uipc_sockbuf.c revision 17675
11541Srgrimes/*
21541Srgrimes * Copyright (c) 1982, 1986, 1988, 1990, 1993
31541Srgrimes *	The Regents of the University of California.  All rights reserved.
41541Srgrimes *
51541Srgrimes * Redistribution and use in source and binary forms, with or without
61541Srgrimes * modification, are permitted provided that the following conditions
71541Srgrimes * are met:
81541Srgrimes * 1. Redistributions of source code must retain the above copyright
91541Srgrimes *    notice, this list of conditions and the following disclaimer.
101541Srgrimes * 2. Redistributions in binary form must reproduce the above copyright
111541Srgrimes *    notice, this list of conditions and the following disclaimer in the
121541Srgrimes *    documentation and/or other materials provided with the distribution.
131541Srgrimes * 3. All advertising materials mentioning features or use of this software
141541Srgrimes *    must display the following acknowledgement:
151541Srgrimes *	This product includes software developed by the University of
161541Srgrimes *	California, Berkeley and its contributors.
171541Srgrimes * 4. Neither the name of the University nor the names of its contributors
181541Srgrimes *    may be used to endorse or promote products derived from this software
191541Srgrimes *    without specific prior written permission.
201541Srgrimes *
211541Srgrimes * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
221541Srgrimes * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
231541Srgrimes * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
241541Srgrimes * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
251541Srgrimes * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
261541Srgrimes * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
271541Srgrimes * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
281541Srgrimes * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
291541Srgrimes * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
301541Srgrimes * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
311541Srgrimes * SUCH DAMAGE.
321541Srgrimes *
331541Srgrimes *	@(#)uipc_socket2.c	8.1 (Berkeley) 6/10/93
3417675Sjulian * $Id: uipc_socket2.c,v 1.12 1996/07/11 16:31:59 wollman Exp $
351541Srgrimes */
361541Srgrimes
371541Srgrimes#include <sys/param.h>
381541Srgrimes#include <sys/systm.h>
3912041Swollman#include <sys/kernel.h>
401541Srgrimes#include <sys/proc.h>
411541Srgrimes#include <sys/file.h>
421541Srgrimes#include <sys/buf.h>
431541Srgrimes#include <sys/malloc.h>
441541Srgrimes#include <sys/mbuf.h>
451541Srgrimes#include <sys/protosw.h>
463308Sphk#include <sys/stat.h>
471541Srgrimes#include <sys/socket.h>
481541Srgrimes#include <sys/socketvar.h>
493308Sphk#include <sys/signalvar.h>
5012041Swollman#include <sys/sysctl.h>
511541Srgrimes
521541Srgrimes/*
531541Srgrimes * Primitive routines for operating on sockets and socket buffers
541541Srgrimes */
551541Srgrimes
5612041Swollmanu_long	sb_max = SB_MAX;		/* XXX should be static */
5712041SwollmanSYSCTL_INT(_kern, KERN_MAXSOCKBUF, maxsockbuf, CTLFLAG_RW, &sb_max, 0, "")
581541Srgrimes
5913267Swollmanstatic	u_long sb_efficiency = 8;	/* parameter for sbreserve() */
6013267SwollmanSYSCTL_INT(_kern, OID_AUTO, sockbuf_waste_factor, CTLFLAG_RW, &sb_efficiency,
6113267Swollman	   0, "");
6213267Swollman
631541Srgrimes/*
641541Srgrimes * Procedures to manipulate state flags of socket
651541Srgrimes * and do appropriate wakeups.  Normal sequence from the
661541Srgrimes * active (originating) side is that soisconnecting() is
671541Srgrimes * called during processing of connect() call,
681541Srgrimes * resulting in an eventual call to soisconnected() if/when the
691541Srgrimes * connection is established.  When the connection is torn down
701541Srgrimes * soisdisconnecting() is called during processing of disconnect() call,
711541Srgrimes * and soisdisconnected() is called when the connection to the peer
721541Srgrimes * is totally severed.  The semantics of these routines are such that
731541Srgrimes * connectionless protocols can call soisconnected() and soisdisconnected()
741541Srgrimes * only, bypassing the in-progress calls when setting up a ``connection''
751541Srgrimes * takes no time.
