2 * Copyright (c) 2005 Jeffrey M. Hsu. All rights reserved.
3 * Copyright (c) 1982, 1986, 1988, 1990, 1993
4 * The Regents of the University of California. All rights reserved.
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions
9 * 1. Redistributions of source code must retain the above copyright
10 * notice, this list of conditions and the following disclaimer.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
14 * 3. Neither the name of the University nor the names of its contributors
15 * may be used to endorse or promote products derived from this software
16 * without specific prior written permission.
18 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
19 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
20 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
21 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
22 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
23 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
24 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
25 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
26 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
27 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30 * @(#)uipc_socket2.c 8.1 (Berkeley) 6/10/93
31 * $FreeBSD: src/sys/kern/uipc_socket2.c,v 1.55.2.17 2002/08/31 19:04:55 dwmalone Exp $
32 * $DragonFly: src/sys/kern/uipc_socket2.c,v 1.33 2008/09/02 16:17:52 dillon Exp $
35 #include "opt_param.h"
36 #include <sys/param.h>
37 #include <sys/systm.h>
38 #include <sys/domain.h>
39 #include <sys/file.h> /* for maxfiles */
40 #include <sys/kernel.h>
43 #include <sys/malloc.h>
45 #include <sys/protosw.h>
46 #include <sys/resourcevar.h>
48 #include <sys/socket.h>
49 #include <sys/socketvar.h>
50 #include <sys/socketops.h>
51 #include <sys/signalvar.h>
52 #include <sys/sysctl.h>
53 #include <sys/event.h>
55 #include <sys/thread2.h>
56 #include <sys/msgport2.h>
57 #include <sys/socketvar2.h>
59 #include <net/netisr2.h>
62 #define KTR_SOWAKEUP KTR_ALL
64 KTR_INFO_MASTER(sowakeup);
65 KTR_INFO(KTR_SOWAKEUP, sowakeup, nconn_start, 0, "newconn sorwakeup start");
66 KTR_INFO(KTR_SOWAKEUP, sowakeup, nconn_end, 1, "newconn sorwakeup end");
67 KTR_INFO(KTR_SOWAKEUP, sowakeup, nconn_wakeupstart, 2, "newconn wakeup start");
68 KTR_INFO(KTR_SOWAKEUP, sowakeup, nconn_wakeupend, 3, "newconn wakeup end");
69 #define logsowakeup(name) KTR_LOG(sowakeup_ ## name)
74 * Primitive routines for operating on sockets and socket buffers
77 u_long sb_max = SB_MAX;
79 SB_MAX * MCLBYTES / (MSIZE + MCLBYTES); /* adjusted sb_max */
81 static u_long sb_efficiency = 8; /* parameter for sbreserve() */
83 /************************************************************************
84 * signalsockbuf procedures *
85 ************************************************************************/
88 * Wait for data to arrive at/drain from a socket buffer.
90 * NOTE: Caller must generally hold the ssb_lock (client side lock) since
91 * WAIT/WAKEUP only works for one client at a time.
93 * NOTE: Caller always retries whatever operation it was waiting on.
96 ssb_wait(struct signalsockbuf *ssb)
102 pflags = (ssb->ssb_flags & SSB_NOINTR) ? 0 : PCATCH;
105 flags = ssb->ssb_flags;
109 * WAKEUP and WAIT interlock each other. We can catch the
110 * race by checking to see if WAKEUP has already been set,
111 * and only setting WAIT if WAKEUP is clear.
113 if (flags & SSB_WAKEUP) {
114 if (atomic_cmpset_int(&ssb->ssb_flags, flags,
115 flags & ~SSB_WAKEUP)) {
123 * Only set WAIT if WAKEUP is clear.
125 tsleep_interlock(&ssb->ssb_cc, pflags);
126 if (atomic_cmpset_int(&ssb->ssb_flags, flags,
128 error = tsleep(&ssb->ssb_cc, pflags | PINTERLOCKED,
129 "sbwait", ssb->ssb_timeo);
137 * Lock a sockbuf already known to be locked;
138 * return any error returned from sleep (EINTR).
