2 * Copyright (c) 1982, 1986, 1988, 1990, 1993
3 * The Regents of the University of California. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 3. All advertising materials mentioning features or use of this software
14 * must display the following acknowledgement:
15 * This product includes software developed by the University of
16 * California, Berkeley and its contributors.
17 * 4. Neither the name of the University nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 * @(#)uipc_socket2.c 8.1 (Berkeley) 6/10/93
34 * $FreeBSD: src/sys/kern/uipc_socket2.c,v 1.55.2.17 2002/08/31 19:04:55 dwmalone Exp $
35 * $DragonFly: src/sys/kern/uipc_socket2.c,v 1.17 2005/04/20 09:28:29 hsu Exp $
38 #include "opt_param.h"
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/domain.h>
42 #include <sys/file.h> /* for maxfiles */
43 #include <sys/kernel.h>
45 #include <sys/malloc.h>
47 #include <sys/protosw.h>
48 #include <sys/resourcevar.h>
50 #include <sys/socket.h>
51 #include <sys/socketvar.h>
52 #include <sys/signalvar.h>
53 #include <sys/sysctl.h>
54 #include <sys/aio.h> /* for aio_swake proto */
55 #include <sys/event.h>
57 #include <sys/thread2.h>
58 #include <sys/msgport2.h>
63 * Primitive routines for operating on sockets and socket buffers
66 u_long sb_max = SB_MAX;
68 SB_MAX * MCLBYTES / (MSIZE + MCLBYTES); /* adjusted sb_max */
70 static u_long sb_efficiency = 8; /* parameter for sbreserve() */
73 * Procedures to manipulate state flags of socket
74 * and do appropriate wakeups. Normal sequence from the
75 * active (originating) side is that soisconnecting() is
76 * called during processing of connect() call,
77 * resulting in an eventual call to soisconnected() if/when the
78 * connection is established. When the connection is torn down
79 * soisdisconnecting() is called during processing of disconnect() call,
80 * and soisdisconnected() is called when the connection to the peer
81 * is totally severed. The semantics of these routines are such that
82 * connectionless protocols can call soisconnected() and soisdisconnected()
83 * only, bypassing the in-progress calls when setting up a ``connection''
86 * From the passive side, a socket is created with
87 * two queues of sockets: so_incomp for connections in progress
88 * and so_comp for connections already made and awaiting user acceptance.
89 * As a protocol is preparing incoming connections, it creates a socket
90 * structure queued on so_incomp by calling sonewconn(). When the connection
91 * is established, soisconnected() is called, and transfers the
92 * socket structure to so_comp, making it available to accept().
94 * If a socket is closed with sockets on either
95 * so_incomp or so_comp, these sockets are dropped.
97 * If higher level protocols are implemented in
98 * the kernel, the wakeups done here will sometimes
99 * cause software-interrupt process scheduling.
107 so->so_state &= ~(SS_ISCONNECTED|SS_ISDISCONNECTING);
108 so->so_state |= SS_ISCONNECTING;
115 struct socket *head = so->so_head;
117 so->so_state &= ~(SS_ISCONNECTING|SS_ISDISCONNECTING|SS_ISCONFIRMING);
118 so->so_state |= SS_ISCONNECTED;
119 if (head && (so->so_state & SS_INCOMP)) {
120 if ((so->so_options & SO_ACCEPTFILTER) != 0) {
121 so->so_upcall = head->so_accf->so_accept_filter->accf_callback;
122 so->so_upcallarg = head->so_accf->so_accept_filter_arg;
123 so->so_rcv.sb_flags |= SB_UPCALL;
124 so->so_options &= ~SO_ACCEPTFILTER;
125 so->so_upcall(so, so->so_upcallarg, 0);
128 TAILQ_REMOVE(&head->so_incomp, so, so_list);
130 so->so_state &= ~SS_INCOMP;
131 TAILQ_INSERT_TAIL(&head->so_comp, so, so_list);
133 so->so_state |= SS_COMP;
135 wakeup_one(&head->so_timeo);
137 wakeup(&so->so_timeo);
144 soisdisconnecting(so)
148 so->so_state &= ~SS_ISCONNECTING;
149 so->so_state |= (SS_ISDISCONNECTING|SS_CANTRCVMORE|SS_CANTSENDMORE);
150 wakeup((caddr_t)&so->so_timeo);
160 so->so_state &= ~(SS_ISCONNECTING|SS_ISCONNECTED|SS_ISDISCONNECTING);
161 so->so_state |= (SS_CANTRCVMORE|SS_CANTSENDMORE|SS_ISDISCONNECTED);
162 wakeup((caddr_t)&so->so_timeo);
163 sbdrop(&so->so_snd, so->so_snd.sb_cc);
169 * When an attempt at a new connection is noted on a socket
170 * which accepts connections, sonewconn is called. If the
171 * connection is possible (subject to space constraints, etc.)
