2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 1982, 1986, 1990, 1993
5 * The Regents of the University of California. All rights reserved.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
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.
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
32 #ifndef _SYS_SOCKETVAR_H_
33 #define _SYS_SOCKETVAR_H_
36 * Socket generation count type. Also used in xinpcb, xtcpcb, xunpcb.
38 typedef uint64_t so_gen_t;
40 #if defined(_KERNEL) || defined(_WANT_SOCKET)
41 #include <sys/queue.h> /* for TAILQ macros */
42 #include <sys/selinfo.h> /* for struct selinfo */
43 #include <sys/_lock.h>
44 #include <sys/_mutex.h>
47 #include <sys/sockbuf.h>
49 #include <sys/caprights.h>
50 #include <sys/sockopt.h>
56 * Kernel structure per socket.
57 * Contains send and receive buffer queues,
58 * handle on protocol and pointer to protocol
59 * private data and error information.
61 typedef int so_upcall_t(struct socket *, void *, int);
62 typedef void so_dtor_t(struct socket *);
68 SQ_INCOMP = 0x0800, /* on sol_incomp */
69 SQ_COMP = 0x1000, /* on sol_comp */
73 * Locking key to struct socket:
74 * (a) constant after allocation, no locking required.
75 * (b) locked by SOCK_LOCK(so).
76 * (cr) locked by SOCK_RECVBUF_LOCK(so)
77 * (cs) locked by SOCK_SENDBUF_LOCK(so)
78 * (e) locked by SOLISTEN_LOCK() of corresponding listening socket.
79 * (f) not locked since integer reads/writes are atomic.
80 * (g) used only as a sleep/wakeup address, no value.
81 * (h) locked by global mutex so_global_mtx.
82 * (k) locked by KTLS workqueue mutex
84 TAILQ_HEAD(accept_queue, socket);
87 volatile u_int so_count; /* (b / refcount) */
88 struct selinfo so_rdsel; /* (b/cr) for so_rcv/so_comp */
89 struct selinfo so_wrsel; /* (b/cs) for so_snd */
90 int so_options; /* (b) from socket call, see socket.h */
91 short so_type; /* (a) generic type, see socket.h */
92 short so_state; /* (b) internal state flags SS_* */
93 void *so_pcb; /* protocol control block */
94 struct vnet *so_vnet; /* (a) network stack instance */
95 struct protosw *so_proto; /* (a) protocol handle */
96 short so_linger; /* time to linger close(2) */
97 short so_timeo; /* (g) connection timeout */
98 u_short so_error; /* (f) error affecting connection */
99 u_short so_rerror; /* (f) error affecting connection */
100 struct sigio *so_sigio; /* [sg] information for async I/O or
101 out of band data (SIGURG) */
102 struct ucred *so_cred; /* (a) user credentials */
103 struct label *so_label; /* (b) MAC label for socket */
104 /* NB: generation count must not be first. */
105 so_gen_t so_gencnt; /* (h) generation count */
106 void *so_emuldata; /* (b) private data for emulators */
107 so_dtor_t *so_dtor; /* (b) optional destructor */
108 struct osd osd; /* Object Specific extensions */
110 * so_fibnum, so_user_cookie and friends can be used to attach
111 * some user-specified metadata to a socket, which then can be
112 * used by the kernel for various actions.
113 * so_user_cookie is used by ipfw/dummynet.
115 int so_fibnum; /* routing domain for this socket */
116 uint32_t so_user_cookie;
118 int so_ts_clock; /* type of the clock used for timestamps */
119 uint32_t so_max_pacing_rate; /* (f) TX rate limit in bytes/s */
122 * Mutexes to prevent interleaving of socket I/O. These have to be
123 * outside of the socket buffers in order to interlock with listen(2).
