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/_lock.h> 45 #include <sys/_mutex.h> 46 #include <sys/_rwlock.h> 47 #include <net/route.h> 48 49 #ifdef _KERNEL 50 #include <sys/lock.h> 51 #include <sys/rwlock.h> 52 #include <net/vnet.h> 53 #include <vm/uma.h> 54 #endif 55 56 #define in6pcb inpcb /* for KAME src sync over BSD*'s */ 57 #define in6p_sp inp_sp /* for KAME src sync over BSD*'s */ 58 59 /* 60 * struct inpcb is the common protocol control block structure used in most 61 * IP transport protocols. 62 * 63 * Pointers to local and foreign host table entries, local and foreign socket 64 * numbers, and pointers up (to a socket structure) and down (to a 65 * protocol-specific control block) are stored here. 66 */ 67 LIST_HEAD(inpcbhead, inpcb); 68 LIST_HEAD(inpcbporthead, inpcbport); 69 typedef uint64_t inp_gen_t; 70 71 /* 72 * PCB with AF_INET6 null bind'ed laddr can receive AF_INET input packet. 73 * So, AF_INET6 null laddr is also used as AF_INET null laddr, by utilizing 74 * the following structure. 75 */ 76 struct in_addr_4in6 { 77 u_int32_t ia46_pad32[3]; 78 struct in_addr ia46_addr4; 79 }; 80 81 /* 82 * NOTE: ipv6 addrs should be 64-bit aligned, per RFC 2553. in_conninfo has 83 * some extra padding to accomplish this. 84 * NOTE 2: tcp_syncache.c uses first 5 32-bit words, which identify fport, 85 * lport, faddr to generate hash, so these fields shouldn't be moved. 86 */ 87 struct in_endpoints { 88 u_int16_t ie_fport; /* foreign port */ 89 u_int16_t ie_lport; /* local port */ 90 /* protocol dependent part, local and foreign addr */ 91 union { 92 /* foreign host table entry */ 93 struct in_addr_4in6 ie46_foreign; 94 struct in6_addr ie6_foreign; 95 } ie_dependfaddr; 96 union { 97 /* local host table entry */ 98 struct in_addr_4in6 ie46_local; 99 struct in6_addr ie6_local; 100 } ie_dependladdr; 101 u_int32_t ie6_zoneid; /* scope zone id */ 102 }; 103 #define ie_faddr ie_dependfaddr.ie46_foreign.ia46_addr4 104 #define ie_laddr ie_dependladdr.ie46_local.ia46_addr4 105 #define ie6_faddr ie_dependfaddr.ie6_foreign 106 #define ie6_laddr ie_dependladdr.ie6_local 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 124 125 #define inc_isipv6 inc_flags /* temp compatibility */ 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. A 140 * few fields are protected by multiple locks as indicated in the locking notes 141 * below. For these fields, all of the listed locks must be write-locked for 142 * any modifications. However, these fields can be safely read while any one of 143 * the listed locks are read-locked. This model can permit greater concurrency 144 * for read operations. For example, connections can be looked up while only 145 * holding a read lock on the global pcblist lock. This is important for 146 * performance when attempting to find the connection for a packet given its IP 147 * and port tuple. 148 * 149 * One noteworthy exception is that the global pcbinfo lock follows a different 150 * set of rules in relation to the inp_list field. Rather than being 151 * write-locked for modifications and read-locked for list iterations, it must 152 * be read-locked during modifications and write-locked during list iterations. 153 * This ensures that the relatively rare global list iterations safely walk a 154 * stable snapshot of connections while allowing more common list modifications 155 * to safely grab the pcblist lock just while adding or removing a connection 156 * from the global list. 157 * 158 * Key: 159 * (b) - Protected by the hpts lock. 160 * (c) - Constant after initialization 161 * (g) - Protected by the pcbgroup lock 162 * (i) - Protected by the inpcb lock 163 * (p) - Protected by the pcbinfo lock for the inpcb 164 * (l) - Protected by the pcblist lock for the inpcb 165 * (h) - Protected by the pcbhash lock for the inpcb 166 * (s) - Protected by another subsystem's locks 167 * (x) - Undefined locking 168 * 169 * Notes on the tcp_hpts: 170 * 171 * First Hpts lock order is 172 * 1) INP_WLOCK() 173 * 2) HPTS_LOCK() i.e. hpts->pmtx 174 * 175 * To insert a TCB on the hpts you *must* be holding the INP_WLOCK(). 