uipc_socket2.c revision 1.21
1/*	$NetBSD: uipc_socket2.c,v 1.21 1997/10/09 13:00:00 mycroft Exp $	*/
2
3/*
4 * Copyright (c) 1982, 1986, 1988, 1990, 1993
5 *	The Regents of the University of California.  All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
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. All advertising materials mentioning features or use of this software
16 *    must display the following acknowledgement:
17 *	This product includes software developed by the University of
18 *	California, Berkeley and its contributors.
19 * 4. Neither the name of the University nor the names of its contributors
20 *    may be used to endorse or promote products derived from this software
21 *    without specific prior written permission.
22 *
23 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 * SUCH DAMAGE.
34 *
35 *	@(#)uipc_socket2.c	8.1 (Berkeley) 6/10/93
36 */
37
38#include <sys/param.h>
39#include <sys/systm.h>
40#include <sys/proc.h>
41#include <sys/file.h>
42#include <sys/buf.h>
43#include <sys/malloc.h>
44#include <sys/mbuf.h>
45#include <sys/protosw.h>
46#include <sys/socket.h>
47#include <sys/socketvar.h>
48#include <sys/signalvar.h>
49
50/*
51 * Primitive routines for operating on sockets and socket buffers
52 */
53
54/* strings for sleep message: */
55const char	netio[] = "netio";
56const char	netcon[] = "netcon";
57const char	netcls[] = "netcls";
58
59u_long	sb_max = SB_MAX;		/* patchable */
60
61/*
62 * Procedures to manipulate state flags of socket
63 * and do appropriate wakeups.  Normal sequence from the
64 * active (originating) side is that soisconnecting() is
65 * called during processing of connect() call,
66 * resulting in an eventual call to soisconnected() if/when the
67 * connection is established.  When the connection is torn down
68 * soisdisconnecting() is called during processing of disconnect() call,
69 * and soisdisconnected() is called when the connection to the peer
70 * is totally severed.  The semantics of these routines are such that
71 * connectionless protocols can call soisconnected() and soisdisconnected()
72 * only, bypassing the in-progress calls when setting up a ``connection''
73 * takes no time.
74 *
75 * From the passive side, a socket is created with
76 * two queues of sockets: so_q0 for connections in progress
77 * and so_q for connections already made and awaiting user acceptance.
78 * As a protocol is preparing incoming connections, it creates a socket
79 * structure queued on so_q0 by calling sonewconn().  When the connection
80 * is established, soisconnected() is called, and transfers the
81 * socket structure to so_q, making it available to accept().
82 *
83 * If a socket is closed with sockets on either
84 * so_q0 or so_q, these sockets are dropped.
85 *
86 * If higher level protocols are implemented in
87 * the kernel, the wakeups done here will sometimes
88 * cause software-interrupt process scheduling.
89 */
90
91void
92soisconnecting(so)
93	register struct socket *so;
94{
95
96	so->so_state &= ~(SS_ISCONNECTED|SS_ISDISCONNECTING);
97	so->so_state |= SS_ISCONNECTING;
98}
99
100void
101soisconnected(so)
102	register struct socket *so;
103{
104	register struct socket *head = so->so_head;
105
106	so->so_state &= ~(SS_ISCONNECTING|SS_ISDISCONNECTING|SS_ISCONFIRMING);
107	so->so_state |= SS_ISCONNECTED;
108	if (head && soqremque(so, 0)) {
109		soqinsque(head, so, 1);
110		sorwakeup(head);
111		wakeup((caddr_t)&head->so_timeo);
112	} else {
113		wakeup((caddr_t)&so->so_timeo);
114		sorwakeup(so);
115		sowwakeup(so);
116	}
117}
118
119void
120soisdisconnecting(so)
121	register struct socket *so;
122{
123
124	so->so_state &= ~SS_ISCONNECTING;
125	so->so_state |= (SS_ISDISCONNECTING|SS_CANTRCVMORE|SS_CANTSENDMORE);
126	wakeup((caddr_t)&so->so_timeo);
127	sowwakeup(so);
128	sorwakeup(so);
129}
130
131void
132soisdisconnected(so)
133	register struct socket *so;
134{
135
136	so->so_state &= ~(SS_ISCONNECTING|SS_ISCONNECTED|SS_ISDISCONNECTING);
137	so->so_state |= (SS_CANTRCVMORE|SS_CANTSENDMORE);
138	wakeup((caddr_t)&so->so_timeo);
139	sowwakeup(so);
140	sorwakeup(so);
141}
142
143/*
144 * When an attempt at a new connection is noted on a socket
145 * which accepts connections, sonewconn is called.  If the
146 * connection is possible (subject to space constraints, etc.)