141 _ssb_lock(struct signalsockbuf *ssb)
147 pflags = (ssb->ssb_flags & SSB_NOINTR) ? 0 : PCATCH;
150 flags = ssb->ssb_flags;
152 if (flags & SSB_LOCK) {
153 tsleep_interlock(&ssb->ssb_flags, pflags);
154 if (atomic_cmpset_int(&ssb->ssb_flags, flags,
156 error = tsleep(&ssb->ssb_flags,
157 pflags | PINTERLOCKED,
163 if (atomic_cmpset_int(&ssb->ssb_flags, flags,
165 lwkt_gettoken(&ssb->ssb_token);
175 * This does the same for sockbufs. Note that the xsockbuf structure,
176 * since it is always embedded in a socket, does not include a self
177 * pointer nor a length. We make this entry point public in case
178 * some other mechanism needs it.
181 ssbtoxsockbuf(struct signalsockbuf *ssb, struct xsockbuf *xsb)
183 xsb->sb_cc = ssb->ssb_cc;
184 xsb->sb_hiwat = ssb->ssb_hiwat;
185 xsb->sb_mbcnt = ssb->ssb_mbcnt;
186 xsb->sb_mbmax = ssb->ssb_mbmax;
187 xsb->sb_lowat = ssb->ssb_lowat;
188 xsb->sb_flags = ssb->ssb_flags;
189 xsb->sb_timeo = ssb->ssb_timeo;
193 /************************************************************************
194 * Procedures which manipulate socket state flags, wakeups, etc. *
195 ************************************************************************
197 * Normal sequence from the active (originating) side is that
198 * soisconnecting() is called during processing of connect() call, resulting
199 * in an eventual call to soisconnected() if/when the connection is
200 * established. When the connection is torn down soisdisconnecting() is
201 * called during processing of disconnect() call, and soisdisconnected() is
202 * called when the connection to the peer is totally severed.
204 * The semantics of these routines are such that connectionless protocols
205 * can call soisconnected() and soisdisconnected() only, bypassing the
206 * in-progress calls when setting up a ``connection'' takes no time.
208 * From the passive side, a socket is created with two queues of sockets:
209 * so_incomp for connections in progress and so_comp for connections
210 * already made and awaiting user acceptance. As a protocol is preparing
211 * incoming connections, it creates a socket structure queued on so_incomp
212 * by calling sonewconn(). When the connection is established,
213 * soisconnected() is called, and transfers the socket structure to so_comp,
214 * making it available to accept().
216 * If a socket is closed with sockets on either so_incomp or so_comp, these
217 * sockets are dropped.
219 * If higher level protocols are implemented in the kernel, the wakeups
220 * done here will sometimes cause software-interrupt process scheduling.
224 soisconnecting(struct socket *so)
226 soclrstate(so, SS_ISCONNECTED | SS_ISDISCONNECTING);
227 sosetstate(so, SS_ISCONNECTING);
231 soisconnected(struct socket *so)
235 while ((head = so->so_head) != NULL) {
236 lwkt_getpooltoken(head);
237 if (so->so_head == head)
239 lwkt_relpooltoken(head);
242 soclrstate(so, SS_ISCONNECTING | SS_ISDISCONNECTING | SS_ISCONFIRMING);
243 sosetstate(so, SS_ISCONNECTED);
244 if (head && (so->so_state & SS_INCOMP)) {
245 if ((so->so_options & SO_ACCEPTFILTER) != 0) {
246 so->so_upcall = head->so_accf->so_accept_filter->accf_callback;
247 so->so_upcallarg = head->so_accf->so_accept_filter_arg;
248 atomic_set_int(&so->so_rcv.ssb_flags, SSB_UPCALL);
249 so->so_options &= ~SO_ACCEPTFILTER;
250 so->so_upcall(so, so->so_upcallarg, 0);
251 lwkt_relpooltoken(head);
256 * Listen socket are not per-cpu.
