uipc_sockbuf.c revision 17096
1/*
2 * Copyright (c) 1982, 1986, 1988, 1990, 1993
3 *	The Regents of the University of California.  All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 *    notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 *    notice, this list of conditions and the following disclaimer in the
12 *    documentation and/or other materials provided with the distribution.
13 * 3. All advertising materials mentioning features or use of this software
14 *    must display the following acknowledgement:
15 *	This product includes software developed by the University of
16 *	California, Berkeley and its contributors.
17 * 4. Neither the name of the University nor the names of its contributors
18 *    may be used to endorse or promote products derived from this software
19 *    without specific prior written permission.
20 *
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * SUCH DAMAGE.
32 *
33 *	@(#)uipc_socket2.c	8.1 (Berkeley) 6/10/93
34 * $Id: uipc_socket2.c,v 1.10 1996/06/12 05:07:35 gpalmer Exp $
35 */
36
37#include <sys/param.h>
38#include <sys/systm.h>
39#include <sys/kernel.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/stat.h>
47#include <sys/socket.h>
48#include <sys/socketvar.h>
49#include <sys/signalvar.h>
50#include <sys/sysctl.h>
51
52/*
53 * Primitive routines for operating on sockets and socket buffers
54 */
55
56u_long	sb_max = SB_MAX;		/* XXX should be static */
57SYSCTL_INT(_kern, KERN_MAXSOCKBUF, maxsockbuf, CTLFLAG_RW, &sb_max, 0, "")
58
59static	u_long sb_efficiency = 8;	/* parameter for sbreserve() */
60SYSCTL_INT(_kern, OID_AUTO, sockbuf_waste_factor, CTLFLAG_RW, &sb_efficiency,
61	   0, "");
62
63/*
64 * Procedures to manipulate state flags of socket
65 * and do appropriate wakeups.  Normal sequence from the
66 * active (originating) side is that soisconnecting() is
67 * called during processing of connect() call,
68 * resulting in an eventual call to soisconnected() if/when the
69 * connection is established.  When the connection is torn down
70 * soisdisconnecting() is called during processing of disconnect() call,
71 * and soisdisconnected() is called when the connection to the peer
72 * is totally severed.  The semantics of these routines are such that
73 * connectionless protocols can call soisconnected() and soisdisconnected()
74 * only, bypassing the in-progress calls when setting up a ``connection''
75 * takes no time.
76 *
77 * From the passive side, a socket is created with
78 * two queues of sockets: so_q0 for connections in progress
79 * and so_q for connections already made and awaiting user acceptance.
80 * As a protocol is preparing incoming connections, it creates a socket
81 * structure queued on so_q0 by calling sonewconn().  When the connection
82 * is established, soisconnected() is called, and transfers the
83 * socket structure to so_q, making it available to accept().
84 *
85 * If a socket is closed with sockets on either
86 * so_q0 or so_q, these sockets are dropped.
87 *
88 * If higher level protocols are implemented in
89 * the kernel, the wakeups done here will sometimes
90 * cause software-interrupt process scheduling.
91 */
92
93void
94soisconnecting(so)
95	register struct socket *so;
96{
97
98	so->so_state &= ~(SS_ISCONNECTED|SS_ISDISCONNECTING);
99	so->so_state |= SS_ISCONNECTING;
100}
101
102void
103soisconnected(so)
104	register struct socket *so;
105{
106	register struct socket *head = so->so_head;
107
108	so->so_state &= ~(SS_ISCONNECTING|SS_ISDISCONNECTING|SS_ISCONFIRMING);
109	so->so_state |= SS_ISCONNECTED;
110	if (head && (so->so_state & SS_INCOMP)) {
111		TAILQ_REMOVE(&head->so_incomp, so, so_list);
112		so->so_state &= ~SS_INCOMP;
113		TAILQ_INSERT_TAIL(&head->so_comp, so, so_list);
114		so->so_state |= SS_COMP;
115		sorwakeup(head);
116		wakeup((caddr_t)&head->so_timeo);
117	} else {
118		wakeup((caddr_t)&so->so_timeo);
119		sorwakeup(so);
120		sowwakeup(so);
121	}
122}
123
124void
125soisdisconnecting(so)
126	register struct socket *so;
127{
128
129	so->so_state &= ~SS_ISCONNECTING;
130	so->so_state |= (SS_ISDISCONNECTING|SS_CANTRCVMORE|SS_CANTSENDMORE);
131	wakeup((caddr_t)&so->so_timeo);
132	sowwakeup(so);
133	sorwakeup(so);
134}
135
136void
137soisdisconnected(so)
138	register struct socket *so;
139{
140
141	so->so_state &= ~(SS_ISCONNECTING|SS_ISCONNECTED|SS_ISDISCONNECTING);
142	so->so_state |= (SS_CANTRCVMORE|SS_CANTSENDMORE);
143	wakeup((caddr_t)&so->so_timeo);
144	sowwakeup(so);
145	sorwakeup(so);
146}
147
148/*
149 * When an attempt at a new connection is noted on a socket
150 * which accepts connections, sonewconn is called.  If the
151 * connection is possible (subject to space constraints, etc.)