125 struct sx so_snd_sx __aligned(CACHE_LINE_SIZE);
126 struct mtx so_snd_mtx;
128 struct sx so_rcv_sx __aligned(CACHE_LINE_SIZE);
129 struct mtx so_rcv_mtx;
132 /* Regular (data flow) socket. */
134 /* (cr, cs) Receive and send buffers. */
135 struct sockbuf so_rcv, so_snd;
137 /* (e) Our place on accept queue. */
138 TAILQ_ENTRY(socket) so_list;
139 struct socket *so_listen; /* (b) */
140 enum socket_qstate so_qstate; /* (b) */
141 /* (b) cached MAC label for peer */
142 struct label *so_peerlabel;
143 u_long so_oobmark; /* chars to oob mark */
145 /* (k) Our place on KTLS RX work queue. */
146 STAILQ_ENTRY(socket) so_ktls_rx_list;
149 * Listening socket, where accepts occur, is so_listen in all
150 * subsidiary sockets. If so_listen is NULL, socket is not
151 * related to an accept. For a listening socket itself
152 * sol_incomp queues partially completed connections, while
153 * sol_comp is a queue of connections ready to be accepted.
154 * If a connection is aborted and it has so_listen set, then
155 * it has to be pulled out of either sol_incomp or sol_comp.
156 * We allow connections to queue up based on current queue
157 * lengths and limit on number of queued connections for this
161 /* (e) queue of partial unaccepted connections */
162 struct accept_queue sol_incomp;
163 /* (e) queue of complete unaccepted connections */
164 struct accept_queue sol_comp;
165 u_int sol_qlen; /* (e) sol_comp length */
166 u_int sol_incqlen; /* (e) sol_incomp length */
167 u_int sol_qlimit; /* (e) queue limit */
169 /* accept_filter(9) optional data */
170 struct accept_filter *sol_accept_filter;
171 void *sol_accept_filter_arg; /* saved filter args */
172 char *sol_accept_filter_str; /* saved user args */
174 /* Optional upcall, for kernel socket. */
175 so_upcall_t *sol_upcall; /* (e) */
176 void *sol_upcallarg; /* (e) */
178 /* Socket buffer parameters, to be copied to
179 * dataflow sockets, accepted from this one. */
182 u_int sol_sbrcv_hiwat;
183 u_int sol_sbsnd_hiwat;
184 short sol_sbrcv_flags;
185 short sol_sbsnd_flags;
186 sbintime_t sol_sbrcv_timeo;
187 sbintime_t sol_sbsnd_timeo;
189 /* Information tracking listen queue overflows. */
190 struct timeval sol_lastover; /* (e) */
191 int sol_overcount; /* (e) */
195 #endif /* defined(_KERNEL) || defined(_WANT_SOCKET) */
200 * Historically, these bits were all kept in the so_state field.
201 * They are now split into separate, lock-specific fields.
202 * so_state maintains basic socket state protected by the socket lock.
203 * so_qstate holds information about the socket accept queues.
204 * Each socket buffer also has a state field holding information
205 * relevant to that socket buffer (can't send, rcv).
206 * Many fields will be read without locks to improve performance and avoid
207 * lock order issues. However, this approach must be used with caution.