258 TAILQ_REMOVE(&head->so_incomp, so, so_list);
260 TAILQ_INSERT_TAIL(&head->so_comp, so, so_list);
262 sosetstate(so, SS_COMP);
263 soclrstate(so, SS_INCOMP);
266 * XXX head may be on a different protocol thread.
267 * sorwakeup()->sowakeup() is hacked atm.
270 wakeup_one(&head->so_timeo);
272 wakeup(&so->so_timeo);
277 lwkt_relpooltoken(head);
281 soisdisconnecting(struct socket *so)
283 soclrstate(so, SS_ISCONNECTING);
284 sosetstate(so, SS_ISDISCONNECTING | SS_CANTRCVMORE | SS_CANTSENDMORE);
285 wakeup((caddr_t)&so->so_timeo);
291 soisdisconnected(struct socket *so)
293 soclrstate(so, SS_ISCONNECTING | SS_ISCONNECTED | SS_ISDISCONNECTING);
294 sosetstate(so, SS_CANTRCVMORE | SS_CANTSENDMORE | SS_ISDISCONNECTED);
295 wakeup((caddr_t)&so->so_timeo);
296 sbdrop(&so->so_snd.sb, so->so_snd.ssb_cc);
302 soisreconnecting(struct socket *so)
304 soclrstate(so, SS_ISDISCONNECTING | SS_ISDISCONNECTED |
305 SS_CANTRCVMORE | SS_CANTSENDMORE);
306 sosetstate(so, SS_ISCONNECTING);
310 soisreconnected(struct socket *so)
312 soclrstate(so, SS_ISDISCONNECTED | SS_CANTRCVMORE | SS_CANTSENDMORE);
317 * Set or change the message port a socket receives commands on.
322 sosetport(struct socket *so, lwkt_port_t port)
328 * When an attempt at a new connection is noted on a socket
329 * which accepts connections, sonewconn is called. If the
330 * connection is possible (subject to space constraints, etc.)
331 * then we allocate a new structure, propoerly linked into the
332 * data structure of the original socket, and return this.
333 * Connstatus may be 0, or SO_ISCONFIRMING, or SO_ISCONNECTED.
335 * The new socket is returned with one ref and so_pcb assigned.
336 * The reference is implied by so_pcb.
339 sonewconn_faddr(struct socket *head, int connstatus,
340 const struct sockaddr *faddr)
344 struct pru_attach_info ai;
346 if (head->so_qlen > 3 * head->so_qlimit / 2)
348 so = soalloc(1, head->so_proto);
353 * Set the port prior to attaching the inpcb to the current
354 * cpu's protocol thread (which should be the current thread
355 * but might not be in all cases). This serializes any pcb ops
356 * which occur to our cpu allowing us to complete the attachment
357 * without racing anything.
359 if (head->so_proto->pr_flags & PR_SYNC_PORT)
360 sosetport(so, &netisr_sync_port);
362 sosetport(so, netisr_cpuport(mycpuid));
363 if ((head->so_options & SO_ACCEPTFILTER) != 0)
366 so->so_type = head->so_type;
367 so->so_options = head->so_options &~ SO_ACCEPTCONN;
368 so->so_linger = head->so_linger;
371 * NOTE: Clearing NOFDREF implies referencing the so with
374 so->so_state = head->so_state | SS_NOFDREF | SS_ASSERTINPROG;
375 so->so_cred = crhold(head->so_cred);
378 ai.fd_rdir = NULL; /* jail code cruft XXX JH */
381 * Reserve space and call pru_attach. We can direct-call the
382 * function since we're already in the protocol thread.