147 * then we allocate a new structure, propoerly linked into the
148 * data structure of the original socket, and return this.
149 * Connstatus may be 0, or SO_ISCONFIRMING, or SO_ISCONNECTED.
150 *
151 * Currently, sonewconn() is defined as sonewconn1() in socketvar.h
152 * to catch calls that are missing the (new) second parameter.
153 */
154struct socket *
155sonewconn1(head, connstatus)
156	register struct socket *head;
157	int connstatus;
158{
159	register struct socket *so;
160	int soqueue = connstatus ? 1 : 0;
161
162	if (head->so_qlen + head->so_q0len > 3 * head->so_qlimit / 2)
163		return ((struct socket *)0);
164	MALLOC(so, struct socket *, sizeof(*so), M_SOCKET, M_DONTWAIT);
165	if (so == NULL)
166		return ((struct socket *)0);
167	bzero((caddr_t)so, sizeof(*so));
168	so->so_type = head->so_type;
169	so->so_options = head->so_options &~ SO_ACCEPTCONN;
170	so->so_linger = head->so_linger;
171	so->so_state = head->so_state | SS_NOFDREF;
172	so->so_proto = head->so_proto;
173	so->so_timeo = head->so_timeo;
174	so->so_pgid = head->so_pgid;
175	(void) soreserve(so, head->so_snd.sb_hiwat, head->so_rcv.sb_hiwat);
176	soqinsque(head, so, soqueue);
177	if ((*so->so_proto->pr_usrreq)(so, PRU_ATTACH,
178	    (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0,
179	    (struct proc *)0)) {
180		(void) soqremque(so, soqueue);
181		(void) free((caddr_t)so, M_SOCKET);
182		return ((struct socket *)0);
183	}
184	if (connstatus) {
185		sorwakeup(head);
186		wakeup((caddr_t)&head->so_timeo);
187		so->so_state |= connstatus;
188	}
189	return (so);
190}
191
192void
193soqinsque(head, so, q)
194	register struct socket *head, *so;
195	int q;
196{
197
198	register struct socket **prev;
199	so->so_head = head;
200	if (q == 0) {
201		head->so_q0len++;
202		so->so_q0 = 0;
203		for (prev = &(head->so_q0); *prev; )
204			prev = &((*prev)->so_q0);
205	} else {
206		head->so_qlen++;
207		so->so_q = 0;
208		for (prev = &(head->so_q); *prev; )
209			prev = &((*prev)->so_q);
210	}
211	*prev = so;
212}
213
214int
215soqremque(so, q)
216	register struct socket *so;
217	int q;
218{
219	register struct socket *head, *prev, *next;
220
221	head = so->so_head;
222	prev = head;
223	for (;;) {
224		next = q ? prev->so_q : prev->so_q0;
225		if (next == so)
226			break;
227		if (next == 0)
228			return (0);
229		prev = next;
230	}
231	if (q == 0) {
232		prev->so_q0 = next->so_q0;
233		head->so_q0len--;
234	} else {
235		prev->so_q = next->so_q;
236		head->so_qlen--;
237	}
238	next->so_q0 = next->so_q = 0;
239	next->so_head = 0;
240	return (1);
241}
242
243/*
244 * Socantsendmore indicates that no more data will be sent on the
245 * socket; it would normally be applied to a socket when the user
246 * informs the system that no more data is to be sent, by the protocol
247 * code (in case PRU_SHUTDOWN).  Socantrcvmore indicates that no more data
248 * will be received, and will normally be applied to the socket by a
249 * protocol when it detects that the peer will send no more data.
250 * Data queued for reading in the socket may yet be read.
