xref: /dflybsd-src/sys/kern/uipc_socket2.c (revision c3c96e4421a1087a390825eac6c01c9ed9182387)
1 /*
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.
5  *
6  * Redistribution and use in source and binary forms, with or without
7  * modification, are permitted provided that the following conditions
8  * are met:
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. All advertising materials mentioning features or use of this software
15  *    must display the following acknowledgement:
16  *	This product includes software developed by the University of
17  *	California, Berkeley and its contributors.
18  * 4. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  *	@(#)uipc_socket2.c	8.1 (Berkeley) 6/10/93
35  * $FreeBSD: src/sys/kern/uipc_socket2.c,v 1.55.2.17 2002/08/31 19:04:55 dwmalone Exp $
36  * $DragonFly: src/sys/kern/uipc_socket2.c,v 1.33 2008/09/02 16:17:52 dillon Exp $
37  */
38 
39 #include "opt_param.h"
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/domain.h>
43 #include <sys/file.h>	/* for maxfiles */
44 #include <sys/kernel.h>
45 #include <sys/proc.h>
46 #include <sys/malloc.h>
47 #include <sys/mbuf.h>
48 #include <sys/protosw.h>
49 #include <sys/resourcevar.h>
50 #include <sys/stat.h>
51 #include <sys/socket.h>
52 #include <sys/socketvar.h>
53 #include <sys/signalvar.h>
54 #include <sys/sysctl.h>
55 #include <sys/aio.h> /* for aio_swake proto */
56 #include <sys/event.h>
57 
58 #include <sys/thread2.h>
59 #include <sys/msgport2.h>
60 
61 int	maxsockets;
62 
63 /*
64  * Primitive routines for operating on sockets and socket buffers
65  */
66 
67 u_long	sb_max = SB_MAX;
68 u_long	sb_max_adj =
69     SB_MAX * MCLBYTES / (MSIZE + MCLBYTES); /* adjusted sb_max */
70 
71 static	u_long sb_efficiency = 8;	/* parameter for sbreserve() */
72 
73 /************************************************************************
74  * signalsockbuf procedures						*
75  ************************************************************************/
76 
77 /*
78  * Wait for data to arrive at/drain from a socket buffer.
79  *
80  * NOTE: Caller must generally hold the ssb_lock (client side lock) since
81  *	 WAIT/WAKEUP only works for one client at a time.
82  *
83  * NOTE: Caller always retries whatever operation it was waiting on.
84  */
85 int
86 ssb_wait(struct signalsockbuf *ssb)
87 {
88 	uint32_t flags;
89 	int pflags;
90 	int error;
91 
92 	pflags = (ssb->ssb_flags & SSB_NOINTR) ? 0 : PCATCH;
93 
94 	for (;;) {
95 		flags = ssb->ssb_flags;
96 		cpu_ccfence();
97 
98 		/*
99 		 * WAKEUP and WAIT interlock eachother.  We can catch the
100 		 * race by checking to see if WAKEUP has already been set,
101 		 * and only setting WAIT if WAKEUP is clear.
102 		 */
103 		if (flags & SSB_WAKEUP) {
104 			if (atomic_cmpset_int(&ssb->ssb_flags, flags,
105 					      flags & ~SSB_WAKEUP)) {
106 				error = 0;
107 				break;
108 			}
109 			continue;
110 		}
111 
112 		/*
113 		 * Only set WAIT if WAKEUP is clear.
114 		 */
115 		tsleep_interlock(&ssb->ssb_cc, pflags);
116 		if (atomic_cmpset_int(&ssb->ssb_flags, flags,
117 				      flags | SSB_WAIT)) {
118 			error = tsleep(&ssb->ssb_cc, pflags | PINTERLOCKED,
119 				       "sbwait", ssb->ssb_timeo);
120 			break;
121 		}
122 	}
123 	return (error);
124 }
125 
126 /*
127  * Lock a sockbuf already known to be locked;
128  * return any error returned from sleep (EINTR).
