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
31 * @(#)socketvar.h 8.3 (Berkeley) 2/19/95
36 #ifndef _SYS_SOCKETVAR_H_
37 #define _SYS_SOCKETVAR_H_
40 * Socket generation count type. Also used in xinpcb, xtcpcb, xunpcb.
42 typedef uint64_t so_gen_t;
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>
51 #include <sys/sockbuf.h>
53 #include <sys/caprights.h>
54 #include <sys/sockopt.h>
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.
65 typedef int so_upcall_t(struct socket *, void *, int);
66 typedef void so_dtor_t(struct socket *);
72 SQ_INCOMP = 0x0800, /* on sol_incomp */
73 SQ_COMP = 0x1000, /* on sol_comp */
77 * Locking key to struct socket:
78 * (a) constant after allocation, no locking required.
79 * (b) locked by SOCK_LOCK(so).
80 * (cr) locked by SOCK_RECVBUF_LOCK(so)
81 * (cs) locked by SOCK_SENDBUF_LOCK(so)
82 * (e) locked by SOLISTEN_LOCK() of corresponding listening socket.
83 * (f) not locked since integer reads/writes are atomic.
84 * (g) used only as a sleep/wakeup address, no value.
85 * (h) locked by global mutex so_global_mtx.
86 * (k) locked by KTLS workqueue mutex
88 TAILQ_HEAD(accept_queue, socket);
91 volatile u_int so_count; /* (b / refcount) */
92 struct selinfo so_rdsel; /* (b/cr) for so_rcv/so_comp */
93 struct selinfo so_wrsel; /* (b/cs) for so_snd */
94 int so_options; /* (b) from socket call, see socket.h */
95 short so_type; /* (a) generic type, see socket.h */
96 short so_state; /* (b) internal state flags SS_* */
97 void *so_pcb; /* protocol control block */
98 struct vnet *so_vnet; /* (a) network stack instance */
99 struct protosw *so_proto; /* (a) protocol handle */
100 short so_linger; /* time to linger close(2) */
101 short so_timeo; /* (g) connection timeout */
102 u_short so_error; /* (f) error affecting connection */
103 u_short so_rerror; /* (f) error affecting connection */
104 struct sigio *so_sigio; /* [sg] information for async I/O or
105 out of band data (SIGURG) */
106 struct ucred *so_cred; /* (a) user credentials */
107 struct label *so_label; /* (b) MAC label for socket */
108 /* NB: generation count must not be first. */
109 so_gen_t so_gencnt; /* (h) generation count */
110 void *so_emuldata; /* (b) private data for emulators */
111 so_dtor_t *so_dtor; /* (b) optional destructor */
112 struct osd osd; /* Object Specific extensions */
114 * so_fibnum, so_user_cookie and friends can be used to attach
115 * some user-specified metadata to a socket, which then can be
116 * used by the kernel for various actions.
117 * so_user_cookie is used by ipfw/dummynet.
119 int so_fibnum; /* routing domain for this socket */
120 uint32_t so_user_cookie;
122 int so_ts_clock; /* type of the clock used for timestamps */
123 uint32_t so_max_pacing_rate; /* (f) TX rate limit in bytes/s */
126 * Mutexes to prevent interleaving of socket I/O. These have to be
127 * outside of the socket buffers in order to interlock with listen(2).
129 struct sx so_snd_sx __aligned(CACHE_LINE_SIZE);
130 struct mtx so_snd_mtx;
132 struct sx so_rcv_sx __aligned(CACHE_LINE_SIZE);
133 struct mtx so_rcv_mtx;
136 /* Regular (data flow) socket. */
138 /* (cr, cs) Receive and send buffers. */
139 struct sockbuf so_rcv, so_snd;
141 /* (e) Our place on accept queue. */
142 TAILQ_ENTRY(socket) so_list;
143 struct socket *so_listen; /* (b) */
144 enum socket_qstate so_qstate; /* (b) */
145 /* (b) cached MAC label for peer */
146 struct label *so_peerlabel;
147 u_long so_oobmark; /* chars to oob mark */
149 /* (k) Our place on KTLS RX work queue. */
150 STAILQ_ENTRY(socket) so_ktls_rx_list;
153 * Listening socket, where accepts occur, is so_listen in all
154 * subsidiary sockets. If so_listen is NULL, socket is not
155 * related to an accept. For a listening socket itself
156 * sol_incomp queues partially completed connections, while
157 * sol_comp is a queue of connections ready to be accepted.
