uipc_socket2.c revision 1.10
1/*	$NetBSD: uipc_socket2.c,v 1.10 1995/08/16 01:03:19 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
49/*
50 * Primitive routines for operating on sockets and socket buffers
51 */
52
53/* strings for sleep message: */
54char	netio[] = "netio";
55char	netcon[] = "netcon";
56char	netcls[] = "netcls";
57
58u_long	sb_max = SB_MAX;		/* patchable */
59
60/*
61 * Procedures to manipulate state flags of socket
62 * and do appropriate wakeups.  Normal sequence from the
63 * active (originating) side is that soisconnecting() is
64 * called during processing of connect() call,
65 * resulting in an eventual call to soisconnected() if/when the
66 * connection is established.  When the connection is torn down
67 * soisdisconnecting() is called during processing of disconnect() call,
68 * and soisdisconnected() is called when the connection to the peer
69 * is totally severed.  The semantics of these routines are such that
70 * connectionless protocols can call soisconnected() and soisdisconnected()
71 * only, bypassing the in-progress calls when setting up a ``connection''
72 * takes no time.
73 *
74 * From the passive side, a socket is created with
75 * two queues of sockets: so_q0 for connections in progress
76 * and so_q for connections already made and awaiting user acceptance.
77 * As a protocol is preparing incoming connections, it creates a socket
78 * structure queued on so_q0 by calling sonewconn().  When the connection
79 * is established, soisconnected() is called, and transfers the
80 * socket structure to so_q, making it available to accept().
81 *
82 * If a socket is closed with sockets on either
83 * so_q0 or so_q, these sockets are dropped.
84 *
85 * If higher level protocols are implemented in
86 * the kernel, the wakeups done here will sometimes
87 * cause software-interrupt process scheduling.
88 */
89
90void
91soisconnecting(so)
92	register struct socket *so;
93{
94
95	so->so_state &= ~(SS_ISCONNECTED|SS_ISDISCONNECTING);
96	so->so_state |= SS_ISCONNECTING;
97}
98
99void
100soisconnected(so)
101	register struct socket *so;
102{
103	register struct socket *head = so->so_head;
104
105	so->so_state &= ~(SS_ISCONNECTING|SS_ISDISCONNECTING|SS_ISCONFIRMING);
106	so->so_state |= SS_ISCONNECTED;
107	if (head && soqremque(so, 0)) {
108		soqinsque(head, so, 1);
109		sorwakeup(head);
110		wakeup((caddr_t)&head->so_timeo);
111	} else {
112		wakeup((caddr_t)&so->so_timeo);
113		sorwakeup(so);
114		sowwakeup(so);
115	}
116}
117
118void
119soisdisconnecting(so)
120	register struct socket *so;
121{
122
123	so->so_state &= ~SS_ISCONNECTING;
124	so->so_state |= (SS_ISDISCONNECTING|SS_CANTRCVMORE|SS_CANTSENDMORE);
125	wakeup((caddr_t)&so->so_timeo);
126	sowwakeup(so);
127	sorwakeup(so);
128}
129
130void
131soisdisconnected(so)
132	register struct socket *so;
133{
134
135	so->so_state &= ~(SS_ISCONNECTING|SS_ISCONNECTED|SS_ISDISCONNECTING);
136	so->so_state |= (SS_CANTRCVMORE|SS_CANTSENDMORE);
137	wakeup((caddr_t)&so->so_timeo);
138	sowwakeup(so);
139	sorwakeup(so);
140}
141
142/*
143 * When an attempt at a new connection is noted on a socket
144 * which accepts connections, sonewconn is called.  If the
145 * connection is possible (subject to space constraints, etc.)
146 * then we allocate a new structure, propoerly linked into the
147 * data structure of the original socket, and return this.
148 * Connstatus may be 0, or SO_ISCONFIRMING, or SO_ISCONNECTED.
149 *
150 * Currently, sonewconn() is defined as sonewconn1() in socketvar.h
151 * to catch calls that are missing the (new) second parameter.