761541Srgrimes *
771541Srgrimes * From the passive side, a socket is created with
781541Srgrimes * two queues of sockets: so_q0 for connections in progress
791541Srgrimes * and so_q for connections already made and awaiting user acceptance.
801541Srgrimes * As a protocol is preparing incoming connections, it creates a socket
811541Srgrimes * structure queued on so_q0 by calling sonewconn().  When the connection
821541Srgrimes * is established, soisconnected() is called, and transfers the
831541Srgrimes * socket structure to so_q, making it available to accept().
848876Srgrimes *
851541Srgrimes * If a socket is closed with sockets on either
861541Srgrimes * so_q0 or so_q, these sockets are dropped.
871541Srgrimes *
881541Srgrimes * If higher level protocols are implemented in
891541Srgrimes * the kernel, the wakeups done here will sometimes
901541Srgrimes * cause software-interrupt process scheduling.
911541Srgrimes */
921541Srgrimes
931549Srgrimesvoid
941541Srgrimessoisconnecting(so)
951541Srgrimes	register struct socket *so;
961541Srgrimes{
971541Srgrimes
981541Srgrimes	so->so_state &= ~(SS_ISCONNECTED|SS_ISDISCONNECTING);
991541Srgrimes	so->so_state |= SS_ISCONNECTING;
1001541Srgrimes}
1011541Srgrimes
1021549Srgrimesvoid
1031541Srgrimessoisconnected(so)
1041541Srgrimes	register struct socket *so;
1051541Srgrimes{
1061541Srgrimes	register struct socket *head = so->so_head;
1071541Srgrimes
1081541Srgrimes	so->so_state &= ~(SS_ISCONNECTING|SS_ISDISCONNECTING|SS_ISCONFIRMING);
1091541Srgrimes	so->so_state |= SS_ISCONNECTED;
11014547Sdg	if (head && (so->so_state & SS_INCOMP)) {
11114547Sdg		TAILQ_REMOVE(&head->so_incomp, so, so_list);
11214547Sdg		so->so_state &= ~SS_INCOMP;
11314547Sdg		TAILQ_INSERT_TAIL(&head->so_comp, so, so_list);
11414547Sdg		so->so_state |= SS_COMP;
1151541Srgrimes		sorwakeup(head);
1161541Srgrimes		wakeup((caddr_t)&head->so_timeo);
1171541Srgrimes	} else {
1181541Srgrimes		wakeup((caddr_t)&so->so_timeo);
1191541Srgrimes		sorwakeup(so);
1201541Srgrimes		sowwakeup(so);
1211541Srgrimes	}
1221541Srgrimes}
1231541Srgrimes
1241549Srgrimesvoid
1251541Srgrimessoisdisconnecting(so)
1261541Srgrimes	register struct socket *so;
1271541Srgrimes{
1281541Srgrimes
1291541Srgrimes	so->so_state &= ~SS_ISCONNECTING;
1301541Srgrimes	so->so_state |= (SS_ISDISCONNECTING|SS_CANTRCVMORE|SS_CANTSENDMORE);
1311541Srgrimes	wakeup((caddr_t)&so->so_timeo);
1321541Srgrimes	sowwakeup(so);
1331541Srgrimes	sorwakeup(so);
1341541Srgrimes}
1351541Srgrimes
1361549Srgrimesvoid
1371541Srgrimessoisdisconnected(so)
1381541Srgrimes	register struct socket *so;
1391541Srgrimes{
1401541Srgrimes
1411541Srgrimes	so->so_state &= ~(SS_ISCONNECTING|SS_ISCONNECTED|SS_ISDISCONNECTING);
1421541Srgrimes	so->so_state |= (SS_CANTRCVMORE|SS_CANTSENDMORE);
1431541Srgrimes	wakeup((caddr_t)&so->so_timeo);
1441541Srgrimes	sowwakeup(so);
1451541Srgrimes	sorwakeup(so);
1461541Srgrimes}
1471541Srgrimes
1481541Srgrimes/*
1491541Srgrimes * When an attempt at a new connection is noted on a socket
1501541Srgrimes * which accepts connections, sonewconn is called.  If the
1511541Srgrimes * connection is possible (subject to space constraints, etc.)