176 * You may check the inp->inp_in_hpts flag without the hpts lock. 177 * The hpts is the only one that will clear this flag holding 178 * only the hpts lock. This means that in your tcp_output() 179 * routine when you test for the inp_in_hpts flag to be 1 180 * it may be transitioning to 0 (by the hpts). 181 * That's ok since that will just mean an extra call to tcp_output 182 * that most likely will find the call you executed 183 * (when the mis-match occured) will have put the TCB back 184 * on the hpts and it will return. If your 185 * call did not add the inp back to the hpts then you will either 186 * over-send or the cwnd will block you from sending more. 187 * 188 * Note you should also be holding the INP_WLOCK() when you 189 * call the remove from the hpts as well. Though usually 190 * you are either doing this from a timer, where you need and have 191 * the INP_WLOCK() or from destroying your TCB where again 192 * you should already have the INP_WLOCK(). 193 * 194 * The inp_hpts_cpu, inp_hpts_cpu_set, inp_input_cpu and 195 * inp_input_cpu_set fields are controlled completely by 196 * the hpts. Do not ever set these. The inp_hpts_cpu_set 197 * and inp_input_cpu_set fields indicate if the hpts has 198 * setup the respective cpu field. It is advised if this 199 * field is 0, to enqueue the packet with the appropriate 200 * hpts_immediate() call. If the _set field is 1, then 201 * you may compare the inp_*_cpu field to the curcpu and 202 * may want to again insert onto the hpts if these fields 203 * are not equal (i.e. you are not on the expected CPU). 204 * 205 * A note on inp_hpts_calls and inp_input_calls, these 206 * flags are set when the hpts calls either the output 207 * or do_segment routines respectively. If the routine 208 * being called wants to use this, then it needs to 209 * clear the flag before returning. The hpts will not 210 * clear the flag. The flags can be used to tell if 211 * the hpts is the function calling the respective 212 * routine. 213 * 214 * A few other notes: 215 * 216 * When a read lock is held, stability of the field is guaranteed; to write 217 * to a field, a write lock must generally be held. 218 * 219 * netinet/netinet6-layer code should not assume that the inp_socket pointer 220 * is safe to dereference without inp_lock being held, even for protocols 221 * other than TCP (where the inpcb persists during TIMEWAIT even after the 222 * socket has been freed), or there may be close(2)-related races. 223 * 224 * The inp_vflag field is overloaded, and would otherwise ideally be (c). 225 * 226 * TODO: Currently only the TCP stack is leveraging the global pcbinfo lock 227 * read-lock usage during modification, this model can be applied to other 228 * protocols (especially SCTP). 229 */ 230 struct icmp6_filter; 231 struct inpcbpolicy; 232 struct m_snd_tag; 233 struct inpcb { 234 /* Cache line #1 (amd64) */ 235 LIST_ENTRY(inpcb) inp_hash; /* (h/i) hash list */ 236 LIST_ENTRY(inpcb) inp_pcbgrouphash; /* (g/i) hash list */ 237 struct rwlock inp_lock; 238 /* Cache line #2 (amd64) */ 239 #define inp_start_zero inp_hpts 240 #define inp_zero_size (sizeof(struct inpcb) - \ 241 offsetof(struct inpcb, inp_start_zero)) 242 TAILQ_ENTRY(inpcb) inp_hpts; /* pacing out queue next lock(b) */ 243 244 uint32_t inp_hpts_request; /* Current hpts request, zero if 245 * fits in the pacing window (i&b). */ 246 /* 247 * Note the next fields are protected by a 248 * different lock (hpts-lock). This means that 249 * they must correspond in size to the smallest 250 * protectable bit field (uint8_t on x86, and 251 * other platfomrs potentially uint32_t?). Also 252 * since CPU switches can occur at different times the two 253 * fields can *not* be collapsed into a signal bit field. 