152 */
153struct socket *
154sonewconn1(head, connstatus)
155	register struct socket *head;
156	int connstatus;
157{
158	register struct socket *so;
159	int soqueue = connstatus ? 1 : 0;
160
161	if (head->so_qlen + head->so_q0len > 3 * head->so_qlimit / 2)
162		return ((struct socket *)0);
163	MALLOC(so, struct socket *, sizeof(*so), M_SOCKET, M_DONTWAIT);
164	if (so == NULL)
165		return ((struct socket *)0);
166	bzero((caddr_t)so, sizeof(*so));
167	so->so_type = head->so_type;
168	so->so_options = head->so_options &~ SO_ACCEPTCONN;
169	so->so_linger = head->so_linger;
170	so->so_state = head->so_state | SS_NOFDREF;
171	so->so_proto = head->so_proto;
172	so->so_timeo = head->so_timeo;
173	so->so_pgid = head->so_pgid;
174	(void) soreserve(so, head->so_snd.sb_hiwat, head->so_rcv.sb_hiwat);
175	soqinsque(head, so, soqueue);
176	if ((*so->so_proto->pr_usrreq)(so, PRU_ATTACH,
177	    (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0)) {
178		(void) soqremque(so, soqueue);
179		(void) free((caddr_t)so, M_SOCKET);
180		return ((struct socket *)0);
181	}
182	if (connstatus) {
183		sorwakeup(head);
184		wakeup((caddr_t)&head->so_timeo);
185		so->so_state |= connstatus;
186	}
187	return (so);
188}
189
190void
191soqinsque(head, so, q)
192	register struct socket *head, *so;
193	int q;
194{
195
196	register struct socket **prev;
197	so->so_head = head;
198	if (q == 0) {
199		head->so_q0len++;
200		so->so_q0 = 0;
201		for (prev = &(head->so_q0); *prev; )
202			prev = &((*prev)->so_q0);
203	} else {
204		head->so_qlen++;
205		so->so_q = 0;
206		for (prev = &(head->so_q); *prev; )
207			prev = &((*prev)->so_q);
208	}
209	*prev = so;
210}
211
212int
213soqremque(so, q)
214	register struct socket *so;
215	int q;
216{
217	register struct socket *head, *prev, *next;
218
219	head = so->so_head;
220	prev = head;
221	for (;;) {
222		next = q ? prev->so_q : prev->so_q0;
223		if (next == so)
224			break;
225		if (next == 0)
226			return (0);
227		prev = next;
228	}
229	if (q == 0) {
230		prev->so_q0 = next->so_q0;
231		head->so_q0len--;
232	} else {
233		prev->so_q = next->so_q;
234		head->so_qlen--;
235	}
236	next->so_q0 = next->so_q = 0;
237	next->so_head = 0;
238	return (1);
239}
240
241/*
242 * Socantsendmore indicates that no more data will be sent on the
243 * socket; it would normally be applied to a socket when the user
244 * informs the system that no more data is to be sent, by the protocol
245 * code (in case PRU_SHUTDOWN).  Socantrcvmore indicates that no more data
246 * will be received, and will normally be applied to the socket by a
247 * protocol when it detects that the peer will send no more data.
248 * Data queued for reading in the socket may yet be read.
249 */
250
251void
252socantsendmore(so)
253	struct socket *so;
254{
255
256	so->so_state |= SS_CANTSENDMORE;
257	sowwakeup(so);
258}
259
260void
261socantrcvmore(so)
262	struct socket *so;
263{
264
265	so->so_state |= SS_CANTRCVMORE;
266	sorwakeup(so);
267}
268
269/*
270 * Wait for data to arrive at/drain from a socket buffer.
271 */
272int
273sbwait(sb)
274	struct sockbuf *sb;
275{
276
277	sb->sb_flags |= SB_WAIT;
278	return (tsleep((caddr_t)&sb->sb_cc,
279	    (sb->sb_flags & SB_NOINTR) ? PSOCK : PSOCK | PCATCH, netio,
280	    sb->sb_timeo));
281}
282
283/*
284 * Lock a sockbuf already known to be locked;
285 * return any error returned from sleep (EINTR).