1521541Srgrimes * then we allocate a new structure, propoerly linked into the
1531541Srgrimes * data structure of the original socket, and return this.
1541541Srgrimes * Connstatus may be 0, or SO_ISCONFIRMING, or SO_ISCONNECTED.
1551541Srgrimes *
1561541Srgrimes * Currently, sonewconn() is defined as sonewconn1() in socketvar.h
1571541Srgrimes * to catch calls that are missing the (new) second parameter.
1581541Srgrimes */
1591541Srgrimesstruct socket *
1601541Srgrimessonewconn1(head, connstatus)
1611541Srgrimes	register struct socket *head;
1621541Srgrimes	int connstatus;
1631541Srgrimes{
1641541Srgrimes	register struct socket *so;
1651541Srgrimes
16614547Sdg	if (head->so_qlen > 3 * head->so_qlimit / 2)
1671541Srgrimes		return ((struct socket *)0);
1681541Srgrimes	MALLOC(so, struct socket *, sizeof(*so), M_SOCKET, M_DONTWAIT);
1698876Srgrimes	if (so == NULL)
1701541Srgrimes		return ((struct socket *)0);
1711541Srgrimes	bzero((caddr_t)so, sizeof(*so));
17214547Sdg	so->so_head = head;
1731541Srgrimes	so->so_type = head->so_type;
1741541Srgrimes	so->so_options = head->so_options &~ SO_ACCEPTCONN;
1751541Srgrimes	so->so_linger = head->so_linger;
1761541Srgrimes	so->so_state = head->so_state | SS_NOFDREF;
1771541Srgrimes	so->so_proto = head->so_proto;
1781541Srgrimes	so->so_timeo = head->so_timeo;
1791541Srgrimes	so->so_pgid = head->so_pgid;
1801541Srgrimes	(void) soreserve(so, head->so_snd.sb_hiwat, head->so_rcv.sb_hiwat);
18114547Sdg	if (connstatus) {
18214547Sdg		TAILQ_INSERT_TAIL(&head->so_comp, so, so_list);
18314547Sdg		so->so_state |= SS_COMP;
18414547Sdg	} else {
18514547Sdg		TAILQ_INSERT_TAIL(&head->so_incomp, so, so_list);
18614547Sdg		so->so_state |= SS_INCOMP;
18714547Sdg	}
18814547Sdg	head->so_qlen++;
18917096Swollman	if ((*so->so_proto->pr_usrreqs->pru_attach)(so, 0)) {
19014547Sdg		if (so->so_state & SS_COMP) {
19114547Sdg			TAILQ_REMOVE(&head->so_comp, so, so_list);
19214547Sdg		} else {
19314547Sdg			TAILQ_REMOVE(&head->so_incomp, so, so_list);
19414547Sdg		}
19514547Sdg		head->so_qlen--;
1961541Srgrimes		(void) free((caddr_t)so, M_SOCKET);
1971541Srgrimes		return ((struct socket *)0);
1981541Srgrimes	}
1991541Srgrimes	if (connstatus) {
2001541Srgrimes		sorwakeup(head);
2011541Srgrimes		wakeup((caddr_t)&head->so_timeo);
2021541Srgrimes		so->so_state |= connstatus;
2031541Srgrimes	}
2041541Srgrimes	return (so);
2051541Srgrimes}
2061541Srgrimes
2071541Srgrimes/*
2081541Srgrimes * Socantsendmore indicates that no more data will be sent on the
2091541Srgrimes * socket; it would normally be applied to a socket when the user
2101541Srgrimes * informs the system that no more data is to be sent, by the protocol
2111541Srgrimes * code (in case PRU_SHUTDOWN).  Socantrcvmore indicates that no more data
2121541Srgrimes * will be received, and will normally be applied to the socket by a
2131541Srgrimes * protocol when it detects that the peer will send no more data.