129  */
130 int
131 _ssb_lock(struct signalsockbuf *ssb)
132 {
133 	uint32_t flags;
134 	int pflags;
135 	int error;
136 
137 	pflags = (ssb->ssb_flags & SSB_NOINTR) ? 0 : PCATCH;
138 
139 	for (;;) {
140 		flags = ssb->ssb_flags;
141 		cpu_ccfence();
142 		if (flags & SSB_LOCK) {
143 			tsleep_interlock(&ssb->ssb_flags, pflags);
144 			if (atomic_cmpset_int(&ssb->ssb_flags, flags,
145 					      flags | SSB_WANT)) {
146 				error = tsleep(&ssb->ssb_flags,
147 					       pflags | PINTERLOCKED,
148 					       "sblock", 0);
149 				if (error)
150 					break;
151 			}
152 		} else {
153 			if (atomic_cmpset_int(&ssb->ssb_flags, flags,
154 					      flags | SSB_LOCK)) {
155 				error = 0;
156 				break;
157 			}
158 		}
159 	}
160 	return (error);
161 }
162 
163 /*
164  * This does the same for sockbufs.  Note that the xsockbuf structure,
165  * since it is always embedded in a socket, does not include a self
166  * pointer nor a length.  We make this entry point public in case
167  * some other mechanism needs it.
168  */
169 void
170 ssbtoxsockbuf(struct signalsockbuf *ssb, struct xsockbuf *xsb)
171 {
172 	xsb->sb_cc = ssb->ssb_cc;
173 	xsb->sb_hiwat = ssb->ssb_hiwat;
174 	xsb->sb_mbcnt = ssb->ssb_mbcnt;
175 	xsb->sb_mbmax = ssb->ssb_mbmax;
176 	xsb->sb_lowat = ssb->ssb_lowat;
177 	xsb->sb_flags = ssb->ssb_flags;
178 	xsb->sb_timeo = ssb->ssb_timeo;
179 }
180 
181 
182 /************************************************************************
183  * Procedures which manipulate socket state flags, wakeups, etc.	*
184  ************************************************************************
185  *
186  * Normal sequence from the active (originating) side is that
187  * soisconnecting() is called during processing of connect() call, resulting
188  * in an eventual call to soisconnected() if/when the connection is
189  * established.  When the connection is torn down soisdisconnecting() is
190  * called during processing of disconnect() call, and soisdisconnected() is
191  * called when the connection to the peer is totally severed.
192  *
193  * The semantics of these routines are such that connectionless protocols
194  * can call soisconnected() and soisdisconnected() only, bypassing the
195  * in-progress calls when setting up a ``connection'' takes no time.
196  *
197  * From the passive side, a socket is created with two queues of sockets:
198  * so_incomp for connections in progress and so_comp for connections
199  * already made and awaiting user acceptance.  As a protocol is preparing
200  * incoming connections, it creates a socket structure queued on so_incomp
201  * by calling sonewconn().  When the connection is established,
202  * soisconnected() is called, and transfers the socket structure to so_comp,
203  * making it available to accept().
204  *
205  * If a socket is closed with sockets on either so_incomp or so_comp, these
206  * sockets are dropped.
207  *
208  * If higher level protocols are implemented in the kernel, the wakeups
209  * done here will sometimes cause software-interrupt process scheduling.