172 * then we allocate a new structure, propoerly linked into the
173 * data structure of the original socket, and return this.
174 * Connstatus may be 0, or SO_ISCONFIRMING, or SO_ISCONNECTED.
177 sonewconn(struct socket *head, int connstatus)
180 struct pru_attach_info ai;
182 if (head->so_qlen > 3 * head->so_qlimit / 2)
183 return ((struct socket *)0);
186 return ((struct socket *)0);
187 if ((head->so_options & SO_ACCEPTFILTER) != 0)
190 so->so_type = head->so_type;
191 so->so_options = head->so_options &~ SO_ACCEPTCONN;
192 so->so_linger = head->so_linger;
193 so->so_state = head->so_state | SS_NOFDREF;
194 so->so_proto = head->so_proto;
195 so->so_timeo = head->so_timeo;
196 so->so_cred = crhold(head->so_cred);
199 ai.fd_rdir = NULL; /* jail code cruft XXX JH */
200 if (soreserve(so, head->so_snd.sb_hiwat, head->so_rcv.sb_hiwat, NULL) ||
201 /* Directly call function since we're already at protocol level. */
202 (*so->so_proto->pr_usrreqs->pru_attach)(so, 0, &ai)) {
204 return ((struct socket *)0);
208 TAILQ_INSERT_TAIL(&head->so_comp, so, so_list);
209 so->so_state |= SS_COMP;
212 if (head->so_incqlen > head->so_qlimit) {
214 sp = TAILQ_FIRST(&head->so_incomp);
217 TAILQ_INSERT_TAIL(&head->so_incomp, so, so_list);
218 so->so_state |= SS_INCOMP;
223 wakeup((caddr_t)&head->so_timeo);
224 so->so_state |= connstatus;
230 * Socantsendmore indicates that no more data will be sent on the
231 * socket; it would normally be applied to a socket when the user
232 * informs the system that no more data is to be sent, by the protocol
233 * code (in case PRU_SHUTDOWN). Socantrcvmore indicates that no more data
234 * will be received, and will normally be applied to the socket by a
235 * protocol when it detects that the peer will send no more data.
236 * Data queued for reading in the socket may yet be read.
244 so->so_state |= SS_CANTSENDMORE;
253 so->so_state |= SS_CANTRCVMORE;
258 * Wait for data to arrive at/drain from a socket buffer.
265 sb->sb_flags |= SB_WAIT;
266 return (tsleep((caddr_t)&sb->sb_cc,
267 ((sb->sb_flags & SB_NOINTR) ? 0 : PCATCH),
273 * Lock a sockbuf already known to be locked;
274 * return any error returned from sleep (EINTR).
282 while (sb->sb_flags & SB_LOCK) {
283 sb->sb_flags |= SB_WANT;
284 error = tsleep((caddr_t)&sb->sb_flags,
285 ((sb->sb_flags & SB_NOINTR) ? 0 : PCATCH),
290 sb->sb_flags |= SB_LOCK;
295 * Wakeup processes waiting on a socket buffer. Do asynchronous notification
296 * via SIGIO if the socket has the SS_ASYNC flag set.
303 struct selinfo *selinfo = &sb->sb_sel;
306 sb->sb_flags &= ~SB_SEL;
307 if (sb->sb_flags & SB_WAIT) {
308 sb->sb_flags &= ~SB_WAIT;
309 wakeup((caddr_t)&sb->sb_cc);
311 if ((so->so_state & SS_ASYNC) && so->so_sigio != NULL)
312 pgsigio(so->so_sigio, SIGIO, 0);
313 if (sb->sb_flags & SB_UPCALL)
314 (*so->so_upcall)(so, so->so_upcallarg, MB_DONTWAIT);
315 if (sb->sb_flags & SB_AIO)
317 KNOTE(&selinfo->si_note, 0);
318 if (sb->sb_flags & SB_MEVENT) {
319 struct netmsg_so_notify *msg, *nmsg;
321 TAILQ_FOREACH_MUTABLE(msg, &selinfo->si_mlist, nm_list, nmsg) {
322 if (msg->nm_predicate((struct netmsg *)msg)) {
323 TAILQ_REMOVE(&selinfo->si_mlist, msg, nm_list);
324 lwkt_replymsg(&msg->nm_lmsg,
325 msg->nm_lmsg.ms_error);
328 if (TAILQ_EMPTY(&sb->sb_sel.si_mlist))
329 sb->sb_flags &= ~SB_MEVENT;
334 * Socket buffer (struct sockbuf) utility routines.