254 */ 255 #if defined(__amd64__) || defined(__i386__) 256 volatile uint8_t inp_in_hpts; /* on output hpts (lock b) */ 257 volatile uint8_t inp_in_input; /* on input hpts (lock b) */ 258 #else 259 volatile uint32_t inp_in_hpts; /* on output hpts (lock b) */ 260 volatile uint32_t inp_in_input; /* on input hpts (lock b) */ 261 #endif 262 volatile uint16_t inp_hpts_cpu; /* Lock (i) */ 263 u_int inp_refcount; /* (i) refcount */ 264 int inp_flags; /* (i) generic IP/datagram flags */ 265 int inp_flags2; /* (i) generic IP/datagram flags #2*/ 266 volatile uint16_t inp_input_cpu; /* Lock (i) */ 267 volatile uint8_t inp_hpts_cpu_set :1, /* on output hpts (i) */ 268 inp_input_cpu_set : 1, /* on input hpts (i) */ 269 inp_hpts_calls :1, /* (i) from output hpts */ 270 inp_input_calls :1, /* (i) from input hpts */ 271 inp_spare_bits2 : 4; 272 uint8_t inp_spare_byte; /* Compiler hole */ 273 void *inp_ppcb; /* (i) pointer to per-protocol pcb */ 274 struct socket *inp_socket; /* (i) back pointer to socket */ 275 uint32_t inp_hptsslot; /* Hpts wheel slot this tcb is Lock(i&b) */ 276 uint32_t inp_hpts_drop_reas; /* reason we are dropping the PCB (lock i&b) */ 277 TAILQ_ENTRY(inpcb) inp_input; /* pacing in queue next lock(b) */ 278 struct inpcbinfo *inp_pcbinfo; /* (c) PCB list info */ 279 struct inpcbgroup *inp_pcbgroup; /* (g/i) PCB group list */ 280 LIST_ENTRY(inpcb) inp_pcbgroup_wild; /* (g/i/h) group wildcard entry */ 281 struct ucred *inp_cred; /* (c) cache of socket cred */ 282 u_int32_t inp_flow; /* (i) IPv6 flow information */ 283 u_char inp_vflag; /* (i) IP version flag (v4/v6) */ 284 u_char inp_ip_ttl; /* (i) time to live proto */ 285 u_char inp_ip_p; /* (c) protocol proto */ 286 u_char inp_ip_minttl; /* (i) minimum TTL or drop */ 287 uint32_t inp_flowid; /* (x) flow id / queue id */ 288 struct m_snd_tag *inp_snd_tag; /* (i) send tag for outgoing mbufs */ 289 uint32_t inp_flowtype; /* (x) M_HASHTYPE value */ 290 uint32_t inp_rss_listen_bucket; /* (x) overridden RSS listen bucket */ 291 292 /* Local and foreign ports, local and foreign addr. */ 293 struct in_conninfo inp_inc; /* (i) list for PCB's local port */ 294 295 /* MAC and IPSEC policy information. */ 296 struct label *inp_label; /* (i) MAC label */ 297 struct inpcbpolicy *inp_sp; /* (s) for IPSEC */ 298 299 /* Protocol-dependent part; options. */ 300 struct { 301 u_char inp_ip_tos; /* (i) type of service proto */ 302 struct mbuf *inp_options; /* (i) IP options */ 303 struct ip_moptions *inp_moptions; /* (i) mcast options */ 304 }; 305 struct { 306 /* (i) IP options */ 307 struct mbuf *in6p_options; 308 /* (i) IP6 options for outgoing packets */ 309 struct ip6_pktopts *in6p_outputopts; 310 /* (i) IP multicast options */ 311 struct ip6_moptions *in6p_moptions; 312 /* (i) ICMPv6 code type filter */ 313 struct icmp6_filter *in6p_icmp6filt; 314 /* (i) IPV6_CHECKSUM setsockopt */ 315 int in6p_cksum; 316 short in6p_hops; 317 }; 318 LIST_ENTRY(inpcb) inp_portlist; /* (i/h) */ 319 struct inpcbport *inp_phd; /* (i/h) head of this list */ 320 inp_gen_t inp_gencnt; /* (c) generation count */ 321 struct llentry *inp_lle; /* cached L2 information */ 322 rt_gen_t inp_rt_cookie; /* generation for route entry */ 323 union { /* cached L3 information */ 324 struct route inp_route; 325 struct route_in6 inp_route6; 326 }; 327 LIST_ENTRY(inpcb) inp_list; /* (p/l) list for all PCBs for proto */ 328 /* (p[w]) for list iteration */ 329 /* (p[r]/l) for addition/removal */ 330 }; 331 #endif /* _KERNEL */ 332 333 #define inp_fport inp_inc.inc_fport 334 #define inp_lport inp_inc.inc_lport 335 #define inp_faddr inp_inc.inc_faddr 336 #define inp_laddr inp_inc.inc_laddr 337 338 #define in6p_faddr inp_inc.inc6_faddr 339 #define in6p_laddr inp_inc.inc6_laddr 340 #define in6p_zoneid inp_inc.inc6_zoneid 341 #define in6p_flowinfo inp_flow 342 343 #define inp_vnet inp_pcbinfo->ipi_vnet 344 345 /* 346 * The range of the generation count, as used in this implementation, is 9e19. 347 * We would have to create 300 billion connections per second for this number 348 * to roll over in a year. This seems sufficiently unlikely that we simply 349 * don't concern ourselves with that possibility. 350 */ 351 352 /* 353 * Interface exported to userland by various protocols which use inpcbs. Hack 354 * alert -- only define if struct xsocket is in scope. 