1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1982, 1986, 1990, 1993 5 * The Regents of the University of California. All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the University nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 * 31 * @(#)socketvar.h 8.3 (Berkeley) 2/19/95 32 * 33 * $FreeBSD$ 34 */ 35 36 #ifndef _SYS_SOCKETVAR_H_ 37 #define _SYS_SOCKETVAR_H_ 38 39 /* 40 * Socket generation count type. Also used in xinpcb, xtcpcb, xunpcb. 41 */ 42 typedef uint64_t so_gen_t; 43 44 #if defined(_KERNEL) || defined(_WANT_SOCKET) 45 #include <sys/queue.h> /* for TAILQ macros */ 46 #include <sys/selinfo.h> /* for struct selinfo */ 47 #include <sys/_lock.h> 48 #include <sys/_mutex.h> 49 #include <sys/osd.h> 50 #include <sys/_sx.h> 51 #include <sys/sockbuf.h> 52 #ifdef _KERNEL 53 #include <sys/caprights.h> 54 #include <sys/sockopt.h> 55 #endif 56 57 struct vnet; 58 59 /* 60 * Kernel structure per socket. 61 * Contains send and receive buffer queues, 62 * handle on protocol and pointer to protocol 63 * private data and error information. 64 */ 65 typedef int so_upcall_t(struct socket *, void *, int); 66 typedef void so_dtor_t(struct socket *); 67 68 struct socket; 69 70 /*- 71 * Locking key to struct socket: 72 * (a) constant after allocation, no locking required. 73 * (b) locked by SOCK_LOCK(so). 74 * (cr) locked by SOCKBUF_LOCK(&so->so_rcv). 75 * (cs) locked by SOCKBUF_LOCK(&so->so_snd). 76 * (e) locked by SOLISTEN_LOCK() of corresponding listening socket. 77 * (f) not locked since integer reads/writes are atomic. 78 * (g) used only as a sleep/wakeup address, no value. 79 * (h) locked by global mutex so_global_mtx. 80 */ 81 TAILQ_HEAD(accept_queue, socket); 82 struct socket { 83 struct mtx so_lock; 84 volatile u_int so_count; /* (b / refcount) */ 85 struct selinfo so_rdsel; /* (b/cr) for so_rcv/so_comp */ 86 struct selinfo so_wrsel; /* (b/cs) for so_snd */ 87 short so_type; /* (a) generic type, see socket.h */ 88 int so_options; /* (b) from socket call, see socket.h */ 89 short so_linger; /* time to linger close(2) */ 90 short so_state; /* (b) internal state flags SS_* */ 91 void *so_pcb; /* protocol control block */ 92 struct vnet *so_vnet; /* (a) network stack instance */ 93 struct protosw *so_proto; /* (a) protocol handle */ 94 short so_timeo; /* (g) connection timeout */ 95 u_short so_error; /* (f) error affecting connection */ 96 struct sigio *so_sigio; /* [sg] information for async I/O or 97 out of band data (SIGURG) */ 98 struct ucred *so_cred; /* (a) user credentials */ 99 struct label *so_label; /* (b) MAC label for socket */ 100 /* NB: generation count must not be first. */ 101 so_gen_t so_gencnt; /* (h) generation count */ 102 void *so_emuldata; /* (b) private data for emulators */ 103 so_dtor_t *so_dtor; /* (b) optional destructor */ 104 struct osd osd; /* Object Specific extensions */ 105 /* 106 * so_fibnum, so_user_cookie and friends can be used to attach 107 * some user-specified metadata to a socket, which then can be 108 * used by the kernel for various actions. 109 * so_user_cookie is used by ipfw/dummynet. 