152 * then we allocate a new structure, propoerly linked into the
153 * data structure of the original socket, and return this.
154 * Connstatus may be 0, or SO_ISCONFIRMING, or SO_ISCONNECTED.
155 *
156 * Currently, sonewconn() is defined as sonewconn1() in socketvar.h
157 * to catch calls that are missing the (new) second parameter.
158 */
159struct socket *
160sonewconn1(head, connstatus)
161	register struct socket *head;
162	int connstatus;
163{
164	register struct socket *so;
165
166	if (head->so_qlen > 3 * head->so_qlimit / 2)
167		return ((struct socket *)0);
168	MALLOC(so, struct socket *, sizeof(*so), M_SOCKET, M_DONTWAIT);
169	if (so == NULL)
170		return ((struct socket *)0);
171	bzero((caddr_t)so, sizeof(*so));
172	so->so_head = head;
173	so->so_type = head->so_type;
174	so->so_options = head->so_options &~ SO_ACCEPTCONN;
175	so->so_linger = head->so_linger;
176	so->so_state = head->so_state | SS_NOFDREF;
177	so->so_proto = head->so_proto;
178	so->so_timeo = head->so_timeo;
179	so->so_pgid = head->so_pgid;
180	(void) soreserve(so, head->so_snd.sb_hiwat, head->so_rcv.sb_hiwat);
181	if (connstatus) {
182		TAILQ_INSERT_TAIL(&head->so_comp, so, so_list);
183		so->so_state |= SS_COMP;
184	} else {
185		TAILQ_INSERT_TAIL(&head->so_incomp, so, so_list);
186		so->so_state |= SS_INCOMP;
187	}
188	head->so_qlen++;
189	if ((*so->so_proto->pr_usrreqs->pru_attach)(so, 0)) {
190		if (so->so_state & SS_COMP) {
191			TAILQ_REMOVE(&head->so_comp, so, so_list);
192		} else {
193			TAILQ_REMOVE(&head->so_incomp, so, so_list);
194		}
195		head->so_qlen--;
196		(void) free((caddr_t)so, M_SOCKET);
197		return ((struct socket *)0);
198	}
199	if (connstatus) {
200		sorwakeup(head);
201		wakeup((caddr_t)&head->so_timeo);
202		so->so_state |= connstatus;
203	}
204	return (so);
205}
206
207/*
208 * Socantsendmore indicates that no more data will be sent on the
209 * socket; it would normally be applied to a socket when the user
210 * informs the system that no more data is to be sent, by the protocol
211 * code (in case PRU_SHUTDOWN).  Socantrcvmore indicates that no more data
212 * will be received, and will normally be applied to the socket by a
213 * protocol when it detects that the peer will send no more data.
214 * Data queued for reading in the socket may yet be read.
215 */
216
217void
218socantsendmore(so)
219	struct socket *so;
220{
221
222	so->so_state |= SS_CANTSENDMORE;
223	sowwakeup(so);
224}
225
226void
227socantrcvmore(so)
228	struct socket *so;
229{
230
231	so->so_state |= SS_CANTRCVMORE;
232	sorwakeup(so);
233}
234
235/*
236 * Wait for data to arrive at/drain from a socket buffer.