355 * Fields prefixed with "xi_" are unique to this structure, and the rest 356 * match fields in the struct inpcb, to ease coding and porting. 357 * 358 * Legend: 359 * (s) - used by userland utilities in src 360 * (p) - used by utilities in ports 361 * (3) - is known to be used by third party software not in ports 362 * (n) - no known usage 363 */ 364 #ifdef _SYS_SOCKETVAR_H_ 365 struct xinpcb { 366 size_t xi_len; /* length of this structure */ 367 struct xsocket xi_socket; /* (s,p) */ 368 struct in_conninfo inp_inc; /* (s,p) */ 369 uint64_t inp_gencnt; /* (s,p) */ 370 union { 371 void *inp_ppcb; /* (s) netstat(1) */ 372 int64_t ph_ppcb; 373 }; 374 int64_t inp_spare64[4]; 375 uint32_t inp_flow; /* (s) */ 376 uint32_t inp_flowid; /* (s) */ 377 uint32_t inp_flowtype; /* (s) */ 378 int32_t inp_flags; /* (s,p) */ 379 int32_t inp_flags2; /* (s) */ 380 int32_t inp_rss_listen_bucket; /* (n) */ 381 int32_t in6p_cksum; /* (n) */ 382 int32_t inp_spare32[4]; 383 uint16_t in6p_hops; /* (n) */ 384 uint8_t inp_ip_tos; /* (n) */ 385 int8_t pad8; 386 uint8_t inp_vflag; /* (s,p) */ 387 uint8_t inp_ip_ttl; /* (n) */ 388 uint8_t inp_ip_p; /* (n) */ 389 uint8_t inp_ip_minttl; /* (n) */ 390 int8_t inp_spare8[4]; 391 } __aligned(8); 392 393 struct xinpgen { 394 size_t xig_len; /* length of this structure */ 395 u_int xig_count; /* number of PCBs at this time */ 396 inp_gen_t xig_gen; /* generation count at this time */ 397 so_gen_t xig_sogen; /* socket generation count this time */ 398 } __aligned(8); 399 #ifdef _KERNEL 400 void in_pcbtoxinpcb(const struct inpcb *, struct xinpcb *); 401 #endif 402 #endif /* _SYS_SOCKETVAR_H_ */ 403 404 struct inpcbport { 405 LIST_ENTRY(inpcbport) phd_hash; 406 struct inpcbhead phd_pcblist; 407 u_short phd_port; 408 }; 409 410 /*- 411 * Global data structure for each high-level protocol (UDP, TCP, ...) in both 412 * IPv4 and IPv6. Holds inpcb lists and information for managing them. 413 * 414 * Each pcbinfo is protected by three locks: ipi_lock, ipi_hash_lock and 415 * ipi_list_lock: 416 * - ipi_lock covering the global pcb list stability during loop iteration, 417 * - ipi_hash_lock covering the hashed lookup tables, 418 * - ipi_list_lock covering mutable global fields (such as the global 419 * pcb list) 420 * 421 * The lock order is: 422 * 423 * ipi_lock (before) 424 * inpcb locks (before) 425 * ipi_list locks (before) 426 * {ipi_hash_lock, pcbgroup locks} 427 * 428 * Locking key: 429 * 430 * (c) Constant or nearly constant after initialisation 431 * (g) Locked by ipi_lock 432 * (l) Locked by ipi_list_lock 433 * (h) Read using either ipi_hash_lock or inpcb lock; write requires both 434 * (p) Protected by one or more pcbgroup locks 435 * (x) Synchronisation properties poorly defined 436 */ 437 struct inpcbinfo { 438 /* 439 * Global lock protecting full inpcb list traversal 440 */ 441 struct rwlock ipi_lock; 442 443 /* 444 * Global list of inpcbs on the protocol. 445 */ 446 struct inpcbhead *ipi_listhead; /* (g/l) */ 447 u_int ipi_count; /* (l) */ 448 449 /* 450 * Generation count -- incremented each time a connection is allocated 451 * or freed. 452 */ 453 u_quad_t ipi_gencnt; /* (l) */ 454 455 /* 456 * Fields associated with port lookup and allocation. 457 */ 458 u_short ipi_lastport; /* (x) */ 459 u_short ipi_lastlow; /* (x) */ 460 u_short ipi_lasthi; /* (x) */ 461 462 /* 463 * UMA zone from which inpcbs are allocated for this protocol. 464 */ 465 struct uma_zone *ipi_zone; /* (c) */ 466 467 /* 468 * Connection groups associated with this protocol. These fields are 469 * constant, but pcbgroup structures themselves are protected by 470 * per-pcbgroup locks. 471 */ 472 struct inpcbgroup *ipi_pcbgroups; /* (c) */ 473 u_int ipi_npcbgroups; /* (c) */ 474 u_int ipi_hashfields; /* (c) */ 475 476 /* 477 * Global lock protecting non-pcbgroup hash lookup tables. 478 */ 479 struct rwlock ipi_hash_lock; 480 481 /* 482 * Global hash of inpcbs, hashed by local and foreign addresses and 483 * port numbers. 