210  */
211 
212 void
213 soisconnecting(struct socket *so)
214 {
215 	so->so_state &= ~(SS_ISCONNECTED|SS_ISDISCONNECTING);
216 	so->so_state |= SS_ISCONNECTING;
217 }
218 
219 void
220 soisconnected(struct socket *so)
221 {
222 	struct socket *head = so->so_head;
223 
224 	so->so_state &= ~(SS_ISCONNECTING|SS_ISDISCONNECTING|SS_ISCONFIRMING);
225 	so->so_state |= SS_ISCONNECTED;
226 	if (head && (so->so_state & SS_INCOMP)) {
227 		if ((so->so_options & SO_ACCEPTFILTER) != 0) {
228 			so->so_upcall = head->so_accf->so_accept_filter->accf_callback;
229 			so->so_upcallarg = head->so_accf->so_accept_filter_arg;
230 			atomic_set_int(&so->so_rcv.ssb_flags, SSB_UPCALL);
231 			so->so_options &= ~SO_ACCEPTFILTER;
232 			so->so_upcall(so, so->so_upcallarg, 0);
233 			return;
234 		}
235 		TAILQ_REMOVE(&head->so_incomp, so, so_list);
236 		head->so_incqlen--;
237 		so->so_state &= ~SS_INCOMP;
238 		TAILQ_INSERT_TAIL(&head->so_comp, so, so_list);
239 		head->so_qlen++;
240 		so->so_state |= SS_COMP;
241 		sorwakeup(head);
242 		wakeup_one(&head->so_timeo);
243 	} else {
244 		wakeup(&so->so_timeo);
245 		sorwakeup(so);
246 		sowwakeup(so);
247 	}
248 }
249 
250 void
251 soisdisconnecting(struct socket *so)
252 {
253 	so->so_state &= ~SS_ISCONNECTING;
254 	so->so_state |= (SS_ISDISCONNECTING|SS_CANTRCVMORE|SS_CANTSENDMORE);
255 	wakeup((caddr_t)&so->so_timeo);
256 	sowwakeup(so);
257 	sorwakeup(so);
258 }
259 
260 void
261 soisdisconnected(struct socket *so)
262 {
263 	so->so_state &= ~(SS_ISCONNECTING|SS_ISCONNECTED|SS_ISDISCONNECTING);
264 	so->so_state |= (SS_CANTRCVMORE|SS_CANTSENDMORE|SS_ISDISCONNECTED);
265 	wakeup((caddr_t)&so->so_timeo);
266 	sbdrop(&so->so_snd.sb, so->so_snd.ssb_cc);
267 	sowwakeup(so);
268 	sorwakeup(so);
269 }
270 
271 void
272 soisreconnecting(struct socket *so)
273 {
274         so->so_state &= ~(SS_ISDISCONNECTING|SS_ISDISCONNECTED|SS_CANTRCVMORE|
275 			SS_CANTSENDMORE);
276 	so->so_state |= SS_ISCONNECTING;
277 }
278 
279 void
280 soisreconnected(struct socket *so)
281 {
282 	so->so_state &= ~(SS_ISDISCONNECTED|SS_CANTRCVMORE|SS_CANTSENDMORE);
283 	soisconnected(so);
284 }
285 
286 /*
287  * Set or change the message port a socket receives commands on.
288  *
289  * XXX
290  */
291 void
292 sosetport(struct socket *so, lwkt_port_t port)
293 {
294 	so->so_port = port;
295 }
296 
297 /*
298  * When an attempt at a new connection is noted on a socket
299  * which accepts connections, sonewconn is called.  If the
300  * connection is possible (subject to space constraints, etc.)
301  * then we allocate a new structure, propoerly linked into the
302  * data structure of the original socket, and return this.
303  * Connstatus may be 0, or SO_ISCONFIRMING, or SO_ISCONNECTED.