336 * Each socket contains two socket buffers: one for sending data and
337 * one for receiving data. Each buffer contains a queue of mbufs,
338 * information about the number of mbufs and amount of data in the
339 * queue, and other fields allowing select() statements and notification
340 * on data availability to be implemented.
342 * Data stored in a socket buffer is maintained as a list of records.
343 * Each record is a list of mbufs chained together with the m_next
344 * field. Records are chained together with the m_nextpkt field. The upper
345 * level routine soreceive() expects the following conventions to be
346 * observed when placing information in the receive buffer:
348 * 1. If the protocol requires each message be preceded by the sender's
349 * name, then a record containing that name must be present before
350 * any associated data (mbuf's must be of type MT_SONAME).
351 * 2. If the protocol supports the exchange of ``access rights'' (really
352 * just additional data associated with the message), and there are
353 * ``rights'' to be received, then a record containing this data
354 * should be present (mbuf's must be of type MT_RIGHTS).
355 * 3. If a name or rights record exists, then it must be followed by
356 * a data record, perhaps of zero length.
358 * Before using a new socket structure it is first necessary to reserve
359 * buffer space to the socket, by calling sbreserve(). This should commit
360 * some of the available buffer space in the system buffer pool for the
361 * socket (currently, it does nothing but enforce limits). The space
362 * should be released by calling sbrelease() when the socket is destroyed.
366 soreserve(struct socket *so, u_long sndcc, u_long rcvcc, struct rlimit *rl)
368 if (sbreserve(&so->so_snd, sndcc, so, rl) == 0)
370 if (sbreserve(&so->so_rcv, rcvcc, so, rl) == 0)
372 if (so->so_rcv.sb_lowat == 0)
373 so->so_rcv.sb_lowat = 1;
374 if (so->so_snd.sb_lowat == 0)
375 so->so_snd.sb_lowat = MCLBYTES;
376 if (so->so_snd.sb_lowat > so->so_snd.sb_hiwat)
377 so->so_snd.sb_lowat = so->so_snd.sb_hiwat;
380 sbrelease(&so->so_snd, so);
386 sysctl_handle_sb_max(SYSCTL_HANDLER_ARGS)
389 u_long old_sb_max = sb_max;
391 error = SYSCTL_OUT(req, arg1, sizeof(int));
392 if (error || !req->newptr)
394 error = SYSCTL_IN(req, arg1, sizeof(int));
397 if (sb_max < MSIZE + MCLBYTES) {
401 sb_max_adj = (u_quad_t)sb_max * MCLBYTES / (MSIZE + MCLBYTES);
406 * Allot mbufs to a sockbuf.
407 * Attempt to scale mbmax so that mbcnt doesn't become limiting
408 * if buffering efficiency is near the normal case.
411 sbreserve(struct sockbuf *sb, u_long cc, struct socket *so, struct rlimit *rl)
415 * rl will only be NULL when we're in an interrupt (eg, in tcp_input)
416 * or when called from netgraph (ie, ngd_attach)
420 if (!chgsbsize(so->so_cred->cr_uidinfo, &sb->sb_hiwat, cc,
421 rl ? rl->rlim_cur : RLIM_INFINITY)) {
424 sb->sb_mbmax = min(cc * sb_efficiency, sb_max);
425 if (sb->sb_lowat > sb->sb_hiwat)
426 sb->sb_lowat = sb->sb_hiwat;
431 * Free mbufs held by a socket, and reserved mbuf space.
440 (void)chgsbsize(so->so_cred->cr_uidinfo, &sb->sb_hiwat, 0,
446 * Routines to add and remove
447 * data from an mbuf queue.
449 * The routines sbappend() or sbappendrecord() are normally called to
450 * append new mbufs to a socket buffer, after checking that adequate
451 * space is available, comparing the function sbspace() with the amount
452 * of data to be added. sbappendrecord() differs from sbappend() in
453 * that data supplied is treated as the beginning of a new record.