251 */
252
253void
254socantsendmore(so)
255	struct socket *so;
256{
257
258	so->so_state |= SS_CANTSENDMORE;
259	sowwakeup(so);
260}
261
262void
263socantrcvmore(so)
264	struct socket *so;
265{
266
267	so->so_state |= SS_CANTRCVMORE;
268	sorwakeup(so);
269}
270
271/*
272 * Wait for data to arrive at/drain from a socket buffer.
273 */
274int
275sbwait(sb)
276	struct sockbuf *sb;
277{
278
279	sb->sb_flags |= SB_WAIT;
280	return (tsleep((caddr_t)&sb->sb_cc,
281	    (sb->sb_flags & SB_NOINTR) ? PSOCK : PSOCK | PCATCH, netio,
282	    sb->sb_timeo));
283}
284
285/*
286 * Lock a sockbuf already known to be locked;
287 * return any error returned from sleep (EINTR).
288 */
289int
290sb_lock(sb)
291	register struct sockbuf *sb;
292{
293	int error;
294
295	while (sb->sb_flags & SB_LOCK) {
296		sb->sb_flags |= SB_WANT;
297		error = tsleep((caddr_t)&sb->sb_flags,
298			       (sb->sb_flags & SB_NOINTR) ?
299					PSOCK : PSOCK|PCATCH, netio, 0);
300		if (error)
301			return (error);
302	}
303	sb->sb_flags |= SB_LOCK;
304	return (0);
305}
306
307/*
308 * Wakeup processes waiting on a socket buffer.
309 * Do asynchronous notification via SIGIO
310 * if the socket has the SS_ASYNC flag set.
311 */
312void
313sowakeup(so, sb)
314	register struct socket *so;
315	register struct sockbuf *sb;
316{
317	struct proc *p;
318
319	selwakeup(&sb->sb_sel);
320	sb->sb_flags &= ~SB_SEL;
321	if (sb->sb_flags & SB_WAIT) {
322		sb->sb_flags &= ~SB_WAIT;
323		wakeup((caddr_t)&sb->sb_cc);
324	}
325	if (so->so_state & SS_ASYNC) {
326		if (so->so_pgid < 0)
327			gsignal(-so->so_pgid, SIGIO);
328		else if (so->so_pgid > 0 && (p = pfind(so->so_pgid)) != 0)
329			psignal(p, SIGIO);
330	}
331}
332
333/*
334 * Socket buffer (struct sockbuf) utility routines.
335 *
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 poll() statements and notification
340 * on data availability to be implemented.
341 *
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:
347 *
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_CONTROL).
355 * 3. If a name or rights record exists, then it must be followed by
356 *    a data record, perhaps of zero length.
357 *
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.
363 */
364
365int
366soreserve(so, sndcc, rcvcc)
367	register struct socket *so;
368	u_long sndcc, rcvcc;
369{
370
371	if (sbreserve(&so->so_snd, sndcc) == 0)
372		goto bad;
373	if (sbreserve(&so->so_rcv, rcvcc) == 0)
374		goto bad2;
375	if (so->so_rcv.sb_lowat == 0)
376		so->so_rcv.sb_lowat = 1;
377	if (so->so_snd.sb_lowat == 0)
378		so->so_snd.sb_lowat = MCLBYTES;
379	if (so->so_snd.sb_lowat > so->so_snd.sb_hiwat)
380		so->so_snd.sb_lowat = so->so_snd.sb_hiwat;
381	return (0);
382bad2:
383	sbrelease(&so->so_snd);
384bad:
385	return (ENOBUFS);
386}
387
388/*
389 * Allot mbufs to a sockbuf.
390 * Attempt to scale mbmax so that mbcnt doesn't become limiting
391 * if buffering efficiency is near the normal case.
392 */
393int
394sbreserve(sb, cc)
395	struct sockbuf *sb;
396	u_long cc;
397{
398
399	if (cc == 0 || cc > sb_max * MCLBYTES / (MSIZE + MCLBYTES))
400		return (0);
401	sb->sb_hiwat = cc;
402	sb->sb_mbmax = min(cc * 2, sb_max);
403	if (sb->sb_lowat > sb->sb_hiwat)
404		sb->sb_lowat = sb->sb_hiwat;
405	return (1);
406}
407
408/*
409 * Free mbufs held by a socket, and reserved mbuf space.