304  */
305 struct socket *
306 sonewconn(struct socket *head, int connstatus)
307 {
308 	struct socket *so;
309 	struct socket *sp;
310 	struct pru_attach_info ai;
311 
312 	if (head->so_qlen > 3 * head->so_qlimit / 2)
313 		return (NULL);
314 	so = soalloc(1);
315 	if (so == NULL)
316 		return (NULL);
317 	if ((head->so_options & SO_ACCEPTFILTER) != 0)
318 		connstatus = 0;
319 	so->so_head = head;
320 	so->so_type = head->so_type;
321 	so->so_options = head->so_options &~ SO_ACCEPTCONN;
322 	so->so_linger = head->so_linger;
323 	so->so_state = head->so_state | SS_NOFDREF;
324 	so->so_proto = head->so_proto;
325 	so->so_cred = crhold(head->so_cred);
326 	ai.sb_rlimit = NULL;
327 	ai.p_ucred = NULL;
328 	ai.fd_rdir = NULL;		/* jail code cruft XXX JH */
329 	if (soreserve(so, head->so_snd.ssb_hiwat, head->so_rcv.ssb_hiwat, NULL) ||
330 	    /* Directly call function since we're already at protocol level. */
331 	    (*so->so_proto->pr_usrreqs->pru_attach)(so, 0, &ai)) {
332 		sodealloc(so);
333 		return (NULL);
334 	}
335 	KKASSERT(so->so_port != NULL);
336 	so->so_rcv.ssb_lowat = head->so_rcv.ssb_lowat;
337 	so->so_snd.ssb_lowat = head->so_snd.ssb_lowat;
338 	so->so_rcv.ssb_timeo = head->so_rcv.ssb_timeo;
339 	so->so_snd.ssb_timeo = head->so_snd.ssb_timeo;
340 	so->so_rcv.ssb_flags |= head->so_rcv.ssb_flags &
341 				(SSB_AUTOSIZE | SSB_AUTOLOWAT);
342 	so->so_snd.ssb_flags |= head->so_snd.ssb_flags &
343 				(SSB_AUTOSIZE | SSB_AUTOLOWAT);
344 	if (connstatus) {
345 		TAILQ_INSERT_TAIL(&head->so_comp, so, so_list);
346 		so->so_state |= SS_COMP;
347 		head->so_qlen++;
348 	} else {
349 		if (head->so_incqlen > head->so_qlimit) {
350 			sp = TAILQ_FIRST(&head->so_incomp);
351 			TAILQ_REMOVE(&head->so_incomp, sp, so_list);
352 			head->so_incqlen--;
353 			sp->so_state &= ~SS_INCOMP;
354 			sp->so_head = NULL;
355 			soaborta(sp);
356 		}
357 		TAILQ_INSERT_TAIL(&head->so_incomp, so, so_list);
358 		so->so_state |= SS_INCOMP;
359 		head->so_incqlen++;
360 	}
361 	if (connstatus) {
362 		sorwakeup(head);
363 		wakeup((caddr_t)&head->so_timeo);
364 		so->so_state |= connstatus;
365 	}
366 	return (so);
367 }
368 
369 /*
370  * Socantsendmore indicates that no more data will be sent on the
371  * socket; it would normally be applied to a socket when the user
372  * informs the system that no more data is to be sent, by the protocol
373  * code (in case PRU_SHUTDOWN).  Socantrcvmore indicates that no more data
374  * will be received, and will normally be applied to the socket by a
375  * protocol when it detects that the peer will send no more data.
376  * Data queued for reading in the socket may yet be read.
377  */
378 void
379 socantsendmore(struct socket *so)
380 {
381 	so->so_state |= SS_CANTSENDMORE;
382 	sowwakeup(so);
383 }
384 
385 void
386 socantrcvmore(struct socket *so)
387 {
388 	so->so_state |= SS_CANTRCVMORE;
389 	sorwakeup(so);
390 }
391 
392 /*
393  * Wakeup processes waiting on a socket buffer.  Do asynchronous notification
394  * via SIGIO if the socket has the SS_ASYNC flag set.
395  *
396  * For users waiting on send/recv try to avoid unnecessary context switch
397  * thrashing.  Particularly for senders of large buffers (needs to be
398  * extended to sel and aio? XXX)
399  */
400 void
401 sowakeup(struct socket *so, struct signalsockbuf *ssb)
402 {
403 	struct kqinfo *kqinfo = &ssb->ssb_kq;
404 	uint32_t flags;
405 
406 	/*
407 	 * Check conditions, set the WAKEUP flag, and clear and signal if
408 	 * the WAIT flag is found to be set.  This interlocks against the
409 	 * client side.