237 */
238int
239sbwait(sb)
240	struct sockbuf *sb;
241{
242
243	sb->sb_flags |= SB_WAIT;
244	return (tsleep((caddr_t)&sb->sb_cc,
245	    (sb->sb_flags & SB_NOINTR) ? PSOCK : PSOCK | PCATCH, "sbwait",
246	    sb->sb_timeo));
247}
248
249/*
250 * Lock a sockbuf already known to be locked;
251 * return any error returned from sleep (EINTR).
252 */
253int
254sb_lock(sb)
255	register struct sockbuf *sb;
256{
257	int error;
258
259	while (sb->sb_flags & SB_LOCK) {
260		sb->sb_flags |= SB_WANT;
261		error = tsleep((caddr_t)&sb->sb_flags,
262		    (sb->sb_flags & SB_NOINTR) ? PSOCK : PSOCK|PCATCH,
263		    "sblock", 0);
264		if (error)
265			return (error);
266	}
267	sb->sb_flags |= SB_LOCK;
268	return (0);
269}
270
271/*
272 * Wakeup processes waiting on a socket buffer.
273 * Do asynchronous notification via SIGIO
274 * if the socket has the SS_ASYNC flag set.
275 */
276void
277sowakeup(so, sb)
278	register struct socket *so;
279	register struct sockbuf *sb;
280{
281	struct proc *p;
282
283	selwakeup(&sb->sb_sel);
284	sb->sb_flags &= ~SB_SEL;
285	if (sb->sb_flags & SB_WAIT) {
286		sb->sb_flags &= ~SB_WAIT;
287		wakeup((caddr_t)&sb->sb_cc);
288	}
289	if (so->so_state & SS_ASYNC) {
290		if (so->so_pgid < 0)
291			gsignal(-so->so_pgid, SIGIO);
292		else if (so->so_pgid > 0 && (p = pfind(so->so_pgid)) != 0)
293			psignal(p, SIGIO);
294	}
295}
296
297/*
298 * Socket buffer (struct sockbuf) utility routines.
299 *
300 * Each socket contains two socket buffers: one for sending data and
301 * one for receiving data.  Each buffer contains a queue of mbufs,
302 * information about the number of mbufs and amount of data in the
303 * queue, and other fields allowing select() statements and notification
304 * on data availability to be implemented.
305 *
306 * Data stored in a socket buffer is maintained as a list of records.
307 * Each record is a list of mbufs chained together with the m_next
308 * field.  Records are chained together with the m_nextpkt field. The upper
309 * level routine soreceive() expects the following conventions to be
310 * observed when placing information in the receive buffer:
311 *
312 * 1. If the protocol requires each message be preceded by the sender's
313 *    name, then a record containing that name must be present before
314 *    any associated data (mbuf's must be of type MT_SONAME).
315 * 2. If the protocol supports the exchange of ``access rights'' (really
316 *    just additional data associated with the message), and there are
317 *    ``rights'' to be received, then a record containing this data
318 *    should be present (mbuf's must be of type MT_RIGHTS).
319 * 3. If a name or rights record exists, then it must be followed by
320 *    a data record, perhaps of zero length.
321 *
322 * Before using a new socket structure it is first necessary to reserve
323 * buffer space to the socket, by calling sbreserve().  This should commit
324 * some of the available buffer space in the system buffer pool for the
325 * socket (currently, it does nothing but enforce limits).  The space
326 * should be released by calling sbrelease() when the socket is destroyed.
327 */
328
329int
330soreserve(so, sndcc, rcvcc)
331	register struct socket *so;
332	u_long sndcc, rcvcc;
333{
334
335	if (sbreserve(&so->so_snd, sndcc) == 0)
336		goto bad;
337	if (sbreserve(&so->so_rcv, rcvcc) == 0)
338		goto bad2;
339	if (so->so_rcv.sb_lowat == 0)
340		so->so_rcv.sb_lowat = 1;
341	if (so->so_snd.sb_lowat == 0)
342		so->so_snd.sb_lowat = MCLBYTES;
343	if (so->so_snd.sb_lowat > so->so_snd.sb_hiwat)
344		so->so_snd.sb_lowat = so->so_snd.sb_hiwat;
345	return (0);
346bad2:
347	sbrelease(&so->so_snd);
348bad:
349	return (ENOBUFS);
350}
351
352/*
353 * Allot mbufs to a sockbuf.
354 * Attempt to scale mbmax so that mbcnt doesn't become limiting
355 * if buffering efficiency is near the normal case.