2141541Srgrimes * Data queued for reading in the socket may yet be read.
2151541Srgrimes */
2161541Srgrimes
2171549Srgrimesvoid
2181541Srgrimessocantsendmore(so)
2191541Srgrimes	struct socket *so;
2201541Srgrimes{
2211541Srgrimes
2221541Srgrimes	so->so_state |= SS_CANTSENDMORE;
2231541Srgrimes	sowwakeup(so);
2241541Srgrimes}
2251541Srgrimes
2261549Srgrimesvoid
2271541Srgrimessocantrcvmore(so)
2281541Srgrimes	struct socket *so;
2291541Srgrimes{
2301541Srgrimes
2311541Srgrimes	so->so_state |= SS_CANTRCVMORE;
2321541Srgrimes	sorwakeup(so);
2331541Srgrimes}
2341541Srgrimes
2351541Srgrimes/*
2361541Srgrimes * Wait for data to arrive at/drain from a socket buffer.
2371541Srgrimes */
2381549Srgrimesint
2391541Srgrimessbwait(sb)
2401541Srgrimes	struct sockbuf *sb;
2411541Srgrimes{
2421541Srgrimes
2431541Srgrimes	sb->sb_flags |= SB_WAIT;
2441541Srgrimes	return (tsleep((caddr_t)&sb->sb_cc,
24512843Sbde	    (sb->sb_flags & SB_NOINTR) ? PSOCK : PSOCK | PCATCH, "sbwait",
2461541Srgrimes	    sb->sb_timeo));
2471541Srgrimes}
2481541Srgrimes
2498876Srgrimes/*
2501541Srgrimes * Lock a sockbuf already known to be locked;
2511541Srgrimes * return any error returned from sleep (EINTR).
2521541Srgrimes */
2531549Srgrimesint
2541541Srgrimessb_lock(sb)
2551541Srgrimes	register struct sockbuf *sb;
2561541Srgrimes{
2571541Srgrimes	int error;
2581541Srgrimes
2591541Srgrimes	while (sb->sb_flags & SB_LOCK) {
2601541Srgrimes		sb->sb_flags |= SB_WANT;
2618876Srgrimes		error = tsleep((caddr_t)&sb->sb_flags,
2621541Srgrimes		    (sb->sb_flags & SB_NOINTR) ? PSOCK : PSOCK|PCATCH,
26312843Sbde		    "sblock", 0);
2643308Sphk		if (error)
2651541Srgrimes			return (error);
2661541Srgrimes	}
2671541Srgrimes	sb->sb_flags |= SB_LOCK;
2681541Srgrimes	return (0);
2691541Srgrimes}
2701541Srgrimes
2711541Srgrimes/*
2721541Srgrimes * Wakeup processes waiting on a socket buffer.
2731541Srgrimes * Do asynchronous notification via SIGIO
2741541Srgrimes * if the socket has the SS_ASYNC flag set.
2751541Srgrimes */
2761549Srgrimesvoid
2771541Srgrimessowakeup(so, sb)
2781541Srgrimes	register struct socket *so;
2791541Srgrimes	register struct sockbuf *sb;
2801541Srgrimes{
2811541Srgrimes	struct proc *p;
2821541Srgrimes
2831541Srgrimes	selwakeup(&sb->sb_sel);
2841541Srgrimes	sb->sb_flags &= ~SB_SEL;
2851541Srgrimes	if (sb->sb_flags & SB_WAIT) {
2861541Srgrimes		sb->sb_flags &= ~SB_WAIT;
2871541Srgrimes		wakeup((caddr_t)&sb->sb_cc);
2881541Srgrimes	}
2891541Srgrimes	if (so->so_state & SS_ASYNC) {
2901541Srgrimes		if (so->so_pgid < 0)
2911541Srgrimes			gsignal(-so->so_pgid, SIGIO);
2921541Srgrimes		else if (so->so_pgid > 0 && (p = pfind(so->so_pgid)) != 0)
2931541Srgrimes			psignal(p, SIGIO);
2941541Srgrimes	}
2951541Srgrimes}
2961541Srgrimes
2971541Srgrimes/*
2981541Srgrimes * Socket buffer (struct sockbuf) utility routines.