410 	 */
411 	for (;;) {
412 		flags = ssb->ssb_flags;
413 		cpu_ccfence();
414 
415 		if ((ssb == &so->so_snd && ssb_space(ssb) >= ssb->ssb_lowat) ||
416 		    (ssb == &so->so_rcv && ssb->ssb_cc >= ssb->ssb_lowat) ||
417 		    (ssb == &so->so_snd && (so->so_state & SS_CANTSENDMORE)) ||
418 		    (ssb == &so->so_rcv && (so->so_state & SS_CANTRCVMORE))
419 		) {
420 			if (atomic_cmpset_int(&ssb->ssb_flags, flags,
421 					  (flags | SSB_WAKEUP) & ~SSB_WAIT)) {
422 				if (flags & SSB_WAIT)
423 					wakeup(&ssb->ssb_cc);
424 				break;
425 			}
426 		} else {
427 			break;
428 		}
429 	}
430 
431 	/*
432 	 * Misc other events
433 	 */
434 	if ((so->so_state & SS_ASYNC) && so->so_sigio != NULL)
435 		pgsigio(so->so_sigio, SIGIO, 0);
436 	if (ssb->ssb_flags & SSB_UPCALL)
437 		(*so->so_upcall)(so, so->so_upcallarg, MB_DONTWAIT);
438 	if (ssb->ssb_flags & SSB_AIO)
439 		aio_swake(so, ssb);
440 	KNOTE(&kqinfo->ki_note, 0);
441 	if (ssb->ssb_flags & SSB_MEVENT) {
442 		struct netmsg_so_notify *msg, *nmsg;
443 
444 		TAILQ_FOREACH_MUTABLE(msg, &kqinfo->ki_mlist, nm_list, nmsg) {
445 			if (msg->nm_predicate(&msg->nm_netmsg)) {
446 				TAILQ_REMOVE(&kqinfo->ki_mlist, msg, nm_list);
447 				lwkt_replymsg(&msg->nm_netmsg.nm_lmsg,
448 					      msg->nm_netmsg.nm_lmsg.ms_error);
449 			}
450 		}
451 		if (TAILQ_EMPTY(&ssb->ssb_kq.ki_mlist))
452 			atomic_clear_int(&ssb->ssb_flags, SSB_MEVENT);
453 	}
454 }
455 
456 /*
457  * Socket buffer (struct signalsockbuf) utility routines.
458  *
459  * Each socket contains two socket buffers: one for sending data and
460  * one for receiving data.  Each buffer contains a queue of mbufs,
461  * information about the number of mbufs and amount of data in the
462  * queue, and other fields allowing kevent()/select()/poll() statements
463  * and notification on data availability to be implemented.
464  *
465  * Data stored in a socket buffer is maintained as a list of records.
466  * Each record is a list of mbufs chained together with the m_next
467  * field.  Records are chained together with the m_nextpkt field. The upper
468  * level routine soreceive() expects the following conventions to be
469  * observed when placing information in the receive buffer:
470  *
471  * 1. If the protocol requires each message be preceded by the sender's
472  *    name, then a record containing that name must be present before
473  *    any associated data (mbuf's must be of type MT_SONAME).
474  * 2. If the protocol supports the exchange of ``access rights'' (really
475  *    just additional data associated with the message), and there are
476  *    ``rights'' to be received, then a record containing this data
477  *    should be present (mbuf's must be of type MT_RIGHTS).
478  * 3. If a name or rights record exists, then it must be followed by
479  *    a data record, perhaps of zero length.