484 */ 485 struct inpcbhead *ipi_hashbase; /* (h) */ 486 u_long ipi_hashmask; /* (h) */ 487 488 /* 489 * Global hash of inpcbs, hashed by only local port number. 490 */ 491 struct inpcbporthead *ipi_porthashbase; /* (h) */ 492 u_long ipi_porthashmask; /* (h) */ 493 494 /* 495 * List of wildcard inpcbs for use with pcbgroups. In the past, was 496 * per-pcbgroup but is now global. All pcbgroup locks must be held 497 * to modify the list, so any is sufficient to read it. 498 */ 499 struct inpcbhead *ipi_wildbase; /* (p) */ 500 u_long ipi_wildmask; /* (p) */ 501 502 /* 503 * Pointer to network stack instance 504 */ 505 struct vnet *ipi_vnet; /* (c) */ 506 507 /* 508 * general use 2 509 */ 510 void *ipi_pspare[2]; 511 512 /* 513 * Global lock protecting global inpcb list, inpcb count, etc. 514 */ 515 struct rwlock ipi_list_lock; 516 }; 517 518 #ifdef _KERNEL 519 /* 520 * Connection groups hold sets of connections that have similar CPU/thread 521 * affinity. Each connection belongs to exactly one connection group. 522 */ 523 struct inpcbgroup { 524 /* 525 * Per-connection group hash of inpcbs, hashed by local and foreign 526 * addresses and port numbers. 527 */ 528 struct inpcbhead *ipg_hashbase; /* (c) */ 529 u_long ipg_hashmask; /* (c) */ 530 531 /* 532 * Notional affinity of this pcbgroup. 533 */ 534 u_int ipg_cpu; /* (p) */ 535 536 /* 537 * Per-connection group lock, not to be confused with ipi_lock. 538 * Protects the hash table hung off the group, but also the global 539 * wildcard list in inpcbinfo. 540 */ 541 struct mtx ipg_lock; 542 } __aligned(CACHE_LINE_SIZE); 543 544 #define INP_LOCK_INIT(inp, d, t) \ 545 rw_init_flags(&(inp)->inp_lock, (t), RW_RECURSE | RW_DUPOK) 546 #define INP_LOCK_DESTROY(inp) rw_destroy(&(inp)->inp_lock) 547 #define INP_RLOCK(inp) rw_rlock(&(inp)->inp_lock) 548 #define INP_WLOCK(inp) rw_wlock(&(inp)->inp_lock) 549 #define INP_TRY_RLOCK(inp) rw_try_rlock(&(inp)->inp_lock) 550 #define INP_TRY_WLOCK(inp) rw_try_wlock(&(inp)->inp_lock) 551 #define INP_RUNLOCK(inp) rw_runlock(&(inp)->inp_lock) 552 #define INP_WUNLOCK(inp) rw_wunlock(&(inp)->inp_lock) 553 #define INP_TRY_UPGRADE(inp) rw_try_upgrade(&(inp)->inp_lock) 554 #define INP_DOWNGRADE(inp) rw_downgrade(&(inp)->inp_lock) 555 #define INP_WLOCKED(inp) rw_wowned(&(inp)->inp_lock) 556 #define INP_LOCK_ASSERT(inp) rw_assert(&(inp)->inp_lock, RA_LOCKED) 557 #define INP_RLOCK_ASSERT(inp) rw_assert(&(inp)->inp_lock, RA_RLOCKED) 558 #define INP_WLOCK_ASSERT(inp) rw_assert(&(inp)->inp_lock, RA_WLOCKED) 559 #define INP_UNLOCK_ASSERT(inp) rw_assert(&(inp)->inp_lock, RA_UNLOCKED) 560 561 /* 562 * These locking functions are for inpcb consumers outside of sys/netinet, 563 * more specifically, they were added for the benefit of TOE drivers. The 564 * macros are reserved for use by the stack. 565 */ 566 void inp_wlock(struct inpcb *); 567 void inp_wunlock(struct inpcb *); 568 void inp_rlock(struct inpcb *); 569 void inp_runlock(struct inpcb *); 570 571 #ifdef INVARIANT_SUPPORT 572 void inp_lock_assert(struct inpcb *); 573 void inp_unlock_assert(struct inpcb *); 574 #else 575 #define inp_lock_assert(inp) do {} while (0) 576 #define inp_unlock_assert(inp) do {} while (0) 577 #endif 578 579 void inp_apply_all(void (*func)(struct inpcb *, void *), void *arg); 580 int inp_ip_tos_get(const struct inpcb *inp); 581 void inp_ip_tos_set(struct inpcb *inp, int val); 582 struct socket * 583 inp_inpcbtosocket(struct inpcb *inp); 584 struct tcpcb * 585 inp_inpcbtotcpcb(struct inpcb *inp); 586 void inp_4tuple_get(struct inpcb *inp, uint32_t *laddr, uint16_t *lp, 587 uint32_t *faddr, uint16_t *fp); 588 short inp_so_options(const struct inpcb *inp); 589 590 #endif /* _KERNEL */ 591 592 #define INP_INFO_LOCK_INIT(ipi, d) \ 593 rw_init_flags(&(ipi)->ipi_lock, (d), RW_RECURSE) 594 #define INP_INFO_LOCK_DESTROY(ipi) rw_destroy(&(ipi)->ipi_lock) 595 #define INP_INFO_RLOCK(ipi) rw_rlock(&(ipi)->ipi_lock) 596 #define INP_INFO_WLOCK(ipi) rw_wlock(&(ipi)->ipi_lock) 597 #define INP_INFO_TRY_RLOCK(ipi) rw_try_rlock(&(ipi)->ipi_lock) 598 #define INP_INFO_TRY_WLOCK(ipi) rw_try_wlock(&(ipi)->ipi_lock) 599 #define