384 if (soreserve(so, head->so_snd.ssb_hiwat,
385 head->so_rcv.ssb_hiwat, NULL) ||
386 so_pru_attach_direct(so, 0, &ai)) {
388 soclrstate(so, SS_ASSERTINPROG);
389 sofree(so); /* remove implied pcb ref */
392 KKASSERT(((so->so_proto->pr_flags & PR_ASYNC_RCVD) == 0 &&
393 so->so_refs == 2) || /* attach + our base ref */
394 ((so->so_proto->pr_flags & PR_ASYNC_RCVD) &&
395 so->so_refs == 3)); /* + async rcvd ref */
397 KKASSERT(so->so_port != NULL);
398 so->so_rcv.ssb_lowat = head->so_rcv.ssb_lowat;
399 so->so_snd.ssb_lowat = head->so_snd.ssb_lowat;
400 so->so_rcv.ssb_timeo = head->so_rcv.ssb_timeo;
401 so->so_snd.ssb_timeo = head->so_snd.ssb_timeo;
403 if (head->so_rcv.ssb_flags & SSB_AUTOLOWAT)
404 so->so_rcv.ssb_flags |= SSB_AUTOLOWAT;
406 so->so_rcv.ssb_flags &= ~SSB_AUTOLOWAT;
408 if (head->so_snd.ssb_flags & SSB_AUTOLOWAT)
409 so->so_snd.ssb_flags |= SSB_AUTOLOWAT;
411 so->so_snd.ssb_flags &= ~SSB_AUTOLOWAT;
413 if (head->so_rcv.ssb_flags & SSB_AUTOSIZE)
414 so->so_rcv.ssb_flags |= SSB_AUTOSIZE;
416 so->so_rcv.ssb_flags &= ~SSB_AUTOSIZE;
418 if (head->so_snd.ssb_flags & SSB_AUTOSIZE)
419 so->so_snd.ssb_flags |= SSB_AUTOSIZE;
421 so->so_snd.ssb_flags &= ~SSB_AUTOSIZE;
424 * Save the faddr, if the information is provided and
425 * the protocol can perform the saving opertation.
427 if (faddr != NULL && so->so_proto->pr_usrreqs->pru_savefaddr != NULL)
428 so->so_proto->pr_usrreqs->pru_savefaddr(so, faddr);
430 lwkt_getpooltoken(head);
432 TAILQ_INSERT_TAIL(&head->so_comp, so, so_list);
433 sosetstate(so, SS_COMP);
436 if (head->so_incqlen > head->so_qlimit) {
437 sp = TAILQ_FIRST(&head->so_incomp);
438 TAILQ_REMOVE(&head->so_incomp, sp, so_list);
440 soclrstate(sp, SS_INCOMP);
444 TAILQ_INSERT_TAIL(&head->so_incomp, so, so_list);
445 sosetstate(so, SS_INCOMP);
448 lwkt_relpooltoken(head);
451 * XXX head may be on a different protocol thread.
452 * sorwakeup()->sowakeup() is hacked atm.
454 logsowakeup(nconn_start);
456 logsowakeup(nconn_end);
458 logsowakeup(nconn_wakeupstart);
459 wakeup((caddr_t)&head->so_timeo);
460 logsowakeup(nconn_wakeupend);
462 sosetstate(so, connstatus);
464 soclrstate(so, SS_ASSERTINPROG);
469 sonewconn(struct socket *head, int connstatus)
471 return sonewconn_faddr(head, connstatus, NULL);
475 * Socantsendmore indicates that no more data will be sent on the
476 * socket; it would normally be applied to a socket when the user
477 * informs the system that no more data is to be sent, by the protocol
478 * code (in case PRU_SHUTDOWN). Socantrcvmore indicates that no more data
479 * will be received, and will normally be applied to the socket by a
480 * protocol when it detects that the peer will send no more data.
481 * Data queued for reading in the socket may yet be read.
484 socantsendmore(struct socket *so)
486 sosetstate(so, SS_CANTSENDMORE);
491 socantrcvmore(struct socket *so)
493 sosetstate(so, SS_CANTRCVMORE);
498 * Wakeup processes waiting on a socket buffer. Do asynchronous notification
499 * via SIGIO if the socket has the SS_ASYNC flag set.
501 * For users waiting on send/recv try to avoid unnecessary context switch
502 * thrashing. Particularly for senders of large buffers (needs to be
503 * extended to sel and aio? XXX)
505 * WARNING! Can be called on a foreign socket from the wrong protocol
506 * thread. aka is called on the 'head' listen socket when
507 * a new connection comes in.