480  *
481  * Before using a new socket structure it is first necessary to reserve
482  * buffer space to the socket, by calling sbreserve().  This should commit
483  * some of the available buffer space in the system buffer pool for the
484  * socket (currently, it does nothing but enforce limits).  The space
485  * should be released by calling ssb_release() when the socket is destroyed.
486  */
487 int
488 soreserve(struct socket *so, u_long sndcc, u_long rcvcc, struct rlimit *rl)
489 {
490 	if (so->so_snd.ssb_lowat == 0)
491 		atomic_set_int(&so->so_snd.ssb_flags, SSB_AUTOLOWAT);
492 	if (ssb_reserve(&so->so_snd, sndcc, so, rl) == 0)
493 		goto bad;
494 	if (ssb_reserve(&so->so_rcv, rcvcc, so, rl) == 0)
495 		goto bad2;
496 	if (so->so_rcv.ssb_lowat == 0)
497 		so->so_rcv.ssb_lowat = 1;
498 	if (so->so_snd.ssb_lowat == 0)
499 		so->so_snd.ssb_lowat = MCLBYTES;
500 	if (so->so_snd.ssb_lowat > so->so_snd.ssb_hiwat)
501 		so->so_snd.ssb_lowat = so->so_snd.ssb_hiwat;
502 	return (0);
503 bad2:
504 	ssb_release(&so->so_snd, so);
505 bad:
506 	return (ENOBUFS);
507 }
508 
509 static int
510 sysctl_handle_sb_max(SYSCTL_HANDLER_ARGS)
511 {
512 	int error = 0;
513 	u_long old_sb_max = sb_max;
514 
515 	error = SYSCTL_OUT(req, arg1, sizeof(int));
516 	if (error || !req->newptr)
517 		return (error);
518 	error = SYSCTL_IN(req, arg1, sizeof(int));
519 	if (error)
520 		return (error);
521 	if (sb_max < MSIZE + MCLBYTES) {
522 		sb_max = old_sb_max;
523 		return (EINVAL);
524 	}
525 	sb_max_adj = (u_quad_t)sb_max * MCLBYTES / (MSIZE + MCLBYTES);
526 	return (0);
527 }
528 
529 /*
530  * Allot mbufs to a signalsockbuf.
531  *
532  * Attempt to scale mbmax so that mbcnt doesn't become limiting
533  * if buffering efficiency is near the normal case.
534  *
535  * sb_max only applies to user-sockets (where rl != NULL).  It does
536  * not apply to kernel sockets or kernel-controlled sockets.  Note
537  * that NFS overrides the sockbuf limits created when nfsd creates
538  * a socket.
539  */
540 int
541 ssb_reserve(struct signalsockbuf *ssb, u_long cc, struct socket *so,
542 	    struct rlimit *rl)
543 {
544 	/*
545 	 * rl will only be NULL when we're in an interrupt (eg, in tcp_input)
546 	 * or when called from netgraph (ie, ngd_attach)
547 	 */
548 	if (rl && cc > sb_max_adj)
549 		cc = sb_max_adj;
550 	if (!chgsbsize(so->so_cred->cr_uidinfo, &ssb->ssb_hiwat, cc,
551 		       rl ? rl->rlim_cur : RLIM_INFINITY)) {
552 		return (0);
553 	}
554 	if (rl)
555 		ssb->ssb_mbmax = min(cc * sb_efficiency, sb_max);
556 	else
557 		ssb->ssb_mbmax = cc * sb_efficiency;
558 
559 	/*
560 	 * AUTOLOWAT is set on send buffers and prevents large writes
561 	 * from generating a huge number of context switches.
562 	 */
563 	if (ssb->ssb_flags & SSB_AUTOLOWAT) {
564 		ssb->ssb_lowat = ssb->ssb_hiwat / 2;
565 		if (ssb->ssb_lowat < MCLBYTES)
566 			ssb->ssb_lowat = MCLBYTES;
567 	}
568 	if (ssb->ssb_lowat > ssb->ssb_hiwat)
569 		ssb->ssb_lowat = ssb->ssb_hiwat;
570 	return (1);
571 }
572 
573 /*
574  * Free mbufs held by a socket, and reserved mbuf space.