356 */
357int
358sbreserve(sb, cc)
359	struct sockbuf *sb;
360	u_long cc;
361{
362
363	if (cc > sb_max * MCLBYTES / (MSIZE + MCLBYTES))
364		return (0);
365	sb->sb_hiwat = cc;
366	sb->sb_mbmax = min(cc * sb_efficiency, sb_max);
367	if (sb->sb_lowat > sb->sb_hiwat)
368		sb->sb_lowat = sb->sb_hiwat;
369	return (1);
370}
371
372/*
373 * Free mbufs held by a socket, and reserved mbuf space.
374 */
375void
376sbrelease(sb)
377	struct sockbuf *sb;
378{
379
380	sbflush(sb);
381	sb->sb_hiwat = sb->sb_mbmax = 0;
382}
383
384/*
385 * Routines to add and remove
386 * data from an mbuf queue.
387 *
388 * The routines sbappend() or sbappendrecord() are normally called to
389 * append new mbufs to a socket buffer, after checking that adequate
390 * space is available, comparing the function sbspace() with the amount
391 * of data to be added.  sbappendrecord() differs from sbappend() in
392 * that data supplied is treated as the beginning of a new record.
393 * To place a sender's address, optional access rights, and data in a
394 * socket receive buffer, sbappendaddr() should be used.  To place
395 * access rights and data in a socket receive buffer, sbappendrights()
396 * should be used.  In either case, the new data begins a new record.
397 * Note that unlike sbappend() and sbappendrecord(), these routines check
398 * for the caller that there will be enough space to store the data.
399 * Each fails if there is not enough space, or if it cannot find mbufs
400 * to store additional information in.
401 *
402 * Reliable protocols may use the socket send buffer to hold data
403 * awaiting acknowledgement.  Data is normally copied from a socket
404 * send buffer in a protocol with m_copy for output to a peer,
405 * and then removing the data from the socket buffer with sbdrop()
406 * or sbdroprecord() when the data is acknowledged by the peer.
407 */
408
409/*
410 * Append mbuf chain m to the last record in the
411 * socket buffer sb.  The additional space associated
412 * the mbuf chain is recorded in sb.  Empty mbufs are
413 * discarded and mbufs are compacted where possible.
414 */
415void
416sbappend(sb, m)
417	struct sockbuf *sb;
418	struct mbuf *m;
419{
420	register struct mbuf *n;
421
422	if (m == 0)
423		return;
424	n = sb->sb_mb;
425	if (n) {
426		while (n->m_nextpkt)
427			n = n->m_nextpkt;
428		do {
429			if (n->m_flags & M_EOR) {
430				sbappendrecord(sb, m); /* XXXXXX!!!! */
431				return;
432			}
433		} while (n->m_next && (n = n->m_next));
434	}
435	sbcompress(sb, m, n);
436}
437
438#ifdef SOCKBUF_DEBUG
439void
440sbcheck(sb)
441	register struct sockbuf *sb;
442{
443	register struct mbuf *m;
444	register int len = 0, mbcnt = 0;
445
446	for (m = sb->sb_mb; m; m = m->m_next) {
447		len += m->m_len;
448		mbcnt += MSIZE;
449		if (m->m_flags & M_EXT)
450			mbcnt += m->m_ext.ext_size;
451		if (m->m_nextpkt)
452			panic("sbcheck nextpkt");
453	}
454	if (len != sb->sb_cc || mbcnt != sb->sb_mbcnt) {
455		printf("cc %d != %d || mbcnt %d != %d\n", len, sb->sb_cc,
456		    mbcnt, sb->sb_mbcnt);
457		panic("sbcheck");
458	}
459}
460#endif
461
462/*
463 * As above, except the mbuf chain
464 * begins a new record.
465 */
466void
467sbappendrecord(sb, m0)
468	register struct sockbuf *sb;
469	register struct mbuf *m0;
470{
471	register struct mbuf *m;
472
473	if (m0 == 0)
474		return;
475	m = sb->sb_mb;
476	if (m)
477		while (m->m_nextpkt)
478			m = m->m_nextpkt;
479	/*
480	 * Put the first mbuf on the queue.
481	 * Note this permits zero length records.
