet *, struct mbuf *); 775 void in_pcboutput_eagain(struct inpcb *); 776 #endif 777 #endif /* _KERNEL */ 778 779 #endif /* !_NETINET_IN_PCB_H_ */ 780