209 #define SS_ISCONNECTED 0x0002 /* socket connected to a peer */
210 #define SS_ISCONNECTING 0x0004 /* in process of connecting to peer */
211 #define SS_ISDISCONNECTING 0x0008 /* in process of disconnecting */
212 #define SS_NBIO 0x0100 /* non-blocking ops */
213 #define SS_ASYNC 0x0200 /* async i/o notify */
214 /* was SS_ISCONFIRMING 0x0400 */
215 #define SS_ISDISCONNECTED 0x2000 /* socket disconnected from peer */
219 #define SOCK_MTX(so) (&(so)->so_lock)
220 #define SOCK_LOCK(so) mtx_lock(&(so)->so_lock)
221 #define SOCK_OWNED(so) mtx_owned(&(so)->so_lock)
222 #define SOCK_UNLOCK(so) mtx_unlock(&(so)->so_lock)
223 #define SOCK_LOCK_ASSERT(so) mtx_assert(&(so)->so_lock, MA_OWNED)
224 #define SOCK_UNLOCK_ASSERT(so) mtx_assert(&(so)->so_lock, MA_NOTOWNED)
226 #define SOLISTENING(sol) (((sol)->so_options & SO_ACCEPTCONN) != 0)
227 #define SOLISTEN_LOCK(sol) do { \
228 mtx_lock(&(sol)->so_lock); \
229 KASSERT(SOLISTENING(sol), \
230 ("%s: %p not listening", __func__, (sol))); \
232 #define SOLISTEN_TRYLOCK(sol) mtx_trylock(&(sol)->so_lock)
233 #define SOLISTEN_UNLOCK(sol) do { \
234 KASSERT(SOLISTENING(sol), \
235 ("%s: %p not listening", __func__, (sol))); \
236 mtx_unlock(&(sol)->so_lock); \
238 #define SOLISTEN_LOCK_ASSERT(sol) do { \
239 mtx_assert(&(sol)->so_lock, MA_OWNED); \
240 KASSERT(SOLISTENING(sol), \
241 ("%s: %p not listening", __func__, (sol))); \
243 #define SOLISTEN_UNLOCK_ASSERT(sol) do { \
244 mtx_assert(&(sol)->so_lock, MA_NOTOWNED); \
245 KASSERT(SOLISTENING(sol), \
246 ("%s: %p not listening", __func__, (sol))); \
250 * Socket buffer locks. These are strongly preferred over SOCKBUF_LOCK(sb)
251 * macros, as we are moving towards protocol specific socket buffers.
253 #define SOCK_RECVBUF_MTX(so) \
255 #define SOCK_RECVBUF_LOCK(so) \
256 mtx_lock(SOCK_RECVBUF_MTX(so))
257 #define SOCK_RECVBUF_UNLOCK(so) \
258 mtx_unlock(SOCK_RECVBUF_MTX(so))
259 #define SOCK_RECVBUF_LOCK_ASSERT(so) \
260 mtx_assert(SOCK_RECVBUF_MTX(so), MA_OWNED)
261 #define SOCK_RECVBUF_UNLOCK_ASSERT(so) \
262 mtx_assert(SOCK_RECVBUF_MTX(so), MA_NOTOWNED)
264 #define SOCK_SENDBUF_MTX(so) \
266 #define SOCK_SENDBUF_LOCK(so) \
267 mtx_lock(SOCK_SENDBUF_MTX(so))
268 #define SOCK_SENDBUF_UNLOCK(so) \
269 mtx_unlock(SOCK_SENDBUF_MTX(so))
270 #define SOCK_SENDBUF_LOCK_ASSERT(so) \
271 mtx_assert(SOCK_SENDBUF_MTX(so), MA_OWNED)
272 #define SOCK_SENDBUF_UNLOCK_ASSERT(so) \
273 mtx_assert(SOCK_SENDBUF_MTX(so), MA_NOTOWNED)
275 #define SOCK_BUF_LOCK(so, which) \
276 mtx_lock(soeventmtx(so, which))
277 #define SOCK_BUF_UNLOCK(so, which) \
278 mtx_unlock(soeventmtx(so, which))
279 #define SOCK_BUF_LOCK_ASSERT(so, which) \
280 mtx_assert(soeventmtx(so, which), MA_OWNED)
281 #define SOCK_BUF_UNLOCK_ASSERT(so, which) \
282 mtx_assert(soeventmtx(so, which), MA_NOTOWNED)
284 static inline struct sockbuf *
285 sobuf(struct socket *so, const sb_which which)
287 return (which == SO_RCV ? &so->so_rcv : &so->so_snd);
290 static inline struct mtx *
291 soeventmtx(struct socket *so, const sb_which which)
293 return (which == SO_RCV ? SOCK_RECVBUF_MTX(so) : SOCK_SENDBUF_MTX(so));
297 * Macros for sockets and socket buffering.