2991541Srgrimes *
3001541Srgrimes * Each socket contains two socket buffers: one for sending data and
3011541Srgrimes * one for receiving data.  Each buffer contains a queue of mbufs,
3021541Srgrimes * information about the number of mbufs and amount of data in the
3031541Srgrimes * queue, and other fields allowing select() statements and notification
3041541Srgrimes * on data availability to be implemented.
3051541Srgrimes *
3061541Srgrimes * Data stored in a socket buffer is maintained as a list of records.
3071541Srgrimes * Each record is a list of mbufs chained together with the m_next
3081541Srgrimes * field.  Records are chained together with the m_nextpkt field. The upper
3091541Srgrimes * level routine soreceive() expects the following conventions to be
3101541Srgrimes * observed when placing information in the receive buffer:
3111541Srgrimes *
3121541Srgrimes * 1. If the protocol requires each message be preceded by the sender's
3131541Srgrimes *    name, then a record containing that name must be present before
3141541Srgrimes *    any associated data (mbuf's must be of type MT_SONAME).
3151541Srgrimes * 2. If the protocol supports the exchange of ``access rights'' (really
3161541Srgrimes *    just additional data associated with the message), and there are
3171541Srgrimes *    ``rights'' to be received, then a record containing this data
3181541Srgrimes *    should be present (mbuf's must be of type MT_RIGHTS).
3191541Srgrimes * 3. If a name or rights record exists, then it must be followed by
3201541Srgrimes *    a data record, perhaps of zero length.
3211541Srgrimes *
3221541Srgrimes * Before using a new socket structure it is first necessary to reserve
3231541Srgrimes * buffer space to the socket, by calling sbreserve().  This should commit
3241541Srgrimes * some of the available buffer space in the system buffer pool for the
3251541Srgrimes * socket (currently, it does nothing but enforce limits).  The space
3261541Srgrimes * should be released by calling sbrelease() when the socket is destroyed.
3271541Srgrimes */
3281541Srgrimes
3291549Srgrimesint
3301541Srgrimessoreserve(so, sndcc, rcvcc)
3311541Srgrimes	register struct socket *so;
3321541Srgrimes	u_long sndcc, rcvcc;
3331541Srgrimes{
3341541Srgrimes
3351541Srgrimes	if (sbreserve(&so->so_snd, sndcc) == 0)
3361541Srgrimes		goto bad;
3371541Srgrimes	if (sbreserve(&so->so_rcv, rcvcc) == 0)
3381541Srgrimes		goto bad2;
3391541Srgrimes	if (so->so_rcv.sb_lowat == 0)
3401541Srgrimes		so->so_rcv.sb_lowat = 1;
3411541Srgrimes	if (so->so_snd.sb_lowat == 0)
3421541Srgrimes		so->so_snd.sb_lowat = MCLBYTES;
3431541Srgrimes	if (so->so_snd.sb_lowat > so->so_snd.sb_hiwat)
3441541Srgrimes		so->so_snd.sb_lowat = so->so_snd.sb_hiwat;
3451541Srgrimes	return (0);
3461541Srgrimesbad2:
3471541Srgrimes	sbrelease(&so->so_snd);
3481541Srgrimesbad:
3491541Srgrimes	return (ENOBUFS);
3501541Srgrimes}
3511541Srgrimes
3521541Srgrimes/*
3531541Srgrimes * Allot mbufs to a sockbuf.