454 * To place a sender's address, optional access rights, and data in a
455 * socket receive buffer, sbappendaddr() should be used. To place
456 * access rights and data in a socket receive buffer, sbappendrights()
457 * should be used. In either case, the new data begins a new record.
458 * Note that unlike sbappend() and sbappendrecord(), these routines check
459 * for the caller that there will be enough space to store the data.
460 * Each fails if there is not enough space, or if it cannot find mbufs
461 * to store additional information in.
463 * Reliable protocols may use the socket send buffer to hold data
464 * awaiting acknowledgement. Data is normally copied from a socket
465 * send buffer in a protocol with m_copy for output to a peer,
466 * and then removing the data from the socket buffer with sbdrop()
467 * or sbdroprecord() when the data is acknowledged by the peer.
471 * Append mbuf chain m to the last record in the
472 * socket buffer sb. The additional space associated
473 * the mbuf chain is recorded in sb. Empty mbufs are
474 * discarded and mbufs are compacted where possible.
482 boolean_t wasempty = (sb->sb_mb == NULL);
491 if (n->m_flags & M_EOR) {
492 sbappendrecord(sb, m); /* XXXXXX!!!! */
495 } while (n->m_next && (n = n->m_next));
497 sbcompress(sb, m, n);
499 sb->sb_lastrecord = sb->sb_mb;
503 * sbappendstream() is an optimized form of sbappend() for protocols
504 * such as TCP that only have one record in the socket buffer, are
505 * not PR_ATOMIC, nor allow MT_CONTROL data. A protocol that uses
506 * sbappendstream() must use sbappendstream() exclusively.
509 sbappendstream(struct sockbuf *sb, struct mbuf *m)
511 KKASSERT(m->m_nextpkt == NULL);
512 sbcompress(sb, m, sb->sb_lastmbuf);
522 u_long len = 0, mbcnt = 0;
524 for (m = sb->sb_mb; m; m = n) {
526 for (; m; m = m->m_next) {
529 if (m->m_flags & M_EXT) /*XXX*/ /* pretty sure this is bogus */
530 mbcnt += m->m_ext.ext_size;
533 if (len != sb->sb_cc || mbcnt != sb->sb_mbcnt) {
534 printf("cc %ld != %ld || mbcnt %ld != %ld\n", len, sb->sb_cc,
535 mbcnt, sb->sb_mbcnt);
542 * As above, except the mbuf chain
543 * begins a new record.
546 sbappendrecord(sb, m0)
557 * Put the first mbuf on the queue.
558 * Note this permits zero length records.
561 sb->sb_lastrecord->m_nextpkt = m0;
564 sb->sb_lastrecord = m0;
568 if (m && (m0->m_flags & M_EOR)) {
569 m0->m_flags &= ~M_EOR;
572 sbcompress(sb, m, m0);
576 * As above except that OOB data
577 * is inserted at the beginning of the sockbuf,
578 * but after any other OOB data.
590 for (mp = &sb->sb_mb; *mp ; mp = &((*mp)->m_nextpkt)) {
596 continue; /* WANT next train */
601 goto again; /* inspect THIS train further */
606 * Put the first mbuf on the queue.
607 * Note this permits zero length records.
612 if (m0->m_nextpkt == NULL)
613 sb->sb_lastrecord = m0;
617 if (m && (m0->m_flags & M_EOR)) {
618 m0->m_flags &= ~M_EOR;
621 sbcompress(sb, m, m0);
625 * Append address and data, and optionally, control (ancillary) data
626 * to the receive queue of a socket. If present,
627 * m0 must include a packet header with total length.
628 * Returns 0 if no space in sockbuf or insufficient mbufs.
631 sbappendaddr(sb, asa, m0, control)
633 const struct sockaddr *asa;
634 struct mbuf *m0, *control;
637 int space = asa->sa_len;
639 if (m0 && (m0->m_flags & M_PKTHDR) == 0)
640 panic("sbappendaddr");
643 space += m0->m_pkthdr.len;
644 for (n = control; n; n = n->m_next) {
646 if (n->m_next == 0) /* keep pointer to last control buf */
649 if (space > sbspace(sb))
651 if (asa->sa_len > MLEN)
653 MGET(m, MB_DONTWAIT, MT_SONAME);
656 m->m_len = asa->sa_len;
657 bcopy(asa, mtod(m, caddr_t), asa->sa_len);
659 n->m_next = m0; /* concatenate data to control */
663 for (n = m; n; n = n->m_next)
667 sb->sb_lastrecord->m_nextpkt = m;
670 sb->sb_lastrecord = m;
676 sbappendcontrol(sb, m0, control)
678 struct mbuf *control, *m0;
684 panic("sbappendcontrol");
685 for (m = control; ; m = m->m_next) {
690 n = m; /* save pointer to last control buffer */
691 for (m = m0; m; m = m->m_next)
693 if (space > sbspace(sb))
695 n->m_next = m0; /* concatenate data to control */
696 for (m = control; m; m = m->m_next)
700 sb->sb_lastrecord->m_nextpkt = control;
703 sb->sb_lastrecord = control;
709 * Compress mbuf chain m into the socket
710 * buffer sb following mbuf n. If n
711 * is null, the buffer is presumed empty.