110 */ 111 int so_fibnum; /* routing domain for this socket */ 112 uint32_t so_user_cookie; 113 114 int so_ts_clock; /* type of the clock used for timestamps */ 115 uint32_t so_max_pacing_rate; /* (f) TX rate limit in bytes/s */ 116 union { 117 /* Regular (data flow) socket. */ 118 struct { 119 /* (cr, cs) Receive and send buffers. */ 120 struct sockbuf so_rcv, so_snd; 121 122 /* (e) Our place on accept queue. */ 123 TAILQ_ENTRY(socket) so_list; 124 struct socket *so_listen; /* (b) */ 125 enum { 126 SQ_NONE = 0, 127 SQ_INCOMP = 0x0800, /* on sol_incomp */ 128 SQ_COMP = 0x1000, /* on sol_comp */ 129 } so_qstate; /* (b) */ 130 131 /* (b) cached MAC label for peer */ 132 struct label *so_peerlabel; 133 u_long so_oobmark; /* chars to oob mark */ 134 }; 135 /* 136 * Listening socket, where accepts occur, is so_listen in all 137 * subsidiary sockets. If so_listen is NULL, socket is not 138 * related to an accept. For a listening socket itself 139 * sol_incomp queues partially completed connections, while 140 * sol_comp is a queue of connections ready to be accepted. 141 * If a connection is aborted and it has so_listen set, then 142 * it has to be pulled out of either sol_incomp or sol_comp. 143 * We allow connections to queue up based on current queue 144 * lengths and limit on number of queued connections for this 145 * socket. 146 */ 147 struct { 148 /* (e) queue of partial unaccepted connections */ 149 struct accept_queue sol_incomp; 150 /* (e) queue of complete unaccepted connections */ 151 struct accept_queue sol_comp; 152 u_int sol_qlen; /* (e) sol_comp length */ 153 u_int sol_incqlen; /* (e) sol_incomp length */ 154 u_int sol_qlimit; /* (e) queue limit */ 155 156 /* accept_filter(9) optional data */ 157 struct accept_filter *sol_accept_filter; 158 void *sol_accept_filter_arg; /* saved filter args */ 159 char *sol_accept_filter_str; /* saved user args */ 160 161 /* Optional upcall, for kernel socket. */ 162 so_upcall_t *sol_upcall; /* (e) */ 163 void *sol_upcallarg; /* (e) */ 164 165 /* Socket buffer parameters, to be copied to 166 * dataflow sockets, accepted from this one. */ 167 int sol_sbrcv_lowat; 168 int sol_sbsnd_lowat; 169 u_int sol_sbrcv_hiwat; 170 u_int sol_sbsnd_hiwat; 171 short sol_sbrcv_flags; 172 short sol_sbsnd_flags; 173 sbintime_t sol_sbrcv_timeo; 174 sbintime_t sol_sbsnd_timeo; 175 }; 176 }; 177 }; 178 #endif /* defined(_KERNEL) || defined(_WANT_SOCKET) */ 179 180 /* 181 * Socket state bits. 182 * 183 * Historically, this bits were all kept in the so_state field. For 184 * locking reasons, they are now in multiple fields, as they are 185 * locked differently. so_state maintains basic socket state protected 186 * by the socket lock. so_qstate holds information about the socket 187 * accept queues. Each socket buffer also has a state field holding 188 * information relevant to that socket buffer (can't send, rcv). Many 189 * fields will be read without locks to improve performance and avoid 190 * lock order issues. However, this approach must be used with caution. 191 */ 192 #define SS_NOFDREF 0x0001 /* no file table ref any more */ 193 #define SS_ISCONNECTED 0x0002 /* socket connected to a peer */ 194 #define SS_ISCONNECTING 0x0004 /* in process of connecting to peer */ 195 #define SS_ISDISCONNECTING 0x0008 /* in process of disconnecting */ 196 #define SS_NBIO 0x0100 /* non-blocking ops */ 197 #define SS_ASYNC 0x0200 /* async i/o notify */ 198 #define SS_ISCONFIRMING 0x0400 /* deciding to accept connection req */ 199 #define SS_ISDISCONNECTED 0x2000 /* socket disconnected from peer */ 200 201 /* 202 * Protocols can mark a socket as SS_PROTOREF to indicate that, following 203 * pru_detach, they still want the socket to persist, and will free it 204 * themselves when they are done. Protocols should only ever call sofree() 205 * following setting this flag in pru_detach(), and never otherwise, as 206 * sofree() bypasses socket reference counting. 