286 */
287int
288sb_lock(sb)
289	register struct sockbuf *sb;
290{
291	int error;
292
293	while (sb->sb_flags & SB_LOCK) {
294		sb->sb_flags |= SB_WANT;
295		if (error = tsleep((caddr_t)&sb->sb_flags,
296		    (sb->sb_flags & SB_NOINTR) ? PSOCK : PSOCK|PCATCH,
297		    netio, 0))
298			return (error);
299	}
300	sb->sb_flags |= SB_LOCK;
301	return (0);
302}
303
304/*
305 * Wakeup processes waiting on a socket buffer.
306 * Do asynchronous notification via SIGIO
307 * if the socket has the SS_ASYNC flag set.
308 */
309void
310sowakeup(so, sb)
311	register struct socket *so;
312	register struct sockbuf *sb;
313{
314	struct proc *p;
315
316	selwakeup(&sb->sb_sel);
317	sb->sb_flags &= ~SB_SEL;
318	if (sb->sb_flags & SB_WAIT) {
319		sb->sb_flags &= ~SB_WAIT;
320		wakeup((caddr_t)&sb->sb_cc);
321	}
322	if (so->so_state & SS_ASYNC) {
323		if (so->so_pgid < 0)
324			gsignal(-so->so_pgid, SIGIO);
325		else if (so->so_pgid > 0 && (p = pfind(so->so_pgid)) != 0)
326			psignal(p, SIGIO);
327	}
328}
329
330/*
331 * Socket buffer (struct sockbuf) utility routines.
332 *
333 * Each socket contains two socket buffers: one for sending data and
334 * one for receiving data.  Each buffer contains a queue of mbufs,
335 * information about the number of mbufs and amount of data in the
336 * queue, and other fields allowing select() statements and notification
337 * on data availability to be implemented.
338 *
339 * Data stored in a socket buffer is maintained as a list of records.
340 * Each record is a list of mbufs chained together with the m_next
341 * field.  Records are chained together with the m_nextpkt field. The upper
342 * level routine soreceive() expects the following conventions to be
343 * observed when placing information in the receive buffer:
344 *
345 * 1. If the protocol requires each message be preceded by the sender's
346 *    name, then a record containing that name must be present before
347 *    any associated data (mbuf's must be of type MT_SONAME).
348 * 2. If the protocol supports the exchange of ``access rights'' (really
349 *    just additional data associated with the message), and there are
350 *    ``rights'' to be received, then a record containing this data
351 *    should be present (mbuf's must be of type MT_CONTROL).
352 * 3. If a name or rights record exists, then it must be followed by
353 *    a data record, perhaps of zero length.
354 *
355 * Before using a new socket structure it is first necessary to reserve
356 * buffer space to the socket, by calling sbreserve().  This should commit
357 * some of the available buffer space in the system buffer pool for the
358 * socket (currently, it does nothing but enforce limits).  The space
359 * should be released by calling sbrelease() when the socket is destroyed.
360 */
361
362int
363soreserve(so, sndcc, rcvcc)
364	register struct socket *so;
365	u_long sndcc, rcvcc;
366{
367
368	if (sbreserve(&so->so_snd, sndcc) == 0)
369		goto bad;
370	if (sbreserve(&so->so_rcv, rcvcc) == 0)
371		goto bad2;
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;
378	return (0);
379bad2:
380	sbrelease(&so->so_snd);
381bad:
382	return (ENOBUFS);
383}
384
385/*
386 * Allot mbufs to a sockbuf.
387 * Attempt to scale mbmax so that mbcnt doesn't become limiting
388 * if buffering efficiency is near the normal case.