158 * If a connection is aborted and it has so_listen set, then
159 * it has to be pulled out of either sol_incomp or sol_comp.
160 * We allow connections to queue up based on current queue
161 * lengths and limit on number of queued connections for this
165 /* (e) queue of partial unaccepted connections */
166 struct accept_queue sol_incomp;
167 /* (e) queue of complete unaccepted connections */
168 struct accept_queue sol_comp;
169 u_int sol_qlen; /* (e) sol_comp length */
170 u_int sol_incqlen; /* (e) sol_incomp length */
171 u_int sol_qlimit; /* (e) queue limit */
173 /* accept_filter(9) optional data */
174 struct accept_filter *sol_accept_filter;
175 void *sol_accept_filter_arg; /* saved filter args */
176 char *sol_accept_filter_str; /* saved user args */
178 /* Optional upcall, for kernel socket. */
179 so_upcall_t *sol_upcall; /* (e) */
180 void *sol_upcallarg; /* (e) */
182 /* Socket buffer parameters, to be copied to
183 * dataflow sockets, accepted from this one. */
186 u_int sol_sbrcv_hiwat;
187 u_int sol_sbsnd_hiwat;
188 short sol_sbrcv_flags;
189 short sol_sbsnd_flags;
190 sbintime_t sol_sbrcv_timeo;
191 sbintime_t sol_sbsnd_timeo;
193 /* Information tracking listen queue overflows. */
194 struct timeval sol_lastover; /* (e) */
195 int sol_overcount; /* (e) */
199 #endif /* defined(_KERNEL) || defined(_WANT_SOCKET) */
204 * Historically, these bits were all kept in the so_state field.
205 * They are now split into separate, lock-specific fields.
206 * so_state maintains basic socket state protected by the socket lock.
207 * so_qstate holds information about the socket accept queues.
208 * Each socket buffer also has a state field holding information
209 * relevant to that socket buffer (can't send, rcv).
210 * Many fields will be read without locks to improve performance and avoid
211 * lock order issues. However, this approach must be used with caution.
213 #define SS_NOFDREF 0x0001 /* no file table ref any more */
214 #define SS_ISCONNECTED 0x0002 /* socket connected to a peer */
215 #define SS_ISCONNECTING 0x0004 /* in process of connecting to peer */
216 #define SS_ISDISCONNECTING 0x0008 /* in process of disconnecting */
217 #define SS_NBIO 0x0100 /* non-blocking ops */
218 #define SS_ASYNC 0x0200 /* async i/o notify */
219 #define SS_ISCONFIRMING 0x0400 /* deciding to accept connection req */
220 #define SS_ISDISCONNECTED 0x2000 /* socket disconnected from peer */
223 * Protocols can mark a socket as SS_PROTOREF to indicate that, following
224 * pru_detach, they still want the socket to persist, and will free it
225 * themselves when they are done. Protocols should only ever call sofree()
226 * following setting this flag in pru_detach(), and never otherwise, as
227 * sofree() bypasses socket reference counting.
229 #define SS_PROTOREF 0x4000 /* strong protocol reference */
233 #define SOCK_MTX(so) (&(so)->so_lock)
234 #define SOCK_LOCK(so) mtx_lock(&(so)->so_lock)
235 #define SOCK_OWNED(so) mtx_owned(&(so)->so_lock)
236 #define SOCK_UNLOCK(so) mtx_unlock(&(so)->so_lock)
237 #define SOCK_LOCK_ASSERT(so) mtx_assert(&(so)->so_lock, MA_OWNED)
238 #define SOCK_UNLOCK_ASSERT(so) mtx_assert(&(so)->so_lock, MA_NOTOWNED)
240 #define SOLISTENING(sol) (((sol)->so_options & SO_ACCEPTCONN) != 0)
241 #define SOLISTEN_LOCK(sol) do { \
242 mtx_lock(&(sol)->so_lock); \
243 KASSERT(SOLISTENING(sol), \
244 ("%s: %p not listening", __func__, (sol))); \
246 #define SOLISTEN_TRYLOCK(sol) mtx_trylock(&(sol)->so_lock)
247 #define SOLISTEN_UNLOCK(sol) do { \
248 KASSERT(SOLISTENING(sol), \
249 ("%s: %p not listening", __func__, (sol))); \
250 mtx_unlock(&(sol)->so_lock); \
252 #define SOLISTEN_LOCK_ASSERT(sol) do { \
253 mtx_assert(&(sol)->so_lock, MA_OWNED); \
254 KASSERT(SOLISTENING(sol), \
255 ("%s: %p not listening", __func__, (sol))); \
259 * Socket buffer locks. These are strongly preferred over SOCKBUF_LOCK(sb)
260 * macros, as we are moving towards protocol specific socket buffers.