575  */
576 void
577 ssb_release(struct signalsockbuf *ssb, struct socket *so)
578 {
579 	sbflush(&ssb->sb);
580 	(void)chgsbsize(so->so_cred->cr_uidinfo, &ssb->ssb_hiwat, 0,
581 	    RLIM_INFINITY);
582 	ssb->ssb_mbmax = 0;
583 }
584 
585 /*
586  * Some routines that return EOPNOTSUPP for entry points that are not
587  * supported by a protocol.  Fill in as needed.
588  */
589 int
590 pru_accept_notsupp(struct socket *so, struct sockaddr **nam)
591 {
592 	return EOPNOTSUPP;
593 }
594 
595 int
596 pru_bind_notsupp(struct socket *so, struct sockaddr *nam, struct thread *td)
597 {
598 	return EOPNOTSUPP;
599 }
600 
601 int
602 pru_connect_notsupp(struct socket *so, struct sockaddr *nam, struct thread *td)
603 {
604 	return EOPNOTSUPP;
605 }
606 
607 int
608 pru_connect2_notsupp(struct socket *so1, struct socket *so2)
609 {
610 	return EOPNOTSUPP;
611 }
612 
613 int
614 pru_control_notsupp(struct socket *so, u_long cmd, caddr_t data,
615 		    struct ifnet *ifp, struct thread *td)
616 {
617 	return EOPNOTSUPP;
618 }
619 
620 int
621 pru_disconnect_notsupp(struct socket *so)
622 {
623 	return EOPNOTSUPP;
624 }
625 
626 int
627 pru_listen_notsupp(struct socket *so, struct thread *td)
628 {
629 	return EOPNOTSUPP;
630 }
631 
632 int
633 pru_peeraddr_notsupp(struct socket *so, struct sockaddr **nam)
634 {
635 	return EOPNOTSUPP;
636 }
637 
638 int
639 pru_rcvd_notsupp(struct socket *so, int flags)
640 {
641 	return EOPNOTSUPP;
642 }
643 
644 int
645 pru_rcvoob_notsupp(struct socket *so, struct mbuf *m, int flags)
646 {
647 	return EOPNOTSUPP;
648 }
649 
650 int
651 pru_shutdown_notsupp(struct socket *so)
652 {
653 	return EOPNOTSUPP;
654 }
655 
656 int
657 pru_sockaddr_notsupp(struct socket *so, struct sockaddr **nam)
658 {
659 	return EOPNOTSUPP;
660 }
661 
662 int
663 pru_sosend_notsupp(struct socket *so, struct sockaddr *addr, struct uio *uio,
664 	   struct mbuf *top, struct mbuf *control, int flags,
665 	   struct thread *td)
666 {
667 	if (top)
668 		m_freem(top);
669 	if (control)
670 		m_freem(control);
671 	return (EOPNOTSUPP);
672 }
673 
674 int
675 pru_soreceive_notsupp(struct socket *so, struct sockaddr **paddr,
676 		      struct uio *uio, struct sockbuf *sio,
677 		      struct mbuf **controlp, int *flagsp)
678 {
679 	return (EOPNOTSUPP);
680 }
681 
682 int
683 pru_ctloutput_notsupp(struct socket *so, struct sockopt *sopt)
684 {
685 	return (EOPNOTSUPP);
686 }
687 
688 /*
689  * This isn't really a ``null'' operation, but it's the default one
690  * and doesn't do anything destructive.