301 * Flags to soiolock().
303 #define SBL_WAIT 0x00000001 /* Wait if not immediately available. */
304 #define SBL_NOINTR 0x00000002 /* Force non-interruptible sleep. */
305 #define SBL_VALID (SBL_WAIT | SBL_NOINTR)
307 #define SBLOCKWAIT(f) (((f) & MSG_DONTWAIT) ? 0 : SBL_WAIT)
309 #define SOCK_IO_SEND_LOCK(so, flags) \
310 soiolock((so), &(so)->so_snd_sx, (flags))
311 #define SOCK_IO_SEND_UNLOCK(so) \
312 soiounlock(&(so)->so_snd_sx)
313 #define SOCK_IO_SEND_OWNED(so) sx_xlocked(&(so)->so_snd_sx)
314 #define SOCK_IO_RECV_LOCK(so, flags) \
315 soiolock((so), &(so)->so_rcv_sx, (flags))
316 #define SOCK_IO_RECV_UNLOCK(so) \
317 soiounlock(&(so)->so_rcv_sx)
318 #define SOCK_IO_RECV_OWNED(so) sx_xlocked(&(so)->so_rcv_sx)
320 /* do we have to send all at once on a socket? */
321 #define sosendallatonce(so) \
322 ((so)->so_proto->pr_flags & PR_ATOMIC)
324 /* can we read something from so? */
325 #define soreadabledata(so) \
326 (sbavail(&(so)->so_rcv) >= (so)->so_rcv.sb_lowat || \
327 (so)->so_error || (so)->so_rerror)
328 #define soreadable(so) \
329 (soreadabledata(so) || ((so)->so_rcv.sb_state & SBS_CANTRCVMORE))
331 /* can we write something to so? */
332 #define sowriteable(so) \
333 ((sbspace(&(so)->so_snd) >= (so)->so_snd.sb_lowat && \
334 (((so)->so_state&SS_ISCONNECTED) || \
335 ((so)->so_proto->pr_flags&PR_CONNREQUIRED)==0)) || \
336 ((so)->so_snd.sb_state & SBS_CANTSENDMORE) || \
340 * soref()/sorele() ref-count the socket structure.
341 * soref() may be called without owning socket lock, but in that case a
342 * caller must own something that holds socket, and so_count must be not 0.
343 * Note that you must still explicitly close the socket, but the last ref
344 * count will free the structure.
346 #define soref(so) refcount_acquire(&(so)->so_count)
347 #define sorele(so) do { \
348 SOCK_UNLOCK_ASSERT(so); \
349 if (!refcount_release_if_not_last(&(so)->so_count)) { \
356 * In sorwakeup() and sowwakeup(), acquire the socket buffer lock to
357 * avoid a non-atomic test-and-wakeup. However, sowakeup is
358 * responsible for releasing the lock if it is called. We unlock only
359 * if we don't call into sowakeup. If any code is introduced that
360 * directly invokes the underlying sowakeup() primitives, it must
361 * maintain the same semantics.
363 #define sorwakeup(so) do { \
364 SOCK_RECVBUF_LOCK(so); \
365 sorwakeup_locked(so); \
368 #define sowwakeup(so) do { \
369 SOCK_SENDBUF_LOCK(so); \
370 sowwakeup_locked(so); \
373 struct accept_filter {
376 (struct socket *so, void *arg, int waitflag);
377 void * (*accf_create)
378 (struct socket *so, char *arg);
381 SLIST_ENTRY(accept_filter) accf_next;
384 #define ACCEPT_FILTER_DEFINE(modname, filtname, cb, create, destroy, ver) \
385 static struct accept_filter modname##_filter = { \
386 .accf_name = filtname, \
387 .accf_callback = cb, \
388 .accf_create = create, \
389 .accf_destroy = destroy, \
391 static moduledata_t modname##_mod = { \
392 .name = __XSTRING(modname), \
393 .evhand = accept_filt_generic_mod_event, \
394 .priv = &modname##_filter, \
396 DECLARE_MODULE(modname, modname##_mod, SI_SUB_DRIVERS, \
398 MODULE_VERSION(modname, ver)
400 #ifdef MALLOC_DECLARE
401 MALLOC_DECLARE(M_ACCF);
402 MALLOC_DECLARE(M_PCB);
403 MALLOC_DECLARE(M_SONAME);
407 * Socket specific helper hook point identifiers
408 * Do not leave holes in the sequence, hook registration is a loop.
