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 SOCKBUF_LOCK(&so->so_rcv).
81 * (cs) locked by SOCKBUF_LOCK(&so->so_snd).
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 short so_type; /* (a) generic type, see socket.h */
95 int so_options; /* (b) from socket call, see socket.h */
96 short so_linger; /* time to linger close(2) */
97 short so_state; /* (b) internal state flags SS_* */
98 void *so_pcb; /* protocol control block */
99 struct vnet *so_vnet; /* (a) network stack instance */
100 struct protosw *so_proto; /* (a) protocol handle */
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 */
125 /* Regular (data flow) socket. */
127 /* (cr, cs) Receive and send buffers. */
128 struct sockbuf so_rcv, so_snd;
130 /* (e) Our place on accept queue. */
131 TAILQ_ENTRY(socket) so_list;
132 struct socket *so_listen; /* (b) */
133 enum socket_qstate so_qstate; /* (b) */
134 /* (b) cached MAC label for peer */
135 struct label *so_peerlabel;
136 u_long so_oobmark; /* chars to oob mark */
138 /* (k) Our place on KTLS RX work queue. */
139 STAILQ_ENTRY(socket) so_ktls_rx_list;
142 * Listening socket, where accepts occur, is so_listen in all
143 * subsidiary sockets. If so_listen is NULL, socket is not
144 * related to an accept. For a listening socket itself
145 * sol_incomp queues partially completed connections, while
146 * sol_comp is a queue of connections ready to be accepted.
147 * If a connection is aborted and it has so_listen set, then
148 * it has to be pulled out of either sol_incomp or sol_comp.
149 * We allow connections to queue up based on current queue
150 * lengths and limit on number of queued connections for this
154 /* (e) queue of partial unaccepted connections */
155 struct accept_queue sol_incomp;
156 /* (e) queue of complete unaccepted connections */
157 struct accept_queue sol_comp;
158 u_int sol_qlen; /* (e) sol_comp length */
159 u_int sol_incqlen; /* (e) sol_incomp length */
160 u_int sol_qlimit; /* (e) queue limit */
162 /* accept_filter(9) optional data */
163 struct accept_filter *sol_accept_filter;
164 void *sol_accept_filter_arg; /* saved filter args */
165 char *sol_accept_filter_str; /* saved user args */
167 /* Optional upcall, for kernel socket. */
168 so_upcall_t *sol_upcall; /* (e) */
169 void *sol_upcallarg; /* (e) */
171 /* Socket buffer parameters, to be copied to
172 * dataflow sockets, accepted from this one. */
175 u_int sol_sbrcv_hiwat;
176 u_int sol_sbsnd_hiwat;
177 short sol_sbrcv_flags;
178 short sol_sbsnd_flags;
179 sbintime_t sol_sbrcv_timeo;
180 sbintime_t sol_sbsnd_timeo;
182 /* Information tracking listen queue overflows. */
183 struct timeval sol_lastover; /* (e) */
184 int sol_overcount; /* (e) */
188 #endif /* defined(_KERNEL) || defined(_WANT_SOCKET) */
193 * Historically, these bits were all kept in the so_state field.
194 * They are now split into separate, lock-specific fields.
195 * so_state maintains basic socket state protected by the socket lock.
196 * so_qstate holds information about the socket accept queues.
197 * Each socket buffer also has a state field holding information
198 * relevant to that socket buffer (can't send, rcv).
199 * Many fields will be read without locks to improve performance and avoid
200 * lock order issues. However, this approach must be used with caution.
202 #define SS_NOFDREF 0x0001 /* no file table ref any more */
203 #define SS_ISCONNECTED 0x0002 /* socket connected to a peer */
204 #define SS_ISCONNECTING 0x0004 /* in process of connecting to peer */
205 #define SS_ISDISCONNECTING 0x0008 /* in process of disconnecting */
206 #define SS_NBIO 0x0100 /* non-blocking ops */
207 #define SS_ASYNC 0x0200 /* async i/o notify */
208 #define SS_ISCONFIRMING 0x0400 /* deciding to accept connection req */
209 #define SS_ISDISCONNECTED 0x2000 /* socket disconnected from peer */
212 * Protocols can mark a socket as SS_PROTOREF to indicate that, following
213 * pru_detach, they still want the socket to persist, and will free it
214 * themselves when they are done. Protocols should only ever call sofree()
215 * following setting this flag in pru_detach(), and never otherwise, as
216 * sofree() bypasses socket reference counting.