722 eor |= m->m_flags & M_EOR;
725 (((o = m->m_next) || (o = n)) &&
726 o->m_type == m->m_type))) {
730 if (n && (n->m_flags & M_EOR) == 0 &&
732 m->m_len <= MCLBYTES / 4 && /* XXX: Don't copy too much */
733 m->m_len <= M_TRAILINGSPACE(n) &&
734 n->m_type == m->m_type) {
735 bcopy(mtod(m, caddr_t), mtod(n, caddr_t) + n->m_len,
737 n->m_len += m->m_len;
738 sb->sb_cc += m->m_len;
749 m->m_flags &= ~M_EOR;
757 printf("semi-panic: sbcompress");
762 * Free all mbufs in a sockbuf.
763 * Check that all resources are reclaimed.
770 if (sb->sb_flags & SB_LOCK)
771 panic("sbflush: locked");
772 while (sb->sb_mbcnt) {
774 * Don't call sbdrop(sb, 0) if the leading mbuf is non-empty:
775 * we would loop forever. Panic instead.
777 if (!sb->sb_cc && (sb->sb_mb == NULL || sb->sb_mb->m_len))
779 sbdrop(sb, (int)sb->sb_cc);
781 KASSERT(!(sb->sb_cc || sb->sb_mb || sb->sb_mbcnt || sb->sb_lastmbuf),
782 ("sbflush: cc %ld || mb %p || mbcnt %ld || lastmbuf %p",
783 sb->sb_cc, sb->sb_mb, sb->sb_mbcnt, sb->sb_lastmbuf));
787 * Drop data from (the front of) a sockbuf.
797 next = (m = sb->sb_mb) ? m->m_nextpkt : 0;
806 if (m->m_len > len) {
816 while (m && m->m_len == 0) {
825 sb->sb_lastmbuf = NULL;
830 * Drop a record off the front of a sockbuf
831 * and move the next record to the front.
841 sb->sb_mb = m->m_nextpkt;
850 * Create a "control" mbuf containing the specified data
851 * with the specified type for presentation on a socket buffer.
854 sbcreatecontrol(p, size, type, level)
862 if (CMSG_SPACE((u_int)size) > MCLBYTES)
863 return ((struct mbuf *) NULL);
864 if ((m = m_get(MB_DONTWAIT, MT_CONTROL)) == NULL)
865 return ((struct mbuf *) NULL);
866 if (CMSG_SPACE((u_int)size) > MLEN) {
867 MCLGET(m, MB_DONTWAIT);
868 if ((m->m_flags & M_EXT) == 0) {
870 return ((struct mbuf *) NULL);
873 cp = mtod(m, struct cmsghdr *);
875 KASSERT(CMSG_SPACE((u_int)size) <= M_TRAILINGSPACE(m),
876 ("sbcreatecontrol: short mbuf"));
878 (void)memcpy(CMSG_DATA(cp), p, size);
879 m->m_len = CMSG_SPACE(size);
880 cp->cmsg_len = CMSG_LEN(size);
881 cp->cmsg_level = level;
882 cp->cmsg_type = type;
887 * Some routines that return EOPNOTSUPP for entry points that are not
888 * supported by a protocol. Fill in as needed.
891 pru_accept_notsupp(struct socket *so, struct sockaddr **nam)
897 pru_connect_notsupp(struct socket *so, struct sockaddr *nam, struct thread *td)
903 pru_connect2_notsupp(struct socket *so1, struct socket *so2)
909 pru_control_notsupp(struct socket *so, u_long cmd, caddr_t data,
910 struct ifnet *ifp, struct thread *td)
916 pru_listen_notsupp(struct socket *so, struct thread *td)
922 pru_rcvd_notsupp(struct socket *so, int flags)
928 pru_rcvoob_notsupp(struct socket *so, struct mbuf *m, int flags)
934 * This isn't really a ``null'' operation, but it's the default one
935 * and doesn't do anything destructive.