410 #define HHOOK_SOCKET_OPT 0
411 #define HHOOK_SOCKET_CREATE 1
412 #define HHOOK_SOCKET_RCV 2
413 #define HHOOK_SOCKET_SND 3
414 #define HHOOK_FILT_SOREAD 4
415 #define HHOOK_FILT_SOWRITE 5
416 #define HHOOK_SOCKET_CLOSE 6
417 #define HHOOK_SOCKET_LAST HHOOK_SOCKET_CLOSE
419 struct socket_hhook_data {
422 void *hctx; /* hook point specific data*/
426 extern int maxsockets;
427 extern u_long sb_max;
428 extern so_gen_t so_gencnt;
439 /* Return values for socket upcalls. */
441 #define SU_ISCONNECTED 1
444 * From uipc_socket and friends
446 int getsockaddr(struct sockaddr **namp, const struct sockaddr *uaddr,
448 int getsock_cap(struct thread *td, int fd, cap_rights_t *rightsp,
449 struct file **fpp, struct filecaps *havecaps);
450 int getsock(struct thread *td, int fd, cap_rights_t *rightsp,
452 void soabort(struct socket *so);
453 int soaccept(struct socket *so, struct sockaddr *sa);
454 int sopeeraddr(struct socket *so, struct sockaddr *sa);
455 int sosockaddr(struct socket *so, struct sockaddr *sa);
456 void soaio_enqueue(struct task *task);
457 void soaio_rcv(void *context, int pending);
458 void soaio_snd(void *context, int pending);
459 int socheckuid(struct socket *so, uid_t uid);
460 int sobind(struct socket *so, struct sockaddr *nam, struct thread *td);
461 int sobindat(int fd, struct socket *so, struct sockaddr *nam,
463 int soclose(struct socket *so);
464 int soconnect(struct socket *so, struct sockaddr *nam, struct thread *td);
465 int soconnectat(int fd, struct socket *so, struct sockaddr *nam,
467 int soconnect2(struct socket *so1, struct socket *so2);
468 int socreate(int dom, struct socket **aso, int type, int proto,
469 struct ucred *cred, struct thread *td);
470 int sodisconnect(struct socket *so);
471 void sodtor_set(struct socket *, so_dtor_t *);
472 struct sockaddr *sodupsockaddr(const struct sockaddr *sa, int mflags);
473 void sohasoutofband(struct socket *so);
474 int solisten(struct socket *so, int backlog, struct thread *td);
475 void solisten_proto(struct socket *so, int backlog);
476 void solisten_proto_abort(struct socket *so);
477 int solisten_proto_check(struct socket *so);
478 bool solisten_enqueue(struct socket *, int);
479 int solisten_dequeue(struct socket *, struct socket **, int);
481 solisten_clone(struct socket *);
483 sonewconn(struct socket *head, int connstatus);
485 sopeeloff(struct socket *);
486 int sopoll(struct socket *so, int events, struct ucred *active_cred,
488 int sopoll_generic(struct socket *so, int events,
489 struct ucred *active_cred, struct thread *td);
490 int soreceive(struct socket *so, struct sockaddr **paddr, struct uio *uio,
491 struct mbuf **mp0, struct mbuf **controlp, int *flagsp);
492 int soreceive_stream(struct socket *so, struct sockaddr **paddr,
493 struct uio *uio, struct mbuf **mp0, struct mbuf **controlp,
495 int soreceive_dgram(struct socket *so, struct sockaddr **paddr,
496 struct uio *uio, struct mbuf **mp0, struct mbuf **controlp,
498 int soreceive_generic(struct socket *so, struct sockaddr **paddr,