218 #define SS_PROTOREF 0x4000 /* strong protocol reference */
222 #define SOCK_MTX(so) (&(so)->so_lock)
223 #define SOCK_LOCK(so) mtx_lock(&(so)->so_lock)
224 #define SOCK_OWNED(so) mtx_owned(&(so)->so_lock)
225 #define SOCK_UNLOCK(so) mtx_unlock(&(so)->so_lock)
226 #define SOCK_LOCK_ASSERT(so) mtx_assert(&(so)->so_lock, MA_OWNED)
227 #define SOCK_UNLOCK_ASSERT(so) mtx_assert(&(so)->so_lock, MA_NOTOWNED)
229 #define SOLISTENING(sol) (((sol)->so_options & SO_ACCEPTCONN) != 0)
230 #define SOLISTEN_LOCK(sol) do { \
231 mtx_lock(&(sol)->so_lock); \
232 KASSERT(SOLISTENING(sol), \
233 ("%s: %p not listening", __func__, (sol))); \
235 #define SOLISTEN_TRYLOCK(sol) mtx_trylock(&(sol)->so_lock)
236 #define SOLISTEN_UNLOCK(sol) do { \
237 KASSERT(SOLISTENING(sol), \
238 ("%s: %p not listening", __func__, (sol))); \
239 mtx_unlock(&(sol)->so_lock); \
241 #define SOLISTEN_LOCK_ASSERT(sol) do { \
242 mtx_assert(&(sol)->so_lock, MA_OWNED); \
243 KASSERT(SOLISTENING(sol), \
244 ("%s: %p not listening", __func__, (sol))); \
248 * Macros for sockets and socket buffering.
254 #define SBL_WAIT 0x00000001 /* Wait if not immediately available. */
255 #define SBL_NOINTR 0x00000002 /* Force non-interruptible sleep. */
256 #define SBL_VALID (SBL_WAIT | SBL_NOINTR)
259 * Do we need to notify the other side when I/O is possible?
261 #define sb_notify(sb) (((sb)->sb_flags & (SB_WAIT | SB_SEL | SB_ASYNC | \
262 SB_UPCALL | SB_AIO | SB_KNOTE)) != 0)
264 /* do we have to send all at once on a socket? */
265 #define sosendallatonce(so) \
266 ((so)->so_proto->pr_flags & PR_ATOMIC)
268 /* can we read something from so? */
269 #define soreadabledata(so) \
270 (sbavail(&(so)->so_rcv) >= (so)->so_rcv.sb_lowat || \
271 (so)->so_error || (so)->so_rerror)
272 #define soreadable(so) \
273 (soreadabledata(so) || ((so)->so_rcv.sb_state & SBS_CANTRCVMORE))
275 /* can we write something to so? */
276 #define sowriteable(so) \
277 ((sbspace(&(so)->so_snd) >= (so)->so_snd.sb_lowat && \
278 (((so)->so_state&SS_ISCONNECTED) || \
279 ((so)->so_proto->pr_flags&PR_CONNREQUIRED)==0)) || \
280 ((so)->so_snd.sb_state & SBS_CANTSENDMORE) || \
284 * soref()/sorele() ref-count the socket structure.
285 * soref() may be called without owning socket lock, but in that case a
286 * caller must own something that holds socket, and so_count must be not 0.
287 * Note that you must still explicitly close the socket, but the last ref
288 * count will free the structure.
290 #define soref(so) refcount_acquire(&(so)->so_count)
291 #define sorele(so) do { \
292 SOCK_LOCK_ASSERT(so); \
293 if (refcount_release(&(so)->so_count)) \
300 * In sorwakeup() and sowwakeup(), acquire the socket buffer lock to
301 * avoid a non-atomic test-and-wakeup. However, sowakeup is
302 * responsible for releasing the lock if it is called. We unlock only
303 * if we don't call into sowakeup. If any code is introduced that
304 * directly invokes the underlying sowakeup() primitives, it must
305 * maintain the same semantics.