262 #define SOCK_RECVBUF_MTX(so) \
264 #define SOCK_RECVBUF_LOCK(so) \
265 mtx_lock(SOCK_RECVBUF_MTX(so))
266 #define SOCK_RECVBUF_UNLOCK(so) \
267 mtx_unlock(SOCK_RECVBUF_MTX(so))
268 #define SOCK_RECVBUF_LOCK_ASSERT(so) \
269 mtx_assert(SOCK_RECVBUF_MTX(so), MA_OWNED)
270 #define SOCK_RECVBUF_UNLOCK_ASSERT(so) \
271 mtx_assert(SOCK_RECVBUF_MTX(so), MA_NOTOWNED)
273 #define SOCK_SENDBUF_MTX(so) \
275 #define SOCK_SENDBUF_LOCK(so) \
276 mtx_lock(SOCK_SENDBUF_MTX(so))
277 #define SOCK_SENDBUF_UNLOCK(so) \
278 mtx_unlock(SOCK_SENDBUF_MTX(so))
279 #define SOCK_SENDBUF_LOCK_ASSERT(so) \
280 mtx_assert(SOCK_SENDBUF_MTX(so), MA_OWNED)
281 #define SOCK_SENDBUF_UNLOCK_ASSERT(so) \
282 mtx_assert(SOCK_SENDBUF_MTX(so), MA_NOTOWNED)
284 #define SOCK_BUF_LOCK(so, which) \
285 mtx_lock(soeventmtx(so, which))
286 #define SOCK_BUF_UNLOCK(so, which) \
287 mtx_unlock(soeventmtx(so, which))
288 #define SOCK_BUF_LOCK_ASSERT(so, which) \
289 mtx_assert(soeventmtx(so, which), MA_OWNED)
290 #define SOCK_BUF_UNLOCK_ASSERT(so, which) \
291 mtx_assert(soeventmtx(so, which), MA_NOTOWNED)
293 static inline struct sockbuf *
294 sobuf(struct socket *so, const sb_which which)
296 return (which == SO_RCV ? &so->so_rcv : &so->so_snd);
299 static inline struct mtx *
300 soeventmtx(struct socket *so, const sb_which which)
302 return (which == SO_RCV ? SOCK_RECVBUF_MTX(so) : SOCK_SENDBUF_MTX(so));
306 * Macros for sockets and socket buffering.
310 * Flags to soiolock().
312 #define SBL_WAIT 0x00000001 /* Wait if not immediately available. */
313 #define SBL_NOINTR 0x00000002 /* Force non-interruptible sleep. */
314 #define SBL_VALID (SBL_WAIT | SBL_NOINTR)
316 #define SBLOCKWAIT(f) (((f) & MSG_DONTWAIT) ? 0 : SBL_WAIT)
318 #define SOCK_IO_SEND_LOCK(so, flags) \
319 soiolock((so), &(so)->so_snd_sx, (flags))
320 #define SOCK_IO_SEND_UNLOCK(so) \
321 soiounlock(&(so)->so_snd_sx)
322 #define SOCK_IO_SEND_OWNED(so) sx_xlocked(&(so)->so_snd_sx)
323 #define SOCK_IO_RECV_LOCK(so, flags) \
324 soiolock((so), &(so)->so_rcv_sx, (flags))
325 #define SOCK_IO_RECV_UNLOCK(so) \
326 soiounlock(&(so)->so_rcv_sx)
327 #define SOCK_IO_RECV_OWNED(so) sx_xlocked(&(so)->so_rcv_sx)
329 /* do we have to send all at once on a socket? */
330 #define sosendallatonce(so) \
331 ((so)->so_proto->pr_flags & PR_ATOMIC)
333 /* can we read something from so? */
334 #define soreadabledata(so) \
335 (sbavail(&(so)->so_rcv) >= (so)->so_rcv.sb_lowat || \
336 (so)->so_error || (so)->so_rerror)
337 #define soreadable(so) \
338 (soreadabledata(so) || ((so)->so_rcv.sb_state & SBS_CANTRCVMORE))
340 /* can we write something to so? */
341 #define sowriteable(so) \
342 ((sbspace(&(so)->so_snd) >= (so)->so_snd.sb_lowat && \
343 (((so)->so_state&SS_ISCONNECTED) || \
344 ((so)->so_proto->pr_flags&PR_CONNREQUIRED)==0)) || \
345 ((so)->so_snd.sb_state & SBS_CANTSENDMORE) || \
349 * soref()/sorele() ref-count the socket structure.