499 struct uio *uio, struct mbuf **mp0, struct mbuf **controlp,
501 void sorele_locked(struct socket *so);
502 void sodealloc(struct socket *);
503 int soreserve(struct socket *so, u_long sndcc, u_long rcvcc);
504 void sorflush(struct socket *so);
505 int sosend(struct socket *so, struct sockaddr *addr, struct uio *uio,
506 struct mbuf *top, struct mbuf *control, int flags,
508 int sousrsend(struct socket *so, struct sockaddr *addr, struct uio *uio,
509 struct mbuf *control, int flags, struct proc *);
510 int sosend_dgram(struct socket *so, struct sockaddr *addr,
511 struct uio *uio, struct mbuf *top, struct mbuf *control,
512 int flags, struct thread *td);
513 int sosend_generic(struct socket *so, struct sockaddr *addr,
514 struct uio *uio, struct mbuf *top, struct mbuf *control,
515 int flags, struct thread *td);
516 int soshutdown(struct socket *so, enum shutdown_how);
517 void soupcall_clear(struct socket *, sb_which);
518 void soupcall_set(struct socket *, sb_which, so_upcall_t, void *);
519 void solisten_upcall_set(struct socket *, so_upcall_t, void *);
520 void sorwakeup_locked(struct socket *);
521 void sowwakeup_locked(struct socket *);
522 void sowakeup_aio(struct socket *, sb_which);
523 void solisten_wakeup(struct socket *);
524 int selsocket(struct socket *so, int events, struct timeval *tv,
526 void soisconnected(struct socket *so);
527 void soisconnecting(struct socket *so);
528 void soisdisconnected(struct socket *so);
529 void soisdisconnecting(struct socket *so);
530 void socantrcvmore(struct socket *so);
531 void socantrcvmore_locked(struct socket *so);
532 void socantsendmore(struct socket *so);
533 void socantsendmore_locked(struct socket *so);
534 void soroverflow(struct socket *so);
535 void soroverflow_locked(struct socket *so);
536 int soiolock(struct socket *so, struct sx *sx, int flags);
537 void soiounlock(struct sx *sx);
540 * Accept filter functions (duh).
542 int accept_filt_add(struct accept_filter *filt);
543 int accept_filt_del(char *name);
544 struct accept_filter *accept_filt_get(char *name);
545 #ifdef ACCEPT_FILTER_MOD
547 SYSCTL_DECL(_net_inet_accf);
549 int accept_filt_generic_mod_event(module_t mod, int event, void *data);
555 * Structure to export socket from kernel to utilities, via sysctl(3).
558 ksize_t xso_len; /* length of this structure */
559 kvaddr_t xso_so; /* kernel address of struct socket */
560 kvaddr_t so_pcb; /* kernel address of struct inpcb */
562 int64_t so_spare64[8];
563 int32_t xso_protocol;
570 int32_t so_spare32[8];
581 uint32_t sb_spare0; /* was sb_mcnt */
582 uint32_t sb_spare1; /* was sb_ccnt */
591 void sotoxsocket(struct socket *so, struct xsocket *xso);
592 void sbtoxsockbuf(struct sockbuf *sb, struct xsockbuf *xsb);
596 * Socket buffer state bits. Exported via libprocstat(3).
598 #define SBS_CANTSENDMORE 0x0010 /* can't send more data to peer */
599 #define SBS_CANTRCVMORE 0x0020 /* can't receive more data from peer */
600 #define SBS_RCVATMARK 0x0040 /* at mark on input */
602 #endif /* !_SYS_SOCKETVAR_H_ */