389 */
390int
391sbreserve(sb, cc)
392	struct sockbuf *sb;
393	u_long cc;
394{
395
396	if (cc > sb_max * MCLBYTES / (MSIZE + MCLBYTES))
397		return (0);
398	sb->sb_hiwat = cc;
399	sb->sb_mbmax = min(cc * 2, sb_max);
400	if (sb->sb_lowat > sb->sb_hiwat)
401		sb->sb_lowat = sb->sb_hiwat;
402	return (1);
403}
404
405/*
406 * Free mbufs held by a socket, and reserved mbuf space.
407 */
408void
409sbrelease(sb)
410	struct sockbuf *sb;
411{
412
413	sbflush(sb);
414	sb->sb_hiwat = sb->sb_mbmax = 0;
415}
416
417/*
418 * Routines to add and remove
419 * data from an mbuf queue.
420 *
421 * The routines sbappend() or sbappendrecord() are normally called to
422 * append new mbufs to a socket buffer, after checking that adequate
423 * space is available, comparing the function sbspace() with the amount
424 * of data to be added.  sbappendrecord() differs from sbappend() in
425 * that data supplied is treated as the beginning of a new record.
426 * To place a sender's address, optional access rights, and data in a
427 * socket receive buffer, sbappendaddr() should be used.  To place
428 * access rights and data in a socket receive buffer, sbappendrights()
429 * should be used.  In either case, the new data begins a new record.
430 * Note that unlike sbappend() and sbappendrecord(), these routines check
431 * for the caller that there will be enough space to store the data.
432 * Each fails if there is not enough space, or if it cannot find mbufs
433 * to store additional information in.
434 *
435 * Reliable protocols may use the socket send buffer to hold data
436 * awaiting acknowledgement.  Data is normally copied from a socket
437 * send buffer in a protocol with m_copy for output to a peer,
438 * and then removing the data from the socket buffer with sbdrop()
439 * or sbdroprecord() when the data is acknowledged by the peer.
440 */
441
442/*
443 * Append mbuf chain m to the last record in the
444 * socket buffer sb.  The additional space associated
445 * the mbuf chain is recorded in sb.  Empty mbufs are
446 * discarded and mbufs are compacted where possible.
447 */
448void
449sbappend(sb, m)
450	struct sockbuf *sb;
451	struct mbuf *m;
452{
453	register struct mbuf *n;
454
455	if (m == 0)
456		return;
457	if (n = sb->sb_mb) {
458		while (n->m_nextpkt)
459			n = n->m_nextpkt;
460		do {
461			if (n->m_flags & M_EOR) {
462				sbappendrecord(sb, m); /* XXXXXX!!!! */
463				return;
464			}
465		} while (n->m_next && (n = n->m_next));
466	}
467	sbcompress(sb, m, n);
468}
469
470#ifdef SOCKBUF_DEBUG
471void
472sbcheck(sb)
473	register struct sockbuf *sb;
474{
475	register struct mbuf *m;
476	register int len = 0, mbcnt = 0;
477
478	for (m = sb->sb_mb; m; m = m->m_next) {
479		len += m->m_len;
480		mbcnt += MSIZE;
481		if (m->m_flags & M_EXT)
482			mbcnt += m->m_ext.ext_size;
483		if (m->m_nextpkt)
484			panic("sbcheck nextpkt");
485	}
486	if (len != sb->sb_cc || mbcnt != sb->sb_mbcnt) {
487		printf("cc %d != %d || mbcnt %d != %d\n", len, sb->sb_cc,
488		    mbcnt, sb->sb_mbcnt);
489		panic("sbcheck");
490	}
491}
492#endif
493
494/*
495 * As above, except the mbuf chain
496 * begins a new record.
497 */
498void
499sbappendrecord(sb, m0)
500	register struct sockbuf *sb;
501	register struct mbuf *m0;
502{
503	register struct mbuf *m;
504
505	if (m0 == 0)
506		return;
507	if (m = sb->sb_mb)
508		while (m->m_nextpkt)
509			m = m->m_nextpkt;
510	/*
511	 * Put the first mbuf on the queue.
512	 * Note this permits zero length records.