410 */
411void
412sbrelease(sb)
413	struct sockbuf *sb;
414{
415
416	sbflush(sb);
417	sb->sb_hiwat = sb->sb_mbmax = 0;
418}
419
420/*
421 * Routines to add and remove
422 * data from an mbuf queue.
423 *
424 * The routines sbappend() or sbappendrecord() are normally called to
425 * append new mbufs to a socket buffer, after checking that adequate
426 * space is available, comparing the function sbspace() with the amount
427 * of data to be added.  sbappendrecord() differs from sbappend() in
428 * that data supplied is treated as the beginning of a new record.
429 * To place a sender's address, optional access rights, and data in a
430 * socket receive buffer, sbappendaddr() should be used.  To place
431 * access rights and data in a socket receive buffer, sbappendrights()
432 * should be used.  In either case, the new data begins a new record.
433 * Note that unlike sbappend() and sbappendrecord(), these routines check
434 * for the caller that there will be enough space to store the data.
435 * Each fails if there is not enough space, or if it cannot find mbufs
436 * to store additional information in.
437 *
438 * Reliable protocols may use the socket send buffer to hold data
439 * awaiting acknowledgement.  Data is normally copied from a socket
440 * send buffer in a protocol with m_copy for output to a peer,
441 * and then removing the data from the socket buffer with sbdrop()
442 * or sbdroprecord() when the data is acknowledged by the peer.
443 */
444
445/*
446 * Append mbuf chain m to the last record in the
447 * socket buffer sb.  The additional space associated
448 * the mbuf chain is recorded in sb.  Empty mbufs are
449 * discarded and mbufs are compacted where possible.
450 */
451void
452sbappend(sb, m)
453	struct sockbuf *sb;
454	struct mbuf *m;
455{
456	register struct mbuf *n;
457
458	if (m == 0)
459		return;
460	if ((n = sb->sb_mb) != NULL) {
461		while (n->m_nextpkt)
462			n = n->m_nextpkt;
463		do {
464			if (n->m_flags & M_EOR) {
465				sbappendrecord(sb, m); /* XXXXXX!!!! */
466				return;
467			}
468		} while (n->m_next && (n = n->m_next));
469	}
470	sbcompress(sb, m, n);
471}
472
473#ifdef SOCKBUF_DEBUG
474void
475sbcheck(sb)
476	register struct sockbuf *sb;
477{
478	register struct mbuf *m;
479	register int len = 0, mbcnt = 0;
480
481	for (m = sb->sb_mb; m; m = m->m_next) {
482		len += m->m_len;
483		mbcnt += MSIZE;
484		if (m->m_flags & M_EXT)
485			mbcnt += m->m_ext.ext_size;
486		if (m->m_nextpkt)
487			panic("sbcheck nextpkt");
488	}
489	if (len != sb->sb_cc || mbcnt != sb->sb_mbcnt) {
490		printf("cc %d != %d || mbcnt %d != %d\n", len, sb->sb_cc,
491		    mbcnt, sb->sb_mbcnt);
492		panic("sbcheck");
493	}
494}
495#endif
496
497/*
498 * As above, except the mbuf chain
499 * begins a new record.
500 */
501void
502sbappendrecord(sb, m0)
503	register struct sockbuf *sb;
504	register struct mbuf *m0;
505{
506	register struct mbuf *m;
507
508	if (m0 == 0)
509		return;
510	if ((m = sb->sb_mb) != NULL)
511		while (m->m_nextpkt)
512			m = m->m_nextpkt;
513	/*
514	 * Put the first mbuf on the queue.
515	 * Note this permits zero length records.
516	 */
517	sballoc(sb, m0);
518	if (m)
519		m->m_nextpkt = m0;
520	else
521		sb->sb_mb = m0;
522	m = m0->m_next;
523	m0->m_next = 0;
524	if (m && (m0->m_flags & M_EOR)) {
525		m0->m_flags &= ~M_EOR;
526		m->m_flags |= M_EOR;
527	}
528	sbcompress(sb, m, m0);
529}
530
531/*
532 * As above except that OOB data
533 * is inserted at the beginning of the sockbuf,
534 * but after any other OOB data.