3541541Srgrimes * Attempt to scale mbmax so that mbcnt doesn't become limiting
3551541Srgrimes * if buffering efficiency is near the normal case.
3561541Srgrimes */
3571549Srgrimesint
3581541Srgrimessbreserve(sb, cc)
3591541Srgrimes	struct sockbuf *sb;
3601541Srgrimes	u_long cc;
3611541Srgrimes{
3621541Srgrimes
3631541Srgrimes	if (cc > sb_max * MCLBYTES / (MSIZE + MCLBYTES))
3641541Srgrimes		return (0);
3651541Srgrimes	sb->sb_hiwat = cc;
36613267Swollman	sb->sb_mbmax = min(cc * sb_efficiency, sb_max);
3671541Srgrimes	if (sb->sb_lowat > sb->sb_hiwat)
3681541Srgrimes		sb->sb_lowat = sb->sb_hiwat;
3691541Srgrimes	return (1);
3701541Srgrimes}
3711541Srgrimes
3721541Srgrimes/*
3731541Srgrimes * Free mbufs held by a socket, and reserved mbuf space.
3741541Srgrimes */
3751549Srgrimesvoid
3761541Srgrimessbrelease(sb)
3771541Srgrimes	struct sockbuf *sb;
3781541Srgrimes{
3791541Srgrimes
3801541Srgrimes	sbflush(sb);
3811541Srgrimes	sb->sb_hiwat = sb->sb_mbmax = 0;
3821541Srgrimes}
3831541Srgrimes
3841541Srgrimes/*
3851541Srgrimes * Routines to add and remove
3861541Srgrimes * data from an mbuf queue.
3871541Srgrimes *
3881541Srgrimes * The routines sbappend() or sbappendrecord() are normally called to
3891541Srgrimes * append new mbufs to a socket buffer, after checking that adequate
3901541Srgrimes * space is available, comparing the function sbspace() with the amount
3911541Srgrimes * of data to be added.  sbappendrecord() differs from sbappend() in
3921541Srgrimes * that data supplied is treated as the beginning of a new record.
3931541Srgrimes * To place a sender's address, optional access rights, and data in a
3941541Srgrimes * socket receive buffer, sbappendaddr() should be used.  To place
3951541Srgrimes * access rights and data in a socket receive buffer, sbappendrights()
3961541Srgrimes * should be used.  In either case, the new data begins a new record.
3971541Srgrimes * Note that unlike sbappend() and sbappendrecord(), these routines check
3981541Srgrimes * for the caller that there will be enough space to store the data.
3991541Srgrimes * Each fails if there is not enough space, or if it cannot find mbufs
4001541Srgrimes * to store additional information in.
4011541Srgrimes *
4021541Srgrimes * Reliable protocols may use the socket send buffer to hold data
4031541Srgrimes * awaiting acknowledgement.  Data is normally copied from a socket
4041541Srgrimes * send buffer in a protocol with m_copy for output to a peer,
4051541Srgrimes * and then removing the data from the socket buffer with sbdrop()
4061541Srgrimes * or sbdroprecord() when the data is acknowledged by the peer.
4071541Srgrimes */
4081541Srgrimes
4091541Srgrimes/*
4101541Srgrimes * Append mbuf chain m to the last record in the
4111541Srgrimes * socket buffer sb.  The additional space associated
4121541Srgrimes * the mbuf chain is recorded in sb.  Empty mbufs are
4131541Srgrimes * discarded and mbufs are compacted where possible.