511 sowakeup(struct socket *so, struct signalsockbuf *ssb)
513 struct kqinfo *kqinfo = &ssb->ssb_kq;
517 * Atomically check the flags. When no special features are being
518 * used, WAIT is clear, and WAKEUP is already set, we can simply
519 * return. The upcoming synchronous waiter will not block.
521 flags = atomic_fetchadd_int(&ssb->ssb_flags, 0);
522 if ((flags & SSB_NOTIFY_MASK) == 0) {
523 if (flags & SSB_WAKEUP)
528 * Check conditions, set the WAKEUP flag, and clear and signal if
529 * the WAIT flag is found to be set. This interlocks against the
535 flags = ssb->ssb_flags;
537 if (ssb->ssb_flags & SSB_PREALLOC)
538 space = ssb_space_prealloc(ssb);
540 space = ssb_space(ssb);
542 if ((ssb == &so->so_snd && space >= ssb->ssb_lowat) ||
543 (ssb == &so->so_rcv && ssb->ssb_cc >= ssb->ssb_lowat) ||
544 (ssb == &so->so_snd && (so->so_state & SS_CANTSENDMORE)) ||
545 (ssb == &so->so_rcv && (so->so_state & SS_CANTRCVMORE))
547 if (atomic_cmpset_int(&ssb->ssb_flags, flags,
548 (flags | SSB_WAKEUP) & ~SSB_WAIT)) {
549 if (flags & SSB_WAIT)
550 wakeup(&ssb->ssb_cc);
561 if ((so->so_state & SS_ASYNC) && so->so_sigio != NULL)
562 pgsigio(so->so_sigio, SIGIO, 0);
563 if (ssb->ssb_flags & SSB_UPCALL)
564 (*so->so_upcall)(so, so->so_upcallarg, MB_DONTWAIT);
565 KNOTE(&kqinfo->ki_note, 0);
568 * This is a bit of a hack. Multiple threads can wind up scanning
569 * ki_mlist concurrently due to the fact that this function can be
570 * called on a foreign socket, so we can't afford to block here.
572 * We need the pool token for (so) (likely the listne socket if
573 * SSB_MEVENT is set) because the predicate function may have
574 * to access the accept queue.
576 if (ssb->ssb_flags & SSB_MEVENT) {
577 struct netmsg_so_notify *msg, *nmsg;
579 lwkt_getpooltoken(so);
580 TAILQ_FOREACH_MUTABLE(msg, &kqinfo->ki_mlist, nm_list, nmsg) {
581 if (msg->nm_predicate(msg)) {
582 TAILQ_REMOVE(&kqinfo->ki_mlist, msg, nm_list);
583 lwkt_replymsg(&msg->base.lmsg,
584 msg->base.lmsg.ms_error);
587 if (TAILQ_EMPTY(&ssb->ssb_kq.ki_mlist))
588 atomic_clear_int(&ssb->ssb_flags, SSB_MEVENT);
589 lwkt_relpooltoken(so);
594 * Socket buffer (struct signalsockbuf) utility routines.
596 * Each socket contains two socket buffers: one for sending data and
597 * one for receiving data. Each buffer contains a queue of mbufs,
598 * information about the number of mbufs and amount of data in the
599 * queue, and other fields allowing kevent()/select()/poll() statements
600 * and notification on data availability to be implemented.
602 * Data stored in a socket buffer is maintained as a list of records.
603 * Each record is a list of mbufs chained together with the m_next
604 * field. Records are chained together with the m_nextpkt field. The upper
605 * level routine soreceive() expects the following conventions to be
606 * observed when placing information in the receive buffer:
608 * 1. If the protocol requires each message be preceded by the sender's
609 * name, then a record containing that name must be present before
610 * any associated data (mbuf's must be of type MT_SONAME).
611 * 2. If the protocol supports the exchange of ``access rights'' (really
612 * just additional data associated with the message), and there are
613 * ``rights'' to be received, then a record containing this data
614 * should be present (mbuf's must be of type MT_RIGHTS).