INP_INFO_TRY_UPGRADE(ipi) rw_try_upgrade(&(ipi)->ipi_lock) 600 #define INP_INFO_WLOCKED(ipi) rw_wowned(&(ipi)->ipi_lock) 601 #define INP_INFO_RUNLOCK(ipi) rw_runlock(&(ipi)->ipi_lock) 602 #define INP_INFO_WUNLOCK(ipi) rw_wunlock(&(ipi)->ipi_lock) 603 #define INP_INFO_LOCK_ASSERT(ipi) rw_assert(&(ipi)->ipi_lock, RA_LOCKED) 604 #define INP_INFO_RLOCK_ASSERT(ipi) rw_assert(&(ipi)->ipi_lock, RA_RLOCKED) 605 #define INP_INFO_WLOCK_ASSERT(ipi) rw_assert(&(ipi)->ipi_lock, RA_WLOCKED) 606 #define INP_INFO_UNLOCK_ASSERT(ipi) rw_assert(&(ipi)->ipi_lock, RA_UNLOCKED) 607 608 #define INP_LIST_LOCK_INIT(ipi, d) \ 609 rw_init_flags(&(ipi)->ipi_list_lock, (d), 0) 610 #define INP_LIST_LOCK_DESTROY(ipi) rw_destroy(&(ipi)->ipi_list_lock) 611 #define INP_LIST_RLOCK(ipi) rw_rlock(&(ipi)->ipi_list_lock) 612 #define INP_LIST_WLOCK(ipi) rw_wlock(&(ipi)->ipi_list_lock) 613 #define INP_LIST_TRY_RLOCK(ipi) rw_try_rlock(&(ipi)->ipi_list_lock) 614 #define INP_LIST_TRY_WLOCK(ipi) rw_try_wlock(&(ipi)->ipi_list_lock) 615 #define INP_LIST_TRY_UPGRADE(ipi) rw_try_upgrade(&(ipi)->ipi_list_lock) 616 #define INP_LIST_RUNLOCK(ipi) rw_runlock(&(ipi)->ipi_list_lock) 617 #define INP_LIST_WUNLOCK(ipi) rw_wunlock(&(ipi)->ipi_list_lock) 618 #define INP_LIST_LOCK_ASSERT(ipi) \ 619 rw_assert(&(ipi)->ipi_list_lock, RA_LOCKED) 620 #define INP_LIST_RLOCK_ASSERT(ipi) \ 621 rw_assert(&(ipi)->ipi_list_lock, RA_RLOCKED) 622 #define INP_LIST_WLOCK_ASSERT(ipi) \ 623 rw_assert(&(ipi)->ipi_list_lock, RA_WLOCKED) 624 #define INP_LIST_UNLOCK_ASSERT(ipi) \ 625 rw_assert(&(ipi)->ipi_list_lock, RA_UNLOCKED) 626 627 #define INP_HASH_LOCK_INIT(ipi, d) \ 628 rw_init_flags(&(ipi)->ipi_hash_lock, (d), 0) 629 #define INP_HASH_LOCK_DESTROY(ipi) rw_destroy(&(ipi)->ipi_hash_lock) 630 #define INP_HASH_RLOCK(ipi) rw_rlock(&(ipi)->ipi_hash_lock) 631 #define INP_HASH_WLOCK(ipi) rw_wlock(&(ipi)->ipi_hash_lock) 632 #define INP_HASH_RUNLOCK(ipi) rw_runlock(&(ipi)->ipi_hash_lock) 633 #define INP_HASH_WUNLOCK(ipi) rw_wunlock(&(ipi)->ipi_hash_lock) 634 #define INP_HASH_LOCK_ASSERT(ipi) rw_assert(&(ipi)->ipi_hash_lock, \ 635 RA_LOCKED) 636 #define INP_HASH_WLOCK_ASSERT(ipi) rw_assert(&(ipi)->ipi_hash_lock, \ 637 RA_WLOCKED) 638 639 #define INP_GROUP_LOCK_INIT(ipg, d) mtx_init(&(ipg)->ipg_lock, (d), NULL, \ 640 MTX_DEF | MTX_DUPOK) 641 #define INP_GROUP_LOCK_DESTROY(ipg) mtx_destroy(&(ipg)->ipg_lock) 642 643 #define INP_GROUP_LOCK(ipg) mtx_lock(&(ipg)->ipg_lock) 644 #define INP_GROUP_LOCK_ASSERT(ipg) mtx_assert(&(ipg)->ipg_lock, MA_OWNED) 645 #define INP_GROUP_UNLOCK(ipg) mtx_unlock(&(ipg)->ipg_lock) 646 647 #define INP_PCBHASH(faddr, lport, fport, mask) \ 648 (((faddr) ^ ((faddr) >> 16) ^ ntohs((lport) ^ (fport))) & (mask)) 649 #define INP_PCBPORTHASH(lport, mask) \ 650 (ntohs((lport)) & (mask)) 651 #define INP6_PCBHASHKEY(faddr) ((faddr)->s6_addr32[3]) 652 653 /* 654 * Flags for inp_vflags -- historically version flags only 655 */ 656 #define INP_IPV4 0x1 657 #define INP_IPV6 0x2 658 #define INP_IPV6PROTO 0x4 /* opened under IPv6 protocol */ 659 660 /* 661 * Flags for inp_flags. 662 */ 663 #define INP_RECVOPTS 0x00000001 /* receive incoming IP options */ 664 #define INP_RECVRETOPTS 0x00000002 /* receive IP options for reply */ 665 #define INP_RECVDSTADDR 0x00000004 /* receive IP dst address */ 666 #define INP_HDRINCL 0x00000008 /* user supplies entire IP header */ 667 #define INP_HIGHPORT 0x00000010 /* user wants "high" port binding */ 668 #define INP_LOWPORT 0x00000020 /* user wants "low" port binding */ 669 #define INP_ANONPORT 0x00000040 /* port chosen for user */ 670 #define INP_RECVIF 0x00000080 /* receive incoming interface */ 671 #define INP_MTUDISC 0x00000100 /* user can do MTU discovery */ 672 /* 0x000200 unused: was INP_FAITH */ 673 #define INP_RECVTTL 0x00000400 /* receive incoming IP TTL */ 674 #define INP_DONTFRAG 0x00000800 /* don't fragment packet */ 675 #define INP_BINDANY 0x00001000 /* allow bind to any address */ 676 #define INP_INHASHLIST 0x00002000 /* in_pcbinshash() has been called */ 677 #define INP_RECVTOS 0x00004000 /* receive incoming