535 */
536void
537sbinsertoob(sb, m0)
538	register struct sockbuf *sb;
539	register struct mbuf *m0;
540{
541	register struct mbuf *m;
542	register struct mbuf **mp;
543
544	if (m0 == 0)
545		return;
546	for (mp = &sb->sb_mb; (m = *mp) != NULL; mp = &((*mp)->m_nextpkt)) {
547	    again:
548		switch (m->m_type) {
549
550		case MT_OOBDATA:
551			continue;		/* WANT next train */
552
553		case MT_CONTROL:
554			if ((m = m->m_next) != NULL)
555				goto again;	/* inspect THIS train further */
556		}
557		break;
558	}
559	/*
560	 * Put the first mbuf on the queue.
561	 * Note this permits zero length records.
562	 */
563	sballoc(sb, m0);
564	m0->m_nextpkt = *mp;
565	*mp = m0;
566	m = m0->m_next;
567	m0->m_next = 0;
568	if (m && (m0->m_flags & M_EOR)) {
569		m0->m_flags &= ~M_EOR;
570		m->m_flags |= M_EOR;
571	}
572	sbcompress(sb, m, m0);
573}
574
575/*
576 * Append address and data, and optionally, control (ancillary) data
577 * to the receive queue of a socket.  If present,
578 * m0 must include a packet header with total length.
579 * Returns 0 if no space in sockbuf or insufficient mbufs.
580 */
581int
582sbappendaddr(sb, asa, m0, control)
583	register struct sockbuf *sb;
584	struct sockaddr *asa;
585	struct mbuf *m0, *control;
586{
587	register struct mbuf *m, *n;
588	int space = asa->sa_len;
589
590if (m0 && (m0->m_flags & M_PKTHDR) == 0)
591panic("sbappendaddr");
592	if (m0)
593		space += m0->m_pkthdr.len;
594	for (n = control; n; n = n->m_next) {
595		space += n->m_len;
596		if (n->m_next == 0)	/* keep pointer to last control buf */
597			break;
598	}
599	if (space > sbspace(sb))
600		return (0);
601	MGET(m, M_DONTWAIT, MT_SONAME);
602	if (m == 0)
603		return (0);
604	if (asa->sa_len > MLEN) {
605		MEXTMALLOC(m, asa->sa_len, M_NOWAIT);
606		if ((m->m_flags & M_EXT) == 0) {
607			m_free(m);
608			return (0);
609		}
610	}
611	m->m_len = asa->sa_len;
612	bcopy((caddr_t)asa, mtod(m, caddr_t), asa->sa_len);
613	if (n)
614		n->m_next = m0;		/* concatenate data to control */
615	else
616		control = m0;
617	m->m_next = control;
618	for (n = m; n; n = n->m_next)
619		sballoc(sb, n);
620	if ((n = sb->sb_mb) != NULL) {
621		while (n->m_nextpkt)
622			n = n->m_nextpkt;
623		n->m_nextpkt = m;
624	} else
625		sb->sb_mb = m;
626	return (1);
627}
628
629int
630sbappendcontrol(sb, m0, control)
631	struct sockbuf *sb;
632	struct mbuf *m0, *control;
633{
634	register struct mbuf *m, *n;
635	int space = 0;
636
637	if (control == 0)
638		panic("sbappendcontrol");
639	for (m = control; ; m = m->m_next) {
640		space += m->m_len;
641		if (m->m_next == 0)
642			break;
643	}
644	n = m;			/* save pointer to last control buffer */
645	for (m = m0; m; m = m->m_next)
646		space += m->m_len;
647	if (space > sbspace(sb))
648		return (0);
649	n->m_next = m0;			/* concatenate data to control */
650	for (m = control; m; m = m->m_next)
651		sballoc(sb, m);
652	if ((n = sb->sb_mb) != NULL) {
653		while (n->m_nextpkt)
654			n = n->m_nextpkt;
655		n->m_nextpkt = control;
656	} else
657		sb->sb_mb = control;
658	return (1);
659}
660
661/*
662 * Compress mbuf chain m into the socket
663 * buffer sb following mbuf n.  If n
664 * is null, the buffer is presumed empty.