4141541Srgrimes */
4151549Srgrimesvoid
4161541Srgrimessbappend(sb, m)
4171541Srgrimes	struct sockbuf *sb;
4181541Srgrimes	struct mbuf *m;
4191541Srgrimes{
4201541Srgrimes	register struct mbuf *n;
4211541Srgrimes
4221541Srgrimes	if (m == 0)
4231541Srgrimes		return;
4243308Sphk	n = sb->sb_mb;
4253308Sphk	if (n) {
4261541Srgrimes		while (n->m_nextpkt)
4271541Srgrimes			n = n->m_nextpkt;
4281541Srgrimes		do {
4291541Srgrimes			if (n->m_flags & M_EOR) {
4301541Srgrimes				sbappendrecord(sb, m); /* XXXXXX!!!! */
4311541Srgrimes				return;
4321541Srgrimes			}
4331541Srgrimes		} while (n->m_next && (n = n->m_next));
4341541Srgrimes	}
4351541Srgrimes	sbcompress(sb, m, n);
4361541Srgrimes}
4371541Srgrimes
4381541Srgrimes#ifdef SOCKBUF_DEBUG
4391549Srgrimesvoid
4401541Srgrimessbcheck(sb)
4411541Srgrimes	register struct sockbuf *sb;
4421541Srgrimes{
4431541Srgrimes	register struct mbuf *m;
4441541Srgrimes	register int len = 0, mbcnt = 0;
4451541Srgrimes
4461541Srgrimes	for (m = sb->sb_mb; m; m = m->m_next) {
4471541Srgrimes		len += m->m_len;
4481541Srgrimes		mbcnt += MSIZE;
44917675Sjulian		if (m->m_flags & M_EXT) /*XXX*/ /* pretty sure this is bogus */
4501541Srgrimes			mbcnt += m->m_ext.ext_size;
4511541Srgrimes		if (m->m_nextpkt)
4521541Srgrimes			panic("sbcheck nextpkt");
4531541Srgrimes	}
4541541Srgrimes	if (len != sb->sb_cc || mbcnt != sb->sb_mbcnt) {
4551541Srgrimes		printf("cc %d != %d || mbcnt %d != %d\n", len, sb->sb_cc,
4561541Srgrimes		    mbcnt, sb->sb_mbcnt);
4571541Srgrimes		panic("sbcheck");
4581541Srgrimes	}
4591541Srgrimes}
4601541Srgrimes#endif
4611541Srgrimes
4621541Srgrimes/*
4631541Srgrimes * As above, except the mbuf chain
4641541Srgrimes * begins a new record.
4651541Srgrimes */
4661549Srgrimesvoid
4671541Srgrimessbappendrecord(sb, m0)
4681541Srgrimes	register struct sockbuf *sb;
4691541Srgrimes	register struct mbuf *m0;
4701541Srgrimes{
4711541Srgrimes	register struct mbuf *m;
4721541Srgrimes
4731541Srgrimes	if (m0 == 0)
4741541Srgrimes		return;
4753308Sphk	m = sb->sb_mb;
4763308Sphk	if (m)
4771541Srgrimes		while (m->m_nextpkt)
4781541Srgrimes			m = m->m_nextpkt;
4791541Srgrimes	/*
4801541Srgrimes	 * Put the first mbuf on the queue.
4811541Srgrimes	 * Note this permits zero length records.
4821541Srgrimes	 */
4831541Srgrimes	sballoc(sb, m0);
4841541Srgrimes	if (m)
4851541Srgrimes		m->m_nextpkt = m0;
4861541Srgrimes	else
4871541Srgrimes		sb->sb_mb = m0;
4881541Srgrimes	m = m0->m_next;
4891541Srgrimes	m0->m_next = 0;
4901541Srgrimes	if (m && (m0->m_flags & M_EOR)) {
4911541Srgrimes		m0->m_flags &= ~M_EOR;
4921541Srgrimes		m->m_flags |= M_EOR;
4931541Srgrimes	}
4941541Srgrimes	sbcompress(sb, m, m0);
4951541Srgrimes}
4961541Srgrimes
4971541Srgrimes/*
4981541Srgrimes * As above except that OOB data
4991541Srgrimes * is inserted at the beginning of the sockbuf,
5001541Srgrimes * but after any other OOB data.