513	 */
514	sballoc(sb, m0);
515	if (m)
516		m->m_nextpkt = m0;
517	else
518		sb->sb_mb = m0;
519	m = m0->m_next;
520	m0->m_next = 0;
521	if (m && (m0->m_flags & M_EOR)) {
522		m0->m_flags &= ~M_EOR;
523		m->m_flags |= M_EOR;
524	}
525	sbcompress(sb, m, m0);
526}
527
528/*
529 * As above except that OOB data
530 * is inserted at the beginning of the sockbuf,
531 * but after any other OOB data.
532 */
533void
534sbinsertoob(sb, m0)
535	register struct sockbuf *sb;
536	register struct mbuf *m0;
537{
538	register struct mbuf *m;
539	register struct mbuf **mp;
540
541	if (m0 == 0)
542		return;
543	for (mp = &sb->sb_mb; m = *mp; mp = &((*mp)->m_nextpkt)) {
544	    again:
545		switch (m->m_type) {
546
547		case MT_OOBDATA:
548			continue;		/* WANT next train */
549
550		case MT_CONTROL:
551			if (m = m->m_next)
552				goto again;	/* inspect THIS train further */
553		}
554		break;
555	}
556	/*
557	 * Put the first mbuf on the queue.
558	 * Note this permits zero length records.
559	 */
560	sballoc(sb, m0);
561	m0->m_nextpkt = *mp;
562	*mp = m0;
563	m = m0->m_next;
564	m0->m_next = 0;
565	if (m && (m0->m_flags & M_EOR)) {
566		m0->m_flags &= ~M_EOR;
567		m->m_flags |= M_EOR;
568	}
569	sbcompress(sb, m, m0);
570}
571
572/*
573 * Append address and data, and optionally, control (ancillary) data
574 * to the receive queue of a socket.  If present,
575 * m0 must include a packet header with total length.
576 * Returns 0 if no space in sockbuf or insufficient mbufs.
577 */
578int
579sbappendaddr(sb, asa, m0, control)
580	register struct sockbuf *sb;
581	struct sockaddr *asa;
582	struct mbuf *m0, *control;
583{
584	register struct mbuf *m, *n;
585	int space = asa->sa_len;
586
587if (m0 && (m0->m_flags & M_PKTHDR) == 0)
588panic("sbappendaddr");
589	if (m0)
590		space += m0->m_pkthdr.len;
591	for (n = control; n; n = n->m_next) {
592		space += n->m_len;
593		if (n->m_next == 0)	/* keep pointer to last control buf */
594			break;
595	}
596	if (space > sbspace(sb))
597		return (0);
598	if (asa->sa_len > MLEN)
599		return (0);
600	MGET(m, M_DONTWAIT, MT_SONAME);
601	if (m == 0)
602		return (0);
603	m->m_len = asa->sa_len;
604	bcopy((caddr_t)asa, mtod(m, caddr_t), asa->sa_len);
605	if (n)
606		n->m_next = m0;		/* concatenate data to control */
607	else
608		control = m0;
609	m->m_next = control;
610	for (n = m; n; n = n->m_next)
611		sballoc(sb, n);
612	if (n = sb->sb_mb) {
613		while (n->m_nextpkt)
614			n = n->m_nextpkt;
615		n->m_nextpkt = m;
616	} else
617		sb->sb_mb = m;
618	return (1);
619}
620
621int
622sbappendcontrol(sb, m0, control)
623	struct sockbuf *sb;
624	struct mbuf *m0, *control;
625{
626	register struct mbuf *m, *n;
627	int space = 0;
628
629	if (control == 0)
630		panic("sbappendcontrol");
631	for (m = control; ; m = m->m_next) {
632		space += m->m_len;
633		if (m->m_next == 0)
634			break;
635	}
636	n = m;			/* save pointer to last control buffer */
637	for (m = m0; m; m = m->m_next)
638		space += m->m_len;
639	if (space > sbspace(sb))
640		return (0);
641	n->m_next = m0;			/* concatenate data to control */
642	for (m = control; m; m = m->m_next)
643		sballoc(sb, m);
644	if (n = sb->sb_mb) {
645		while (n->m_nextpkt)
646			n = n->m_nextpkt;
647		n->m_nextpkt = control;
648	} else
649		sb->sb_mb = control;
650	return (1);
651}
652
653/*
654 * Compress mbuf chain m into the socket
655 * buffer sb following mbuf n.  If n
656 * is null, the buffer is presumed empty.