307 #define sorwakeup_locked(so) do { \
308 SOCKBUF_LOCK_ASSERT(&(so)->so_rcv); \
309 if (sb_notify(&(so)->so_rcv)) \
310 sowakeup((so), &(so)->so_rcv); \
312 SOCKBUF_UNLOCK(&(so)->so_rcv); \
315 #define sorwakeup(so) do { \
316 SOCKBUF_LOCK(&(so)->so_rcv); \
317 sorwakeup_locked(so); \
320 #define sowwakeup_locked(so) do { \
321 SOCKBUF_LOCK_ASSERT(&(so)->so_snd); \
322 if (sb_notify(&(so)->so_snd)) \
323 sowakeup((so), &(so)->so_snd); \
325 SOCKBUF_UNLOCK(&(so)->so_snd); \
328 #define sowwakeup(so) do { \
329 SOCKBUF_LOCK(&(so)->so_snd); \
330 sowwakeup_locked(so); \
333 struct accept_filter {
336 (struct socket *so, void *arg, int waitflag);
337 void * (*accf_create)
338 (struct socket *so, char *arg);
341 SLIST_ENTRY(accept_filter) accf_next;
344 #define ACCEPT_FILTER_DEFINE(modname, filtname, cb, create, destroy, ver) \
345 static struct accept_filter modname##_filter = { \
346 .accf_name = filtname, \
347 .accf_callback = cb, \
348 .accf_create = create, \
349 .accf_destroy = destroy, \
351 static moduledata_t modname##_mod = { \
352 .name = __XSTRING(modname), \
353 .evhand = accept_filt_generic_mod_event, \
354 .priv = &modname##_filter, \
356 DECLARE_MODULE(modname, modname##_mod, SI_SUB_DRIVERS, \
358 MODULE_VERSION(modname, ver)
360 #ifdef MALLOC_DECLARE
361 MALLOC_DECLARE(M_ACCF);
362 MALLOC_DECLARE(M_PCB);
363 MALLOC_DECLARE(M_SONAME);
367 * Socket specific helper hook point identifiers
368 * Do not leave holes in the sequence, hook registration is a loop.
370 #define HHOOK_SOCKET_OPT 0
371 #define HHOOK_SOCKET_CREATE 1
372 #define HHOOK_SOCKET_RCV 2
373 #define HHOOK_SOCKET_SND 3
374 #define HHOOK_FILT_SOREAD 4
375 #define HHOOK_FILT_SOWRITE 5
376 #define HHOOK_SOCKET_CLOSE 6
377 #define HHOOK_SOCKET_LAST HHOOK_SOCKET_CLOSE
379 struct socket_hhook_data {
382 void *hctx; /* hook point specific data*/
386 extern int maxsockets;
387 extern u_long sb_max;
388 extern so_gen_t so_gencnt;
398 /* 'which' values for socket upcalls. */
402 /* Return values for socket upcalls. */
404 #define SU_ISCONNECTED 1
407 * From uipc_socket and friends
409 int getsockaddr(struct sockaddr **namp, const struct sockaddr *uaddr,
411 int getsock_cap(struct thread *td, int fd, cap_rights_t *rightsp,
412 struct file **fpp, u_int *fflagp, struct filecaps *havecaps);
413 void soabort(struct socket *so);
414 int soaccept(struct socket *so, struct sockaddr **nam);
415 void soaio_enqueue(struct task *task);
416 void soaio_rcv(void *context, int pending);
417 void soaio_snd(void *context, int pending);
418 int socheckuid(struct socket *so, uid_t uid);
419 int sobind(struct socket *so, struct sockaddr *nam, struct thread *td);
420 int sobindat(int fd, struct socket *so, struct sockaddr *nam,
422 int soclose(struct socket *so);
423 int soconnect(struct socket *so, struct sockaddr *nam, struct thread *td);
424 int soconnectat(int fd, struct socket *so, struct sockaddr *nam,
426 int soconnect2(struct socket *so1, struct socket *so2);
427 int socreate(int dom, struct socket **aso, int type, int proto,
428 struct ucred *cred, struct thread *td);
429 int sodisconnect(struct socket *so);
430 void sodtor_set(struct socket *, so_dtor_t *);
431 struct sockaddr *sodupsockaddr(const struct sockaddr *sa, int mflags);
432 void sofree(struct socket *so);
433 void sohasoutofband(struct socket *so);
434 int solisten(struct socket *so, int backlog, struct thread *td);
435 void solisten_proto(struct socket *so, int backlog);