938 pru_sense_null(struct socket *so, struct stat *sb)
940 sb->st_blksize = so->so_snd.sb_hiwat;
945 * Make a copy of a sockaddr in a malloced buffer of type M_SONAME. Callers
946 * of this routine assume that it always succeeds, so we have to use a
947 * blockable allocation even though we might be called from a critical thread.
950 dup_sockaddr(const struct sockaddr *sa)
952 struct sockaddr *sa2;
954 sa2 = malloc(sa->sa_len, M_SONAME, M_INTWAIT);
955 bcopy(sa, sa2, sa->sa_len);
960 * Create an external-format (``xsocket'') structure using the information
961 * in the kernel-format socket structure pointed to by so. This is done
962 * to reduce the spew of irrelevant information over this interface,
963 * to isolate user code from changes in the kernel structure, and
964 * potentially to provide information-hiding if we decide that
965 * some of this information should be hidden from users.
968 sotoxsocket(struct socket *so, struct xsocket *xso)
970 xso->xso_len = sizeof *xso;
972 xso->so_type = so->so_type;
973 xso->so_options = so->so_options;
974 xso->so_linger = so->so_linger;
975 xso->so_state = so->so_state;
976 xso->so_pcb = so->so_pcb;
977 xso->xso_protocol = so->so_proto->pr_protocol;
978 xso->xso_family = so->so_proto->pr_domain->dom_family;
979 xso->so_qlen = so->so_qlen;
980 xso->so_incqlen = so->so_incqlen;
981 xso->so_qlimit = so->so_qlimit;
982 xso->so_timeo = so->so_timeo;
983 xso->so_error = so->so_error;
984 xso->so_pgid = so->so_sigio ? so->so_sigio->sio_pgid : 0;
985 xso->so_oobmark = so->so_oobmark;
986 sbtoxsockbuf(&so->so_snd, &xso->so_snd);
987 sbtoxsockbuf(&so->so_rcv, &xso->so_rcv);
988 xso->so_uid = so->so_cred->cr_uid;
992 * This does the same for sockbufs. Note that the xsockbuf structure,
993 * since it is always embedded in a socket, does not include a self
994 * pointer nor a length. We make this entry point public in case
995 * some other mechanism needs it.
998 sbtoxsockbuf(struct sockbuf *sb, struct xsockbuf *xsb)
1000 xsb->sb_cc = sb->sb_cc;
1001 xsb->sb_hiwat = sb->sb_hiwat;
1002 xsb->sb_mbcnt = sb->sb_mbcnt;
1003 xsb->sb_mbmax = sb->sb_mbmax;
1004 xsb->sb_lowat = sb->sb_lowat;
1005 xsb->sb_flags = sb->sb_flags;
1006 xsb->sb_timeo = sb->sb_timeo;
1010 * Here is the definition of some of the basic objects in the kern.ipc
1011 * branch of the MIB.
1013 SYSCTL_NODE(_kern, KERN_IPC, ipc, CTLFLAG_RW, 0, "IPC");
1015 /* This takes the place of kern.maxsockbuf, which moved to kern.ipc. */
1017 SYSCTL_INT(_kern, KERN_DUMMY, dummy, CTLFLAG_RW, &dummy, 0, "");
1018 SYSCTL_OID(_kern_ipc, KIPC_MAXSOCKBUF, maxsockbuf, CTLTYPE_INT|CTLFLAG_RW,
1019 &sb_max, 0, sysctl_handle_sb_max, "I", "Maximum socket buffer size");
1020 SYSCTL_INT(_kern_ipc, OID_AUTO, maxsockets, CTLFLAG_RD,
1021 &maxsockets, 0, "Maximum number of sockets avaliable");
1022 SYSCTL_INT(_kern_ipc, KIPC_SOCKBUF_WASTE, sockbuf_waste_factor, CTLFLAG_RW,
1023 &sb_efficiency, 0, "");
1026 * Initialise maxsockets
1028 static void init_maxsockets(void *ignored)
1030 TUNABLE_INT_FETCH("kern.ipc.maxsockets", &maxsockets);
1031 maxsockets = imax(maxsockets, imax(maxfiles, nmbclusters));
1033 SYSINIT(param, SI_SUB_TUNABLES, SI_ORDER_ANY, init_maxsockets, NULL);