350 * soref() may be called without owning socket lock, but in that case a
351 * caller must own something that holds socket, and so_count must be not 0.
352 * Note that you must still explicitly close the socket, but the last ref
353 * count will free the structure.
355 #define soref(so) refcount_acquire(&(so)->so_count)
356 #define sorele(so) do { \
357 SOCK_UNLOCK_ASSERT(so); \
358 if (!refcount_release_if_not_last(&(so)->so_count)) { \
365 * In sorwakeup() and sowwakeup(), acquire the socket buffer lock to
366 * avoid a non-atomic test-and-wakeup. However, sowakeup is
367 * responsible for releasing the lock if it is called. We unlock only
368 * if we don't call into sowakeup. If any code is introduced that
369 * directly invokes the underlying sowakeup() primitives, it must
370 * maintain the same semantics.
372 #define sorwakeup(so) do { \
373 SOCK_RECVBUF_LOCK(so); \
374 sorwakeup_locked(so); \
377 #define sowwakeup(so) do { \
378 SOCK_SENDBUF_LOCK(so); \
379 sowwakeup_locked(so); \
382 struct accept_filter {
385 (struct socket *so, void *arg, int waitflag);
386 void * (*accf_create)
387 (struct socket *so, char *arg);
390 SLIST_ENTRY(accept_filter) accf_next;
393 #define ACCEPT_FILTER_DEFINE(modname, filtname, cb, create, destroy, ver) \
394 static struct accept_filter modname##_filter = { \
395 .accf_name = filtname, \
396 .accf_callback = cb, \
397 .accf_create = create, \
398 .accf_destroy = destroy, \
400 static moduledata_t modname##_mod = { \
401 .name = __XSTRING(modname), \
402 .evhand = accept_filt_generic_mod_event, \
403 .priv = &modname##_filter, \
405 DECLARE_MODULE(modname, modname##_mod, SI_SUB_DRIVERS, \
407 MODULE_VERSION(modname, ver)
409 #ifdef MALLOC_DECLARE
410 MALLOC_DECLARE(M_ACCF);
411 MALLOC_DECLARE(M_PCB);
412 MALLOC_DECLARE(M_SONAME);
416 * Socket specific helper hook point identifiers
417 * Do not leave holes in the sequence, hook registration is a loop.
419 #define HHOOK_SOCKET_OPT 0
420 #define HHOOK_SOCKET_CREATE 1
421 #define HHOOK_SOCKET_RCV 2
422 #define HHOOK_SOCKET_SND 3
423 #define HHOOK_FILT_SOREAD 4
424 #define HHOOK_FILT_SOWRITE 5
425 #define HHOOK_SOCKET_CLOSE 6
426 #define HHOOK_SOCKET_LAST HHOOK_SOCKET_CLOSE
428 struct socket_hhook_data {
431 void *hctx; /* hook point specific data*/
435 extern int maxsockets;
436 extern u_long sb_max;
437 extern so_gen_t so_gencnt;
447 /* Return values for socket upcalls. */
449 #define SU_ISCONNECTED 1
452 * From uipc_socket and friends
454 int getsockaddr(struct sockaddr **namp, const struct sockaddr *uaddr,
456 int getsock_cap(struct thread *td, int fd, cap_rights_t *rightsp,
457 struct file **fpp, u_int *fflagp, struct filecaps *havecaps);
458 void soabort(struct socket *so);
459 int soaccept(struct socket *so, struct sockaddr **nam);
460 void soaio_enqueue(struct task *task);
461 void soaio_rcv(void *context, int pending);
462 void soaio_snd(void *context, int pending);
463 int socheckuid(struct socket *so, uid_t uid);
464 int sobind(struct socket *so, struct sockaddr *nam, struct thread *td);
465 int sobindat(int fd, struct socket *so, struct sockaddr *nam,
467 int soclose(struct socket *so);
468 int soconnect(struct socket *so, struct sockaddr *nam, struct thread *td);
469 int soconnectat(int fd, struct socket *so, struct sockaddr *nam,
471 int soconnect2(struct socket *so1, struct socket *so2);
472 int socreate(int dom, struct socket **aso, int type, int proto,
473 struct ucred *cred, struct thread *td);
474 int sodisconnect(struct socket *so);
475 void sodtor_set(struct socket *, so_dtor_t *);
476 struct sockaddr *sodupsockaddr(const struct sockaddr *sa, int mflags);
477 void sofree(struct socket *so);
478 void sohasoutofband(struct socket *so);
479 int solisten(struct socket *so, int backlog, struct thread *td);