665 */
666void
667sbcompress(sb, m, n)
668	register struct sockbuf *sb;
669	register struct mbuf *m, *n;
670{
671	register int eor = 0;
672	register struct mbuf *o;
673
674	while (m) {
675		eor |= m->m_flags & M_EOR;
676		if (m->m_len == 0 &&
677		    (eor == 0 ||
678		     (((o = m->m_next) || (o = n)) &&
679		      o->m_type == m->m_type))) {
680			m = m_free(m);
681			continue;
682		}
683		if (n && (n->m_flags & (M_EXT | M_EOR)) == 0 &&
684		    (n->m_data + n->m_len + m->m_len) < &n->m_dat[MLEN] &&
685		    n->m_type == m->m_type) {
686			bcopy(mtod(m, caddr_t), mtod(n, caddr_t) + n->m_len,
687			    (unsigned)m->m_len);
688			n->m_len += m->m_len;
689			sb->sb_cc += m->m_len;
690			m = m_free(m);
691			continue;
692		}
693		if (n)
694			n->m_next = m;
695		else
696			sb->sb_mb = m;
697		sballoc(sb, m);
698		n = m;
699		m->m_flags &= ~M_EOR;
700		m = m->m_next;
701		n->m_next = 0;
702	}
703	if (eor) {
704		if (n)
705			n->m_flags |= eor;
706		else
707			printf("semi-panic: sbcompress\n");
708	}
709}
710
711/*
712 * Free all mbufs in a sockbuf.
713 * Check that all resources are reclaimed.
714 */
715void
716sbflush(sb)
717	register struct sockbuf *sb;
718{
719
720	if (sb->sb_flags & SB_LOCK)
721		panic("sbflush");
722	while (sb->sb_mbcnt)
723		sbdrop(sb, (int)sb->sb_cc);
724	if (sb->sb_cc || sb->sb_mb)
725		panic("sbflush 2");
726}
727
728/*
729 * Drop data from (the front of) a sockbuf.
730 */
731void
732sbdrop(sb, len)
733	register struct sockbuf *sb;
734	register int len;
735{
736	register struct mbuf *m, *mn;
737	struct mbuf *next;
738
739	next = (m = sb->sb_mb) ? m->m_nextpkt : 0;
740	while (len > 0) {
741		if (m == 0) {
742			if (next == 0)
743				panic("sbdrop");
744			m = next;
745			next = m->m_nextpkt;
746			continue;
747		}
748		if (m->m_len > len) {
749			m->m_len -= len;
750			m->m_data += len;
751			sb->sb_cc -= len;
752			break;
753		}
754		len -= m->m_len;
755		sbfree(sb, m);
756		MFREE(m, mn);
757		m = mn;
758	}
759	while (m && m->m_len == 0) {
760		sbfree(sb, m);
761		MFREE(m, mn);
762		m = mn;
763	}
764	if (m) {
765		sb->sb_mb = m;
766		m->m_nextpkt = next;
767	} else
768		sb->sb_mb = next;
769}
770
771/*
772 * Drop a record off the front of a sockbuf
773 * and move the next record to the front.
774 */
775void
776sbdroprecord(sb)
777	register struct sockbuf *sb;
778{
779	register struct mbuf *m, *mn;
780
781	m = sb->sb_mb;
782	if (m) {
783		sb->sb_mb = m->m_nextpkt;
784		do {
785			sbfree(sb, m);
786			MFREE(m, mn);
787		} while ((m = mn) != NULL);
788	}
789}
790
791/*
792 * Create a "control" mbuf containing the specified data
793 * with the specified type for presentation on a socket buffer.
794 */
795struct mbuf *
796sbcreatecontrol(p, size, type, level)
797	caddr_t p;
798	register int size;
799	int type, level;
800{
801	register struct cmsghdr *cp;
802	struct mbuf *m;
803
804	if ((m = m_get(M_DONTWAIT, MT_CONTROL)) == NULL)
805		return ((struct mbuf *) NULL);
806	cp = mtod(m, struct cmsghdr *);
807	bcopy(p, CMSG_DATA(cp), size);
808	size += sizeof(*cp);
809	m->m_len = size;
810	cp->cmsg_len = size;
811	cp->cmsg_level = level;
812	cp->cmsg_type = type;
813	return (m);
814}
815