657 */
658void
659sbcompress(sb, m, n)
660	register struct sockbuf *sb;
661	register struct mbuf *m, *n;
662{
663	register int eor = 0;
664	register struct mbuf *o;
665
666	while (m) {
667		eor |= m->m_flags & M_EOR;
668		if (m->m_len == 0 &&
669		    (eor == 0 ||
670		     (((o = m->m_next) || (o = n)) &&
671		      o->m_type == m->m_type))) {
672			m = m_free(m);
673			continue;
674		}
675		if (n && (n->m_flags & (M_EXT | M_EOR)) == 0 &&
676		    (n->m_data + n->m_len + m->m_len) < &n->m_dat[MLEN] &&
677		    n->m_type == m->m_type) {
678			bcopy(mtod(m, caddr_t), mtod(n, caddr_t) + n->m_len,
679			    (unsigned)m->m_len);
680			n->m_len += m->m_len;
681			sb->sb_cc += m->m_len;
682			m = m_free(m);
683			continue;
684		}
685		if (n)
686			n->m_next = m;
687		else
688			sb->sb_mb = m;
689		sballoc(sb, m);
690		n = m;
691		m->m_flags &= ~M_EOR;
692		m = m->m_next;
693		n->m_next = 0;
694	}
695	if (eor) {
696		if (n)
697			n->m_flags |= eor;
698		else
699			printf("semi-panic: sbcompress\n");
700	}
701}
702
703/*
704 * Free all mbufs in a sockbuf.
705 * Check that all resources are reclaimed.
706 */
707void
708sbflush(sb)
709	register struct sockbuf *sb;
710{
711
712	if (sb->sb_flags & SB_LOCK)
713		panic("sbflush");
714	while (sb->sb_mbcnt)
715		sbdrop(sb, (int)sb->sb_cc);
716	if (sb->sb_cc || sb->sb_mb)
717		panic("sbflush 2");
718}
719
720/*
721 * Drop data from (the front of) a sockbuf.
722 */
723void
724sbdrop(sb, len)
725	register struct sockbuf *sb;
726	register int len;
727{
728	register struct mbuf *m, *mn;
729	struct mbuf *next;
730
731	next = (m = sb->sb_mb) ? m->m_nextpkt : 0;
732	while (len > 0) {
733		if (m == 0) {
734			if (next == 0)
735				panic("sbdrop");
736			m = next;
737			next = m->m_nextpkt;
738			continue;
739		}
740		if (m->m_len > len) {
741			m->m_len -= len;
742			m->m_data += len;
743			sb->sb_cc -= len;
744			break;
745		}
746		len -= m->m_len;
747		sbfree(sb, m);
748		MFREE(m, mn);
749		m = mn;
750	}
751	while (m && m->m_len == 0) {
752		sbfree(sb, m);
753		MFREE(m, mn);
754		m = mn;
755	}
756	if (m) {
757		sb->sb_mb = m;
758		m->m_nextpkt = next;
759	} else
760		sb->sb_mb = next;
761}
762
763/*
764 * Drop a record off the front of a sockbuf
765 * and move the next record to the front.
766 */
767void
768sbdroprecord(sb)
769	register struct sockbuf *sb;
770{
771	register struct mbuf *m, *mn;
772
773	m = sb->sb_mb;
774	if (m) {
775		sb->sb_mb = m->m_nextpkt;
776		do {
777			sbfree(sb, m);
778			MFREE(m, mn);
779		} while (m = mn);
780	}
781}
782