480 void solisten_proto(struct socket *so, int backlog);
481 void solisten_proto_abort(struct socket *so);
482 int solisten_proto_check(struct socket *so);
483 int solisten_dequeue(struct socket *, struct socket **, int);
485 sonewconn(struct socket *head, int connstatus);
487 sopeeloff(struct socket *);
488 int sopoll(struct socket *so, int events, struct ucred *active_cred,
490 int sopoll_generic(struct socket *so, int events,
491 struct ucred *active_cred, struct thread *td);
492 int soreceive(struct socket *so, struct sockaddr **paddr, struct uio *uio,
493 struct mbuf **mp0, struct mbuf **controlp, int *flagsp);
494 int soreceive_stream(struct socket *so, struct sockaddr **paddr,
495 struct uio *uio, struct mbuf **mp0, struct mbuf **controlp,
497 int soreceive_dgram(struct socket *so, struct sockaddr **paddr,
498 struct uio *uio, struct mbuf **mp0, struct mbuf **controlp,
500 int soreceive_generic(struct socket *so, struct sockaddr **paddr,
501 struct uio *uio, struct mbuf **mp0, struct mbuf **controlp,
503 void sorele_locked(struct socket *so);
504 int soreserve(struct socket *so, u_long sndcc, u_long rcvcc);
505 void sorflush(struct socket *so);
506 int sosend(struct socket *so, struct sockaddr *addr, struct uio *uio,
507 struct mbuf *top, struct mbuf *control, int flags,
509 int sosend_dgram(struct socket *so, struct sockaddr *addr,
510 struct uio *uio, struct mbuf *top, struct mbuf *control,
511 int flags, struct thread *td);
512 int sosend_generic(struct socket *so, struct sockaddr *addr,
513 struct uio *uio, struct mbuf *top, struct mbuf *control,
514 int flags, struct thread *td);
515 int soshutdown(struct socket *so, int how);
516 void soupcall_clear(struct socket *, sb_which);
517 void soupcall_set(struct socket *, sb_which, so_upcall_t, void *);
518 void solisten_upcall_set(struct socket *, so_upcall_t, void *);
519 void sorwakeup_locked(struct socket *);
520 void sowwakeup_locked(struct socket *);
521 void sowakeup_aio(struct socket *, sb_which);
522 void solisten_wakeup(struct socket *);
523 int selsocket(struct socket *so, int events, struct timeval *tv,
525 void soisconnected(struct socket *so);
526 void soisconnecting(struct socket *so);
527 void soisdisconnected(struct socket *so);
528 void soisdisconnecting(struct socket *so);
529 void socantrcvmore(struct socket *so);
530 void socantrcvmore_locked(struct socket *so);
531 void socantsendmore(struct socket *so);
532 void socantsendmore_locked(struct socket *so);
533 void soroverflow(struct socket *so);
534 void soroverflow_locked(struct socket *so);
535 int soiolock(struct socket *so, struct sx *sx, int flags);
536 void soiounlock(struct sx *sx);
539 * Accept filter functions (duh).
541 int accept_filt_add(struct accept_filter *filt);
542 int accept_filt_del(char *name);
543 struct accept_filter *accept_filt_get(char *name);
544 #ifdef ACCEPT_FILTER_MOD
546 SYSCTL_DECL(_net_inet_accf);
548 int accept_filt_generic_mod_event(module_t mod, int event, void *data);
554 * Structure to export socket from kernel to utilities, via sysctl(3).
557 ksize_t xso_len; /* length of this structure */
558 kvaddr_t xso_so; /* kernel address of struct socket */
559 kvaddr_t so_pcb; /* kernel address of struct inpcb */
561 int64_t so_spare64[8];
562 int32_t xso_protocol;
569 int32_t so_spare32[8];
590 void sotoxsocket(struct socket *so, struct xsocket *xso);
591 void sbtoxsockbuf(struct sockbuf *sb, struct xsockbuf *xsb);
595 * Socket buffer state bits. Exported via libprocstat(3).
597 #define SBS_CANTSENDMORE 0x0010 /* can't send more data to peer */
598 #define SBS_CANTRCVMORE 0x0020 /* can't receive more data from peer */
599 #define SBS_RCVATMARK 0x0040 /* at mark on input */
601 #endif /* !_SYS_SOCKETVAR_H_ */