482	 */
483	sballoc(sb, m0);
484	if (m)
485		m->m_nextpkt = m0;
486	else
487		sb->sb_mb = m0;
488	m = m0->m_next;
489	m0->m_next = 0;
490	if (m && (m0->m_flags & M_EOR)) {
491		m0->m_flags &= ~M_EOR;
492		m->m_flags |= M_EOR;
493	}
494	sbcompress(sb, m, m0);
495}
496
497/*
498 * As above except that OOB data
499 * is inserted at the beginning of the sockbuf,
500 * but after any other OOB data.
501 */
502void
503sbinsertoob(sb, m0)
504	register struct sockbuf *sb;
505	register struct mbuf *m0;
506{
507	register struct mbuf *m;
508	register struct mbuf **mp;
509
510	if (m0 == 0)
511		return;
512	for (mp = &sb->sb_mb; *mp ; mp = &((*mp)->m_nextpkt)) {
513	    m = *mp;
514	    again:
515		switch (m->m_type) {
516
517		case MT_OOBDATA:
518			continue;		/* WANT next train */
519
520		case MT_CONTROL:
521			m = m->m_next;
522			if (m)
523				goto again;	/* inspect THIS train further */
524		}
525		break;
526	}
527	/*
528	 * Put the first mbuf on the queue.
529	 * Note this permits zero length records.
530	 */
531	sballoc(sb, m0);
532	m0->m_nextpkt = *mp;
533	*mp = m0;
534	m = m0->m_next;
535	m0->m_next = 0;
536	if (m && (m0->m_flags & M_EOR)) {
537		m0->m_flags &= ~M_EOR;
538		m->m_flags |= M_EOR;
539	}
540	sbcompress(sb, m, m0);
541}
542
543/*
544 * Append address and data, and optionally, control (ancillary) data
545 * to the receive queue of a socket.  If present,
546 * m0 must include a packet header with total length.
547 * Returns 0 if no space in sockbuf or insufficient mbufs.
548 */
549int
550sbappendaddr(sb, asa, m0, control)
551	register struct sockbuf *sb;
552	struct sockaddr *asa;
553	struct mbuf *m0, *control;
554{
555	register struct mbuf *m, *n;
556	int space = asa->sa_len;
557
558if (m0 && (m0->m_flags & M_PKTHDR) == 0)
559panic("sbappendaddr");
560	if (m0)
561		space += m0->m_pkthdr.len;
562	for (n = control; n; n = n->m_next) {
563		space += n->m_len;
564		if (n->m_next == 0)	/* keep pointer to last control buf */
565			break;
566	}
567	if (space > sbspace(sb))
568		return (0);
569	if (asa->sa_len > MLEN)
570		return (0);
571	MGET(m, M_DONTWAIT, MT_SONAME);
572	if (m == 0)
573		return (0);
574	m->m_len = asa->sa_len;
575	bcopy((caddr_t)asa, mtod(m, caddr_t), asa->sa_len);
576	if (n)
577		n->m_next = m0;		/* concatenate data to control */
578	else
579		control = m0;
580	m->m_next = control;
581	for (n = m; n; n = n->m_next)
582		sballoc(sb, n);
583	n = sb->sb_mb;
584	if (n) {
585		while (n->m_nextpkt)
586			n = n->m_nextpkt;
587		n->m_nextpkt = m;
588	} else
589		sb->sb_mb = m;
590	return (1);
591}
592
593int
594sbappendcontrol(sb, m0, control)
595	struct sockbuf *sb;
596	struct mbuf *control, *m0;
597{
598	register struct mbuf *m, *n;
599	int space = 0;
600
601	if (control == 0)
602		panic("sbappendcontrol");
603	for (m = control; ; m = m->m_next) {
604		space += m->m_len;
605		if (m->m_next == 0)
606			break;
607	}
608	n = m;			/* save pointer to last control buffer */
609	for (m = m0; m; m = m->m_next)
610		space += m->m_len;
611	if (space > sbspace(sb))
612		return (0);
613	n->m_next = m0;			/* concatenate data to control */
614	for (m = control; m; m = m->m_next)
615		sballoc(sb, m);
616	n = sb->sb_mb;
617	if (n) {
618		while (n->m_nextpkt)
619			n = n->m_nextpkt;
620		n->m_nextpkt = control;
621	} else
622		sb->sb_mb = control;
623	return (1);
624}
625
626/*
627 * Compress mbuf chain m into the socket
628 * buffer sb following mbuf n.  If n
629 * is null, the buffer is presumed empty.