436 int solisten_proto_check(struct socket *so);
437 int solisten_dequeue(struct socket *, struct socket **, int);
439 sonewconn(struct socket *head, int connstatus);
441 sopeeloff(struct socket *);
442 int sopoll(struct socket *so, int events, struct ucred *active_cred,
444 int sopoll_generic(struct socket *so, int events,
445 struct ucred *active_cred, struct thread *td);
446 int soreceive(struct socket *so, struct sockaddr **paddr, struct uio *uio,
447 struct mbuf **mp0, struct mbuf **controlp, int *flagsp);
448 int soreceive_stream(struct socket *so, struct sockaddr **paddr,
449 struct uio *uio, struct mbuf **mp0, struct mbuf **controlp,
451 int soreceive_dgram(struct socket *so, struct sockaddr **paddr,
452 struct uio *uio, struct mbuf **mp0, struct mbuf **controlp,
454 int soreceive_generic(struct socket *so, struct sockaddr **paddr,
455 struct uio *uio, struct mbuf **mp0, struct mbuf **controlp,
457 int soreserve(struct socket *so, u_long sndcc, u_long rcvcc);
458 void sorflush(struct socket *so);
459 int sosend(struct socket *so, struct sockaddr *addr, struct uio *uio,
460 struct mbuf *top, struct mbuf *control, int flags,
462 int sosend_dgram(struct socket *so, struct sockaddr *addr,
463 struct uio *uio, struct mbuf *top, struct mbuf *control,
464 int flags, struct thread *td);
465 int sosend_generic(struct socket *so, struct sockaddr *addr,
466 struct uio *uio, struct mbuf *top, struct mbuf *control,
467 int flags, struct thread *td);
468 int soshutdown(struct socket *so, int how);
469 void soupcall_clear(struct socket *, int);
470 void soupcall_set(struct socket *, int, so_upcall_t, void *);
471 void solisten_upcall_set(struct socket *, so_upcall_t, void *);
472 void sowakeup(struct socket *so, struct sockbuf *sb);
473 void sowakeup_aio(struct socket *so, struct sockbuf *sb);
474 void solisten_wakeup(struct socket *);
475 int selsocket(struct socket *so, int events, struct timeval *tv,
477 void soisconnected(struct socket *so);
478 void soisconnecting(struct socket *so);
479 void soisdisconnected(struct socket *so);
480 void soisdisconnecting(struct socket *so);
481 void socantrcvmore(struct socket *so);
482 void socantrcvmore_locked(struct socket *so);
483 void socantsendmore(struct socket *so);
484 void socantsendmore_locked(struct socket *so);
485 void soroverflow(struct socket *so);
486 void soroverflow_locked(struct socket *so);
489 * Accept filter functions (duh).
491 int accept_filt_add(struct accept_filter *filt);
492 int accept_filt_del(char *name);
493 struct accept_filter *accept_filt_get(char *name);
494 #ifdef ACCEPT_FILTER_MOD
496 SYSCTL_DECL(_net_inet_accf);
498 int accept_filt_generic_mod_event(module_t mod, int event, void *data);
504 * Structure to export socket from kernel to utilities, via sysctl(3).
507 ksize_t xso_len; /* length of this structure */
508 kvaddr_t xso_so; /* kernel address of struct socket */
509 kvaddr_t so_pcb; /* kernel address of struct inpcb */
511 int64_t so_spare64[8];
512 int32_t xso_protocol;
519 int32_t so_spare32[8];
540 void sotoxsocket(struct socket *so, struct xsocket *xso);
541 void sbtoxsockbuf(struct sockbuf *sb, struct xsockbuf *xsb);
545 * Socket buffer state bits. Exported via libprocstat(3).
547 #define SBS_CANTSENDMORE 0x0010 /* can't send more data to peer */
548 #define SBS_CANTRCVMORE 0x0020 /* can't receive more data from peer */
549 #define SBS_RCVATMARK 0x0040 /* at mark on input */
551 #endif /* !_SYS_SOCKETVAR_H_ */