615 * 3. If a name or rights record exists, then it must be followed by
616 * a data record, perhaps of zero length.
618 * Before using a new socket structure it is first necessary to reserve
619 * buffer space to the socket, by calling sbreserve(). This should commit
620 * some of the available buffer space in the system buffer pool for the
621 * socket (currently, it does nothing but enforce limits). The space
622 * should be released by calling ssb_release() when the socket is destroyed.
625 soreserve(struct socket *so, u_long sndcc, u_long rcvcc, struct rlimit *rl)
627 if (so->so_snd.ssb_lowat == 0)
628 atomic_set_int(&so->so_snd.ssb_flags, SSB_AUTOLOWAT);
629 if (ssb_reserve(&so->so_snd, sndcc, so, rl) == 0)
631 if (ssb_reserve(&so->so_rcv, rcvcc, so, rl) == 0)
633 if (so->so_rcv.ssb_lowat == 0)
634 so->so_rcv.ssb_lowat = 1;
635 if (so->so_snd.ssb_lowat == 0)
636 so->so_snd.ssb_lowat = MCLBYTES;
637 if (so->so_snd.ssb_lowat > so->so_snd.ssb_hiwat)
638 so->so_snd.ssb_lowat = so->so_snd.ssb_hiwat;
641 ssb_release(&so->so_snd, so);
647 sysctl_handle_sb_max(SYSCTL_HANDLER_ARGS)
650 u_long old_sb_max = sb_max;
652 error = SYSCTL_OUT(req, arg1, sizeof(int));
653 if (error || !req->newptr)
655 error = SYSCTL_IN(req, arg1, sizeof(int));
658 if (sb_max < MSIZE + MCLBYTES) {
662 sb_max_adj = (u_quad_t)sb_max * MCLBYTES / (MSIZE + MCLBYTES);
667 * Allot mbufs to a signalsockbuf.
669 * Attempt to scale mbmax so that mbcnt doesn't become limiting
670 * if buffering efficiency is near the normal case.
672 * sb_max only applies to user-sockets (where rl != NULL). It does
673 * not apply to kernel sockets or kernel-controlled sockets. Note
674 * that NFS overrides the sockbuf limits created when nfsd creates
678 ssb_reserve(struct signalsockbuf *ssb, u_long cc, struct socket *so,
682 * rl will only be NULL when we're in an interrupt (eg, in tcp_input)
683 * or when called from netgraph (ie, ngd_attach)
685 if (rl && cc > sb_max_adj)
687 if (!chgsbsize(so->so_cred->cr_uidinfo, &ssb->ssb_hiwat, cc,
688 rl ? rl->rlim_cur : RLIM_INFINITY)) {
692 ssb->ssb_mbmax = min(cc * sb_efficiency, sb_max);
694 ssb->ssb_mbmax = cc * sb_efficiency;
697 * AUTOLOWAT is set on send buffers and prevents large writes
698 * from generating a huge number of context switches.
700 if (ssb->ssb_flags & SSB_AUTOLOWAT) {
701 ssb->ssb_lowat = ssb->ssb_hiwat / 4;
702 if (ssb->ssb_lowat < MCLBYTES)
703 ssb->ssb_lowat = MCLBYTES;
705 if (ssb->ssb_lowat > ssb->ssb_hiwat)
706 ssb->ssb_lowat = ssb->ssb_hiwat;
711 * Free mbufs held by a socket, and reserved mbuf space.
714 ssb_release(struct signalsockbuf *ssb, struct socket *so)
717 (void)chgsbsize(so->so_cred->cr_uidinfo, &ssb->ssb_hiwat, 0,
723 * Some routines that return EOPNOTSUPP for entry points that are not
724 * supported by a protocol. Fill in as needed.