207 */ 208 #define SS_PROTOREF 0x4000 /* strong protocol reference */ 209 210 #ifdef _KERNEL 211 212 #define SOCK_MTX(so) &(so)->so_lock 213 #define SOCK_LOCK(so) mtx_lock(&(so)->so_lock) 214 #define SOCK_OWNED(so) mtx_owned(&(so)->so_lock) 215 #define SOCK_UNLOCK(so) mtx_unlock(&(so)->so_lock) 216 #define SOCK_LOCK_ASSERT(so) mtx_assert(&(so)->so_lock, MA_OWNED) 217 #define SOCK_UNLOCK_ASSERT(so) mtx_assert(&(so)->so_lock, MA_NOTOWNED) 218 219 #define SOLISTENING(sol) (((sol)->so_options & SO_ACCEPTCONN) != 0) 220 #define SOLISTEN_LOCK(sol) do { \ 221 mtx_lock(&(sol)->so_lock); \ 222 KASSERT(SOLISTENING(sol), \ 223 ("%s: %p not listening", __func__, (sol))); \ 224 } while (0) 225 #define SOLISTEN_TRYLOCK(sol) mtx_trylock(&(sol)->so_lock) 226 #define SOLISTEN_UNLOCK(sol) do { \ 227 KASSERT(SOLISTENING(sol), \ 228 ("%s: %p not listening", __func__, (sol))); \ 229 mtx_unlock(&(sol)->so_lock); \ 230 } while (0) 231 #define SOLISTEN_LOCK_ASSERT(sol) do { \ 232 mtx_assert(&(sol)->so_lock, MA_OWNED); \ 233 KASSERT(SOLISTENING(sol), \ 234 ("%s: %p not listening", __func__, (sol))); \ 235 } while (0) 236 237 /* 238 * Macros for sockets and socket buffering. 239 */ 240 241 /* 242 * Flags to sblock(). 243 */ 244 #define SBL_WAIT 0x00000001 /* Wait if not immediately available. */ 245 #define SBL_NOINTR 0x00000002 /* Force non-interruptible sleep. */ 246 #define SBL_VALID (SBL_WAIT | SBL_NOINTR) 247 248 /* 249 * Do we need to notify the other side when I/O is possible? 250 */ 251 #define sb_notify(sb) (((sb)->sb_flags & (SB_WAIT | SB_SEL | SB_ASYNC | \ 252 SB_UPCALL | SB_AIO | SB_KNOTE)) != 0) 253 254 /* do we have to send all at once on a socket? */ 255 #define sosendallatonce(so) \ 256 ((so)->so_proto->pr_flags & PR_ATOMIC) 257 258 /* can we read something from so? */ 259 #define soreadabledata(so) \ 260 (sbavail(&(so)->so_rcv) >= (so)->so_rcv.sb_lowat || (so)->so_error) 261 #define soreadable(so) \ 262 (soreadabledata(so) || ((so)->so_rcv.sb_state & SBS_CANTRCVMORE)) 263 264 /* can we write something to so? */ 265 #define sowriteable(so) \ 266 ((sbspace(&(so)->so_snd) >= (so)->so_snd.sb_lowat && \ 267 (((so)->so_state&SS_ISCONNECTED) || \ 268 ((so)->so_proto->pr_flags&PR_CONNREQUIRED)==0)) || \ 269 ((so)->so_snd.sb_state & SBS_CANTSENDMORE) || \ 270 (so)->so_error) 271 272 /* 273 * soref()/sorele() ref-count the socket structure. 274 * soref() may be called without owning socket lock, but in that case a 275 * caller must own something that holds socket, and so_count must be not 0. 276 * Note that you must still explicitly close the socket, but the last ref 277 * count will free the structure. 278 */ 279 #define soref(so) refcount_acquire(&(so)->so_count) 280 #define sorele(so) do { \ 281 SOCK_LOCK_ASSERT(so); \ 282 if (refcount_release(&(so)->so_count)) \ 283 sofree(so); \ 284 else \ 285 SOCK_UNLOCK(so); \ 286 } while (0) 287 288 /* 289 * In sorwakeup() and sowwakeup(), acquire the socket buffer lock to 290 * avoid a non-atomic test-and-wakeup. However, sowakeup is 291 * responsible for releasing the lock if it is called. We unlock only 292 * if we don't call into sowakeup. If any code is introduced that 293 * directly invokes the underlying sowakeup() primitives, it must 294 * maintain the same semantics. 295 */ 296 #define sorwakeup_locked(so) do { \ 297 SOCKBUF_LOCK_ASSERT(&(so)->so_rcv); \ 298 if (sb_notify(&(so)->so_rcv)) \ 299 sowakeup((so), &(so)->so_rcv); \ 300 else \ 301 SOCKBUF_UNLOCK(&(so)->so_rcv); \ 302 } while (0) 303 304 #define sorwakeup(so) do { \ 305 SOCKBUF_LOCK(&(so)->so_rcv); \ 306 sorwakeup_locked(so); \ 307 } while (0) 308 309 #define sowwakeup_locked(so) do { \ 310 SOCKBUF_LOCK_ASSERT(&(so)->so_snd); \ 311 if (sb_notify(&(so)->so_snd)) \ 312 sowakeup((so), &(so)->so_snd); \ 313 else \ 314 SOCKBUF_UNLOCK(&(so)->so_snd); \ 315 } while (0) 