630 */
631void
632sbcompress(sb, m, n)
633	register struct sockbuf *sb;
634	register struct mbuf *m, *n;
635{
636	register int eor = 0;
637	register struct mbuf *o;
638
639	while (m) {
640		eor |= m->m_flags & M_EOR;
641		if (m->m_len == 0 &&
642		    (eor == 0 ||
643		     (((o = m->m_next) || (o = n)) &&
644		      o->m_type == m->m_type))) {
645			m = m_free(m);
646			continue;
647		}
648		if (n && (n->m_flags & (M_EXT | M_EOR)) == 0 &&
649		    (n->m_data + n->m_len + m->m_len) < &n->m_dat[MLEN] &&
650		    n->m_type == m->m_type) {
651			bcopy(mtod(m, caddr_t), mtod(n, caddr_t) + n->m_len,
652			    (unsigned)m->m_len);
653			n->m_len += m->m_len;
654			sb->sb_cc += m->m_len;
655			m = m_free(m);
656			continue;
657		}
658		if (n)
659			n->m_next = m;
660		else
661			sb->sb_mb = m;
662		sballoc(sb, m);
663		n = m;
664		m->m_flags &= ~M_EOR;
665		m = m->m_next;
666		n->m_next = 0;
667	}
668	if (eor) {
669		if (n)
670			n->m_flags |= eor;
671		else
672			printf("semi-panic: sbcompress\n");
673	}
674}
675
676/*
677 * Free all mbufs in a sockbuf.
678 * Check that all resources are reclaimed.
679 */
680void
681sbflush(sb)
682	register struct sockbuf *sb;
683{
684
685	if (sb->sb_flags & SB_LOCK)
686		panic("sbflush");
687	while (sb->sb_mbcnt)
688		sbdrop(sb, (int)sb->sb_cc);
689	if (sb->sb_cc || sb->sb_mb)
690		panic("sbflush 2");
691}
692
693/*
694 * Drop data from (the front of) a sockbuf.
695 */
696void
697sbdrop(sb, len)
698	register struct sockbuf *sb;
699	register int len;
700{
701	register struct mbuf *m, *mn;
702	struct mbuf *next;
703
704	next = (m = sb->sb_mb) ? m->m_nextpkt : 0;
705	while (len > 0) {
706		if (m == 0) {
707			if (next == 0)
708				panic("sbdrop");
709			m = next;
710			next = m->m_nextpkt;
711			continue;
712		}
713		if (m->m_len > len) {
714			m->m_len -= len;
715			m->m_data += len;
716			sb->sb_cc -= len;
717			break;
718		}
719		len -= m->m_len;
720		sbfree(sb, m);
721		MFREE(m, mn);
722		m = mn;
723	}
724	while (m && m->m_len == 0) {
725		sbfree(sb, m);
726		MFREE(m, mn);
727		m = mn;
728	}
729	if (m) {
730		sb->sb_mb = m;
731		m->m_nextpkt = next;
732	} else
733		sb->sb_mb = next;
734}
735
736/*
737 * Drop a record off the front of a sockbuf
738 * and move the next record to the front.
739 */
740void
741sbdroprecord(sb)
742	register struct sockbuf *sb;
743{
744	register struct mbuf *m, *mn;
745
746	m = sb->sb_mb;
747	if (m) {
748		sb->sb_mb = m->m_nextpkt;
749		do {
750			sbfree(sb, m);
751			MFREE(m, mn);
752			m = mn;
753		} while (m);
754	}
755}
756
757#ifdef PRU_OLDSTYLE
758/*
759 * The following routines mediate between the old-style `pr_usrreq'
760 * protocol implementations and the new-style `struct pr_usrreqs'
761 * calling convention.