727 pr_generic_notsupp(netmsg_t msg)
729 lwkt_replymsg(&msg->lmsg, EOPNOTSUPP);
733 pru_sosend_notsupp(struct socket *so, struct sockaddr *addr, struct uio *uio,
734 struct mbuf *top, struct mbuf *control, int flags,
745 pru_soreceive_notsupp(struct socket *so, struct sockaddr **paddr,
746 struct uio *uio, struct sockbuf *sio,
747 struct mbuf **controlp, int *flagsp)
753 * This isn't really a ``null'' operation, but it's the default one
754 * and doesn't do anything destructive.
757 pru_sense_null(netmsg_t msg)
759 msg->sense.nm_stat->st_blksize = msg->base.nm_so->so_snd.ssb_hiwat;
760 lwkt_replymsg(&msg->lmsg, 0);
764 * Make a copy of a sockaddr in a malloced buffer of type M_SONAME. Callers
765 * of this routine assume that it always succeeds, so we have to use a
766 * blockable allocation even though we might be called from a critical thread.
769 dup_sockaddr(const struct sockaddr *sa)
771 struct sockaddr *sa2;
773 sa2 = kmalloc(sa->sa_len, M_SONAME, M_INTWAIT);
774 bcopy(sa, sa2, sa->sa_len);
779 * Create an external-format (``xsocket'') structure using the information
780 * in the kernel-format socket structure pointed to by so. This is done
781 * to reduce the spew of irrelevant information over this interface,
782 * to isolate user code from changes in the kernel structure, and
783 * potentially to provide information-hiding if we decide that
784 * some of this information should be hidden from users.
787 sotoxsocket(struct socket *so, struct xsocket *xso)
789 xso->xso_len = sizeof *xso;
791 xso->so_type = so->so_type;
792 xso->so_options = so->so_options;
793 xso->so_linger = so->so_linger;
794 xso->so_state = so->so_state;
795 xso->so_pcb = so->so_pcb;
796 xso->xso_protocol = so->so_proto->pr_protocol;
797 xso->xso_family = so->so_proto->pr_domain->dom_family;
798 xso->so_qlen = so->so_qlen;
799 xso->so_incqlen = so->so_incqlen;
800 xso->so_qlimit = so->so_qlimit;
801 xso->so_timeo = so->so_timeo;
802 xso->so_error = so->so_error;
803 xso->so_pgid = so->so_sigio ? so->so_sigio->sio_pgid : 0;
804 xso->so_oobmark = so->so_oobmark;
805 ssbtoxsockbuf(&so->so_snd, &xso->so_snd);
806 ssbtoxsockbuf(&so->so_rcv, &xso->so_rcv);
807 xso->so_uid = so->so_cred->cr_uid;
811 * Here is the definition of some of the basic objects in the kern.ipc
814 SYSCTL_NODE(_kern, KERN_IPC, ipc, CTLFLAG_RW, 0, "IPC");
817 * This takes the place of kern.maxsockbuf, which moved to kern.ipc.
819 * NOTE! sb_max only applies to user-created socket buffers.
822 SYSCTL_INT(_kern, KERN_DUMMY, dummy, CTLFLAG_RW, &dummy, 0, "");
823 SYSCTL_OID(_kern_ipc, KIPC_MAXSOCKBUF, maxsockbuf, CTLTYPE_INT|CTLFLAG_RW,
824 &sb_max, 0, sysctl_handle_sb_max, "I", "Maximum socket buffer size");
825 SYSCTL_INT(_kern_ipc, OID_AUTO, maxsockets, CTLFLAG_RD,
826 &maxsockets, 0, "Maximum number of sockets available");
827 SYSCTL_INT(_kern_ipc, KIPC_SOCKBUF_WASTE, sockbuf_waste_factor, CTLFLAG_RW,
829 "Socket buffer limit scaler");
832 * Initialize maxsockets
835 init_maxsockets(void *ignored)
837 TUNABLE_INT_FETCH("kern.ipc.maxsockets", &maxsockets);
838 maxsockets = imax(maxsockets, imax(maxfiles, nmbclusters));
840 SYSINIT(param, SI_BOOT1_TUNABLES, SI_ORDER_ANY,
841 init_maxsockets, NULL);