316 317 #define sowwakeup(so) do { \ 318 SOCKBUF_LOCK(&(so)->so_snd); \ 319 sowwakeup_locked(so); \ 320 } while (0) 321 322 struct accept_filter { 323 char accf_name[16]; 324 int (*accf_callback) 325 (struct socket *so, void *arg, int waitflag); 326 void * (*accf_create) 327 (struct socket *so, char *arg); 328 void (*accf_destroy) 329 (struct socket *so); 330 SLIST_ENTRY(accept_filter) accf_next; 331 }; 332 333 #ifdef MALLOC_DECLARE 334 MALLOC_DECLARE(M_ACCF); 335 MALLOC_DECLARE(M_PCB); 336 MALLOC_DECLARE(M_SONAME); 337 #endif 338 339 /* 340 * Socket specific helper hook point identifiers 341 * Do not leave holes in the sequence, hook registration is a loop. 342 */ 343 #define HHOOK_SOCKET_OPT 0 344 #define HHOOK_SOCKET_CREATE 1 345 #define HHOOK_SOCKET_RCV 2 346 #define HHOOK_SOCKET_SND 3 347 #define HHOOK_FILT_SOREAD 4 348 #define HHOOK_FILT_SOWRITE 5 349 #define HHOOK_SOCKET_CLOSE 6 350 #define HHOOK_SOCKET_LAST HHOOK_SOCKET_CLOSE 351 352 struct socket_hhook_data { 353 struct socket *so; 354 struct mbuf *m; 355 void *hctx; /* hook point specific data*/ 356 int status; 357 }; 358 359 extern int maxsockets; 360 extern u_long sb_max; 361 extern so_gen_t so_gencnt; 362 363 struct file; 364 struct filecaps; 365 struct filedesc; 366 struct mbuf; 367 struct sockaddr; 368 struct ucred; 369 struct uio; 370 371 /* 'which' values for socket upcalls. */ 372 #define SO_RCV 1 373 #define SO_SND 2 374 375 /* Return values for socket upcalls. */ 376 #define SU_OK 0 377 #define SU_ISCONNECTED 1 378 379 /* 380 * From uipc_socket and friends 381 */ 382 int getsockaddr(struct sockaddr **namp, caddr_t uaddr, size_t len); 383 int getsock_cap(struct thread *td, int fd, cap_rights_t *rightsp, 384 struct file **fpp, u_int *fflagp, struct filecaps *havecaps); 385 void soabort(struct socket *so); 386 int soaccept(struct socket *so, struct sockaddr **nam); 387 void soaio_enqueue(struct task *task); 388 void soaio_rcv(void *context, int pending); 389 void soaio_snd(void *context, int pending); 390 int socheckuid(struct socket *so, uid_t uid); 391 int sobind(struct socket *so, struct sockaddr *nam, struct thread *td); 392 int sobindat(int fd, struct socket *so, struct sockaddr *nam, 393 struct thread *td); 394 int soclose(struct socket *so); 395 int soconnect(struct socket *so, struct sockaddr *nam, struct thread *td); 396 int soconnectat(int fd, struct socket *so, struct sockaddr *nam, 397 struct thread *td); 398 int soconnect2(struct socket *so1, struct socket *so2); 399 int socreate(int dom, struct socket **aso, int type, int proto, 400 struct ucred *cred, struct thread *td); 401 int sodisconnect(struct socket *so); 402 void sodtor_set(struct socket *, so_dtor_t *); 403 struct sockaddr *sodupsockaddr(const struct sockaddr *sa, int mflags); 404 void sofree(struct socket *so); 405 void sohasoutofband(struct socket *so); 406 int solisten(struct socket *so, int backlog, struct thread *td); 407 void solisten_proto(struct socket *so, int backlog); 408 int solisten_proto_check(struct socket *so); 409 int solisten_dequeue(struct socket *, struct socket **, int); 410 struct socket * 411 sonewconn(struct socket *head, int connstatus); 412 struct socket * 413 sopeeloff(struct socket *); 414 int sopoll(struct socket *so, int events, struct ucred *active_cred, 415 struct thread *td); 416 int sopoll_generic(struct socket *so, int events, 417 struct ucred *active_cred, struct thread *td); 418 int soreceive(struct socket *so, struct sockaddr **paddr, struct uio *uio, 419 struct mbuf **mp0, struct mbuf **controlp, int *flagsp); 420 int soreceive_stream(struct socket *so, struct sockaddr **paddr, 