IP TOS */ 678 #define IN6P_IPV6_V6ONLY 0x00008000 /* restrict AF_INET6 socket for v6 */ 679 #define IN6P_PKTINFO 0x00010000 /* receive IP6 dst and I/F */ 680 #define IN6P_HOPLIMIT 0x00020000 /* receive hoplimit */ 681 #define IN6P_HOPOPTS 0x00040000 /* receive hop-by-hop options */ 682 #define IN6P_DSTOPTS 0x00080000 /* receive dst options after rthdr */ 683 #define IN6P_RTHDR 0x00100000 /* receive routing header */ 684 #define IN6P_RTHDRDSTOPTS 0x00200000 /* receive dstoptions before rthdr */ 685 #define IN6P_TCLASS 0x00400000 /* receive traffic class value */ 686 #define IN6P_AUTOFLOWLABEL 0x00800000 /* attach flowlabel automatically */ 687 #define INP_TIMEWAIT 0x01000000 /* in TIMEWAIT, ppcb is tcptw */ 688 #define INP_ONESBCAST 0x02000000 /* send all-ones broadcast */ 689 #define INP_DROPPED 0x04000000 /* protocol drop flag */ 690 #define INP_SOCKREF 0x08000000 /* strong socket reference */ 691 #define INP_RESERVED_0 0x10000000 /* reserved field */ 692 #define INP_RESERVED_1 0x20000000 /* reserved field */ 693 #define IN6P_RFC2292 0x40000000 /* used RFC2292 API on the socket */ 694 #define IN6P_MTU 0x80000000 /* receive path MTU */ 695 696 #define INP_CONTROLOPTS (INP_RECVOPTS|INP_RECVRETOPTS|INP_RECVDSTADDR|\ 697 INP_RECVIF|INP_RECVTTL|INP_RECVTOS|\ 698 IN6P_PKTINFO|IN6P_HOPLIMIT|IN6P_HOPOPTS|\ 699 IN6P_DSTOPTS|IN6P_RTHDR|IN6P_RTHDRDSTOPTS|\ 700 IN6P_TCLASS|IN6P_AUTOFLOWLABEL|IN6P_RFC2292|\ 701 IN6P_MTU) 702 703 /* 704 * Flags for inp_flags2. 705 */ 706 #define INP_LLE_VALID 0x00000001 /* cached lle is valid */ 707 #define INP_RT_VALID 0x00000002 /* cached rtentry is valid */ 708 #define INP_PCBGROUPWILD 0x00000004 /* in pcbgroup wildcard list */ 709 #define INP_REUSEPORT 0x00000008 /* SO_REUSEPORT option is set */ 710 #define INP_FREED 0x00000010 /* inp itself is not valid */ 711 #define INP_REUSEADDR 0x00000020 /* SO_REUSEADDR option is set */ 712 #define INP_BINDMULTI 0x00000040 /* IP_BINDMULTI option is set */ 713 #define INP_RSS_BUCKET_SET 0x00000080 /* IP_RSS_LISTEN_BUCKET is set */ 714 #define INP_RECVFLOWID 0x00000100 /* populate recv datagram with flow info */ 715 #define INP_RECVRSSBUCKETID 0x00000200 /* populate recv datagram with bucket id */ 716 #define INP_RATE_LIMIT_CHANGED 0x00000400 /* rate limit needs attention */ 717 #define INP_ORIGDSTADDR 0x00000800 /* receive IP dst address/port */ 718 #define INP_CANNOT_DO_ECN 0x00001000 /* The stack does not do ECN */ 719 720 /* 721 * Flags passed to in_pcblookup*() functions. 722 */ 723 #define INPLOOKUP_WILDCARD 0x00000001 /* Allow wildcard sockets. */ 724 #define INPLOOKUP_RLOCKPCB 0x00000002 /* Return inpcb read-locked. */ 725 #define INPLOOKUP_WLOCKPCB 0x00000004 /* Return inpcb write-locked. */ 726 727 #define INPLOOKUP_MASK (INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB | \ 728 INPLOOKUP_WLOCKPCB) 729 730 #define sotoinpcb(so) ((struct inpcb *)(so)->so_pcb) 731 #define sotoin6pcb(so) sotoinpcb(so) /* for KAME src sync over BSD*'s */ 732 733 #define INP_SOCKAF(so) so->so_proto->pr_domain->dom_family 734 735 #define INP_CHECK_SOCKAF(so, af) (INP_SOCKAF(so) == af) 736 737 /* 738 * Constants for pcbinfo.ipi_hashfields. 739 */ 740 #define IPI_HASHFIELDS_NONE 0 741 #define IPI_HASHFIELDS_2TUPLE 1 742 #define IPI_HASHFIELDS_4TUPLE 2 743 744 #ifdef _KERNEL 745 VNET_DECLARE(int, ipport_reservedhigh); 746 VNET_DECLARE(int, ipport_reservedlow); 747 VNET_DECLARE(int, ipport_lowfirstauto); 748 VNET_DECLARE(int, ipport_lowlastauto); 749 VNET_DECLARE(int, ipport_firstauto); 750 VNET_DECLARE(int, ipport_lastauto); 751 VNET_DECLARE(int, ipport_hifirstauto); 752 VNET_DECLARE(int, ipport_hilastauto); 753 VNET_DECLARE(int, ipport_randomized); 754 VNET_DECLARE(int, ipport_randomcps); 755 VNET_DECLARE(int, ipport_randomtime); 756 VNET_DECLARE(int, ipport_stoprandom); 757 VNET_DECLARE(int, ipport_tcpallocs); 758 759 #define V_ipport_reservedhigh VNET(ipport_reservedhigh) 760 #define V_ipport_reservedlow VNET(ipport_reservedlow) 761 #define V_ipport_lowfirstauto VNET(ipport_lowfirstauto) 762 #define V_ipport_lowlastauto