691  */
692 int
693 pru_sense_null(struct socket *so, struct stat *sb)
694 {
695 	sb->st_blksize = so->so_snd.ssb_hiwat;
696 	return 0;
697 }
698 
699 /*
700  * Make a copy of a sockaddr in a malloced buffer of type M_SONAME.  Callers
701  * of this routine assume that it always succeeds, so we have to use a
702  * blockable allocation even though we might be called from a critical thread.
703  */
704 struct sockaddr *
705 dup_sockaddr(const struct sockaddr *sa)
706 {
707 	struct sockaddr *sa2;
708 
709 	sa2 = kmalloc(sa->sa_len, M_SONAME, M_INTWAIT);
710 	bcopy(sa, sa2, sa->sa_len);
711 	return (sa2);
712 }
713 
714 /*
715  * Create an external-format (``xsocket'') structure using the information
716  * in the kernel-format socket structure pointed to by so.  This is done
717  * to reduce the spew of irrelevant information over this interface,
718  * to isolate user code from changes in the kernel structure, and
719  * potentially to provide information-hiding if we decide that
720  * some of this information should be hidden from users.
721  */
722 void
723 sotoxsocket(struct socket *so, struct xsocket *xso)
724 {
725 	xso->xso_len = sizeof *xso;
726 	xso->xso_so = so;
727 	xso->so_type = so->so_type;
728 	xso->so_options = so->so_options;
729 	xso->so_linger = so->so_linger;
730 	xso->so_state = so->so_state;
731 	xso->so_pcb = so->so_pcb;
732 	xso->xso_protocol = so->so_proto->pr_protocol;
733 	xso->xso_family = so->so_proto->pr_domain->dom_family;
734 	xso->so_qlen = so->so_qlen;
735 	xso->so_incqlen = so->so_incqlen;
736 	xso->so_qlimit = so->so_qlimit;
737 	xso->so_timeo = so->so_timeo;
738 	xso->so_error = so->so_error;
739 	xso->so_pgid = so->so_sigio ? so->so_sigio->sio_pgid : 0;
740 	xso->so_oobmark = so->so_oobmark;
741 	ssbtoxsockbuf(&so->so_snd, &xso->so_snd);
742 	ssbtoxsockbuf(&so->so_rcv, &xso->so_rcv);
743 	xso->so_uid = so->so_cred->cr_uid;
744 }
745 
746 /*
747  * Here is the definition of some of the basic objects in the kern.ipc
748  * branch of the MIB.
749  */
750 SYSCTL_NODE(_kern, KERN_IPC, ipc, CTLFLAG_RW, 0, "IPC");
751 
752 /*
753  * This takes the place of kern.maxsockbuf, which moved to kern.ipc.
754  *
755  * NOTE! sb_max only applies to user-created socket buffers.
756  */
757 static int dummy;
758 SYSCTL_INT(_kern, KERN_DUMMY, dummy, CTLFLAG_RW, &dummy, 0, "");
759 SYSCTL_OID(_kern_ipc, KIPC_MAXSOCKBUF, maxsockbuf, CTLTYPE_INT|CTLFLAG_RW,
760     &sb_max, 0, sysctl_handle_sb_max, "I", "Maximum socket buffer size");
761 SYSCTL_INT(_kern_ipc, OID_AUTO, maxsockets, CTLFLAG_RD,
762     &maxsockets, 0, "Maximum number of sockets available");
763 SYSCTL_INT(_kern_ipc, KIPC_SOCKBUF_WASTE, sockbuf_waste_factor, CTLFLAG_RW,
764     &sb_efficiency, 0, "");
765 
766 /*
767  * Initialize maxsockets
768  */
769 static void
770 init_maxsockets(void *ignored)
771 {
772     TUNABLE_INT_FETCH("kern.ipc.maxsockets", &maxsockets);
773     maxsockets = imax(maxsockets, imax(maxfiles, nmbclusters));
774 }
775 SYSINIT(param, SI_BOOT1_TUNABLES, SI_ORDER_ANY,
776 	init_maxsockets, NULL);
777 
778