421 struct uio *uio, struct mbuf **mp0, struct mbuf **controlp, 422 int *flagsp); 423 int soreceive_dgram(struct socket *so, struct sockaddr **paddr, 424 struct uio *uio, struct mbuf **mp0, struct mbuf **controlp, 425 int *flagsp); 426 int soreceive_generic(struct socket *so, struct sockaddr **paddr, 427 struct uio *uio, struct mbuf **mp0, struct mbuf **controlp, 428 int *flagsp); 429 int soreserve(struct socket *so, u_long sndcc, u_long rcvcc); 430 void sorflush(struct socket *so); 431 int sosend(struct socket *so, struct sockaddr *addr, struct uio *uio, 432 struct mbuf *top, struct mbuf *control, int flags, 433 struct thread *td); 434 int sosend_dgram(struct socket *so, struct sockaddr *addr, 435 struct uio *uio, struct mbuf *top, struct mbuf *control, 436 int flags, struct thread *td); 437 int sosend_generic(struct socket *so, struct sockaddr *addr, 438 struct uio *uio, struct mbuf *top, struct mbuf *control, 439 int flags, struct thread *td); 440 int soshutdown(struct socket *so, int how); 441 void soupcall_clear(struct socket *, int); 442 void soupcall_set(struct socket *, int, so_upcall_t, void *); 443 void solisten_upcall_set(struct socket *, so_upcall_t, void *); 444 void sowakeup(struct socket *so, struct sockbuf *sb); 445 void sowakeup_aio(struct socket *so, struct sockbuf *sb); 446 void solisten_wakeup(struct socket *); 447 int selsocket(struct socket *so, int events, struct timeval *tv, 448 struct thread *td); 449 void soisconnected(struct socket *so); 450 void soisconnecting(struct socket *so); 451 void soisdisconnected(struct socket *so); 452 void soisdisconnecting(struct socket *so); 453 void socantrcvmore(struct socket *so); 454 void socantrcvmore_locked(struct socket *so); 455 void socantsendmore(struct socket *so); 456 void socantsendmore_locked(struct socket *so); 457 458 /* 459 * Accept filter functions (duh). 460 */ 461 int accept_filt_add(struct accept_filter *filt); 462 int accept_filt_del(char *name); 463 struct accept_filter *accept_filt_get(char *name); 464 #ifdef ACCEPT_FILTER_MOD 465 #ifdef SYSCTL_DECL 466 SYSCTL_DECL(_net_inet_accf); 467 #endif 468 int accept_filt_generic_mod_event(module_t mod, int event, void *data); 469 #endif 470 471 #endif /* _KERNEL */ 472 473 /* 474 * Structure to export socket from kernel to utilities, via sysctl(3). 475 */ 476 struct xsocket { 477 ksize_t xso_len; /* length of this structure */ 478 kvaddr_t xso_so; /* kernel address of struct socket */ 479 kvaddr_t so_pcb; /* kernel address of struct inpcb */ 480 uint64_t so_oobmark; 481 int64_t so_spare64[8]; 482 int32_t xso_protocol; 483 int32_t xso_family; 484 uint32_t so_qlen; 485 uint32_t so_incqlen; 486 uint32_t so_qlimit; 487 pid_t so_pgid; 488 uid_t so_uid; 489 int32_t so_spare32[8]; 490 int16_t so_type; 491 int16_t so_options; 492 int16_t so_linger; 493 int16_t so_state; 494 int16_t so_timeo; 495 uint16_t so_error; 496 struct xsockbuf { 497 uint32_t sb_cc; 498 uint32_t sb_hiwat; 499 uint32_t sb_mbcnt; 500 uint32_t sb_mcnt; 501 uint32_t sb_ccnt; 502 uint32_t sb_mbmax; 503 int32_t sb_lowat; 504 int32_t sb_timeo; 505 int16_t sb_flags; 506 } so_rcv, so_snd; 507 }; 508 509 #ifdef _KERNEL 510 void sotoxsocket(struct socket *so, struct xsocket *xso); 511 void sbtoxsockbuf(struct sockbuf *sb, struct xsockbuf *xsb); 512 #endif 513 514 /* 515 * Socket buffer state bits. Exported via libprocstat(3). 516 */ 517 #define SBS_CANTSENDMORE 0x0010 /* can't send more data to peer */ 518 #define SBS_CANTRCVMORE 0x0020 /* can't receive more data from peer */ 519 #define SBS_RCVATMARK 0x0040 /* at mark on input */ 520 521 #endif /* !_SYS_SOCKETVAR_H_ */ 522