VNET(ipport_lowlastauto) 763 #define V_ipport_firstauto VNET(ipport_firstauto) 764 #define V_ipport_lastauto VNET(ipport_lastauto) 765 #define V_ipport_hifirstauto VNET(ipport_hifirstauto) 766 #define V_ipport_hilastauto VNET(ipport_hilastauto) 767 #define V_ipport_randomized VNET(ipport_randomized) 768 #define V_ipport_randomcps VNET(ipport_randomcps) 769 #define V_ipport_randomtime VNET(ipport_randomtime) 770 #define V_ipport_stoprandom VNET(ipport_stoprandom) 771 #define V_ipport_tcpallocs VNET(ipport_tcpallocs) 772 773 void in_pcbinfo_destroy(struct inpcbinfo *); 774 void in_pcbinfo_init(struct inpcbinfo *, const char *, struct inpcbhead *, 775 int, int, char *, uma_init, u_int); 776 777 int in_pcbbind_check_bindmulti(const struct inpcb *ni, 778 const struct inpcb *oi); 779 780 struct inpcbgroup * 781 in_pcbgroup_byhash(struct inpcbinfo *, u_int, uint32_t); 782 struct inpcbgroup * 783 in_pcbgroup_byinpcb(struct inpcb *); 784 struct inpcbgroup * 785 in_pcbgroup_bytuple(struct inpcbinfo *, struct in_addr, u_short, 786 struct in_addr, u_short); 787 void in_pcbgroup_destroy(struct inpcbinfo *); 788 int in_pcbgroup_enabled(struct inpcbinfo *); 789 void in_pcbgroup_init(struct inpcbinfo *, u_int, int); 790 void in_pcbgroup_remove(struct inpcb *); 791 void in_pcbgroup_update(struct inpcb *); 792 void in_pcbgroup_update_mbuf(struct inpcb *, struct mbuf *); 793 794 void in_pcbpurgeif0(struct inpcbinfo *, struct ifnet *); 795 int in_pcballoc(struct socket *, struct inpcbinfo *); 796 int in_pcbbind(struct inpcb *, struct sockaddr *, struct ucred *); 797 int in_pcb_lport(struct inpcb *, struct in_addr *, u_short *, 798 struct ucred *, int); 799 int in_pcbbind_setup(struct inpcb *, struct sockaddr *, in_addr_t *, 800 u_short *, struct ucred *); 801 int in_pcbconnect(struct inpcb *, struct sockaddr *, struct ucred *); 802 int in_pcbconnect_mbuf(struct inpcb *, struct sockaddr *, struct ucred *, 803 struct mbuf *); 804 int in_pcbconnect_setup(struct inpcb *, struct sockaddr *, in_addr_t *, 805 u_short *, in_addr_t *, u_short *, struct inpcb **, 806 struct ucred *); 807 void in_pcbdetach(struct inpcb *); 808 void in_pcbdisconnect(struct inpcb *); 809 void in_pcbdrop(struct inpcb *); 810 void in_pcbfree(struct inpcb *); 811 int in_pcbinshash(struct inpcb *); 812 int in_pcbinshash_nopcbgroup(struct inpcb *); 813 int in_pcbladdr(struct inpcb *, struct in_addr *, struct in_addr *, 814 struct ucred *); 815 struct inpcb * 816 in_pcblookup_local(struct inpcbinfo *, 817 struct in_addr, u_short, int, struct ucred *); 818 struct inpcb * 819 in_pcblookup(struct inpcbinfo *, struct in_addr, u_int, 820 struct in_addr, u_int, int, struct ifnet *); 821 struct inpcb * 822 in_pcblookup_mbuf(struct inpcbinfo *, struct in_addr, u_int, 823 struct in_addr, u_int, int, struct ifnet *, struct mbuf *); 824 void in_pcbnotifyall(struct inpcbinfo *pcbinfo, struct in_addr, 825 int, struct inpcb *(*)(struct inpcb *, int)); 826 void in_pcbref(struct inpcb *); 827 void in_pcbrehash(struct inpcb *); 828 void in_pcbrehash_mbuf(struct inpcb *, struct mbuf *); 829 int in_pcbrele(struct inpcb *); 830 int in_pcbrele_rlocked(struct inpcb *); 831 int in_pcbrele_wlocked(struct inpcb *); 832 void in_losing(struct inpcb *); 833 void in_pcbsetsolabel(struct socket *so); 834 int in_getpeeraddr(struct socket *so, struct sockaddr **nam); 835 int in_getsockaddr(struct socket *so, struct sockaddr **nam); 836 struct sockaddr * 837 in_sockaddr(in_port_t port, struct in_addr *addr); 838 void in_pcbsosetlabel(struct socket *so); 839 #ifdef RATELIMIT 840 int in_pcbattach_txrtlmt(struct inpcb *, struct ifnet *, uint32_t, uint32_t, uint32_t); 841 void in_pcbdetach_txrtlmt(struct inpcb *); 842 int in_pcbmodify_txrtlmt(struct inpcb *, uint32_t); 843 int in_pcbquery_txrtlmt(struct inpcb *, uint32_t *); 844 int in_pcbquery_txrlevel(struct inpcb *, uint32_t *); 845 void in_pcboutput_txrtlmt(struct inpcb *, struct ifnet *, struct mbuf *); 846 void in_pcboutput_eagain(struct inpcb *); 847 #endif 848 #endif /* _KERNEL */ 849 850 #endif /* !_NETINET_IN_PCB_H_ */ 851