762 */
763
764/* syntactic sugar */
765#define	nomb	(struct mbuf *)0
766
767static int
768old_abort(struct socket *so)
769{
770	return so->so_proto->pr_ousrreq(so, PRU_ABORT, nomb, nomb, nomb);
771}
772
773static int
774old_accept(struct socket *so, struct mbuf *nam)
775{
776	return so->so_proto->pr_ousrreq(so, PRU_ACCEPT, nomb,  nam, nomb);
777}
778
779static int
780old_attach(struct socket *so, int proto)
781{
782	return so->so_proto->pr_ousrreq(so, PRU_ATTACH, nomb,
783				       (struct mbuf *)proto, /* XXX */
784				       nomb);
785}
786
787static int
788old_bind(struct socket *so, struct mbuf *nam)
789{
790	return so->so_proto->pr_ousrreq(so, PRU_BIND, nomb, nam, nomb);
791}
792
793static int
794old_connect(struct socket *so, struct mbuf *nam)
795{
796	return so->so_proto->pr_ousrreq(so, PRU_CONNECT, nomb, nam, nomb);
797}
798
799static int
800old_connect2(struct socket *so1, struct socket *so2)
801{
802	return so1->so_proto->pr_ousrreq(so1, PRU_CONNECT2, nomb,
803				       (struct mbuf *)so2, nomb);
804}
805
806static int
807old_control(struct socket *so, int cmd, caddr_t data, struct ifnet *ifp)
808{
809	return so->so_proto->pr_ousrreq(so, PRU_CONTROL, (struct mbuf *)cmd,
810				       (struct mbuf *)data,
811				       (struct mbuf *)ifp);
812}
813
814static int
815old_detach(struct socket *so)
816{
817	return so->so_proto->pr_ousrreq(so, PRU_DETACH, nomb, nomb, nomb);
818}
819
820static int
821old_disconnect(struct socket *so)
822{
823	return so->so_proto->pr_ousrreq(so, PRU_DISCONNECT, nomb, nomb, nomb);
824}
825
826static int
827old_listen(struct socket *so)
828{
829	return so->so_proto->pr_ousrreq(so, PRU_LISTEN, nomb, nomb, nomb);
830}
831
832static int
833old_peeraddr(struct socket *so, struct mbuf *nam)
834{
835	return so->so_proto->pr_ousrreq(so, PRU_PEERADDR, nomb, nam, nomb);
836}
837
838static int
839old_rcvd(struct socket *so, int flags)
840{
841	return so->so_proto->pr_ousrreq(so, PRU_RCVD, nomb,
842				       (struct mbuf *)flags, /* XXX */
843				       nomb);
844}
845
846static int
847old_rcvoob(struct socket *so, struct mbuf *m, int flags)
848{
849	return so->so_proto->pr_ousrreq(so, PRU_RCVOOB, m,
850				       (struct mbuf *)flags, /* XXX */
851				       nomb);
852}
853
854static int
855old_send(struct socket *so, int flags, struct mbuf *m, struct mbuf *addr,
856	 struct mbuf *control)
857{
858	int req;
859
860	if (flags & PRUS_OOB) {
861		req = PRU_SENDOOB;
862	} else if(flags & PRUS_EOF) {
863		req = PRU_SEND_EOF;
864	} else {
865		req = PRU_SEND;
866	}
867	return so->so_proto->pr_ousrreq(so, req, m, addr, control);
868}
869
870static int
871old_sense(struct socket *so, struct stat *sb)
872{
873	return so->so_proto->pr_ousrreq(so, PRU_SENSE, (struct mbuf *)sb,
874				       nomb, nomb);
875}
876
877static int
878old_shutdown(struct socket *so)
879{
880	return so->so_proto->pr_ousrreq(so, PRU_SHUTDOWN, nomb, nomb, nomb);
881}
882
883static int
884old_sockaddr(struct socket *so, struct mbuf *nam)
885{
886	return so->so_proto->pr_ousrreq(so, PRU_SOCKADDR, nomb, nam, nomb);
887}
888
889struct pr_usrreqs pru_oldstyle = {
890	old_abort, old_accept, old_attach, old_bind, old_connect,
891	old_connect2, old_control, old_detach, old_disconnect,
892	old_listen, old_peeraddr, old_rcvd, old_rcvoob, old_send,
893	old_sense, old_shutdown, old_sockaddr
894};
895
896#endif /* PRU_OLDSTYLE */
897
898/*
899 * Some routines that return EOPNOTSUPP for entry points that are not
900 * supported by a protocol.  Fill in as needed.
901 */
902int
903pru_connect2_notsupp(struct socket *so1, struct socket *so2)
904{
905	return EOPNOTSUPP;
906}
907