5011541Srgrimes */
5021549Srgrimesvoid
5031541Srgrimessbinsertoob(sb, m0)
5041541Srgrimes	register struct sockbuf *sb;
5051541Srgrimes	register struct mbuf *m0;
5061541Srgrimes{
5071541Srgrimes	register struct mbuf *m;
5081541Srgrimes	register struct mbuf **mp;
5091541Srgrimes
5101541Srgrimes	if (m0 == 0)
5111541Srgrimes		return;
5123308Sphk	for (mp = &sb->sb_mb; *mp ; mp = &((*mp)->m_nextpkt)) {
5133308Sphk	    m = *mp;
5141541Srgrimes	    again:
5151541Srgrimes		switch (m->m_type) {
5161541Srgrimes
5171541Srgrimes		case MT_OOBDATA:
5181541Srgrimes			continue;		/* WANT next train */
5191541Srgrimes
5201541Srgrimes		case MT_CONTROL:
5213308Sphk			m = m->m_next;
5223308Sphk			if (m)
5231541Srgrimes				goto again;	/* inspect THIS train further */
5241541Srgrimes		}
5251541Srgrimes		break;
5261541Srgrimes	}
5271541Srgrimes	/*
5281541Srgrimes	 * Put the first mbuf on the queue.
5291541Srgrimes	 * Note this permits zero length records.
5301541Srgrimes	 */
5311541Srgrimes	sballoc(sb, m0);
5321541Srgrimes	m0->m_nextpkt = *mp;
5331541Srgrimes	*mp = m0;
5341541Srgrimes	m = m0->m_next;
5351541Srgrimes	m0->m_next = 0;
5361541Srgrimes	if (m && (m0->m_flags & M_EOR)) {
5371541Srgrimes		m0->m_flags &= ~M_EOR;
5381541Srgrimes		m->m_flags |= M_EOR;
5391541Srgrimes	}
5401541Srgrimes	sbcompress(sb, m, m0);
5411541Srgrimes}
5421541Srgrimes
5431541Srgrimes/*
5441541Srgrimes * Append address and data, and optionally, control (ancillary) data
5451541Srgrimes * to the receive queue of a socket.  If present,
5461541Srgrimes * m0 must include a packet header with total length.
5471541Srgrimes * Returns 0 if no space in sockbuf or insufficient mbufs.
5481541Srgrimes */
5491549Srgrimesint
5501541Srgrimessbappendaddr(sb, asa, m0, control)
5511541Srgrimes	register struct sockbuf *sb;
5521541Srgrimes	struct sockaddr *asa;
5531541Srgrimes	struct mbuf *m0, *control;
5541541Srgrimes{
5551541Srgrimes	register struct mbuf *m, *n;
5561541Srgrimes	int space = asa->sa_len;
5571541Srgrimes
5581541Srgrimesif (m0 && (m0->m_flags & M_PKTHDR) == 0)
5591541Srgrimespanic("sbappendaddr");
5601541Srgrimes	if (m0)
5611541Srgrimes		space += m0->m_pkthdr.len;
5621541Srgrimes	for (n = control; n; n = n->m_next) {
5631541Srgrimes		space += n->m_len;
5641541Srgrimes		if (n->m_next == 0)	/* keep pointer to last control buf */
5651541Srgrimes			break;
5661541Srgrimes	}
5671541Srgrimes	if (space > sbspace(sb))
5681541Srgrimes		return (0);
5691541Srgrimes	if (asa->sa_len > MLEN)
5701541Srgrimes		return (0);
5711541Srgrimes	MGET(m, M_DONTWAIT, MT_SONAME);
5721541Srgrimes	if (m == 0)
5731541Srgrimes		return (0);
5741541Srgrimes	m->m_len = asa->sa_len;
5751541Srgrimes	bcopy((caddr_t)asa, mtod(m, caddr_t), asa->sa_len);
5761541Srgrimes	if (n)
5771541Srgrimes		n->m_next = m0;		/* concatenate data to control */
5781541Srgrimes	else
5791541Srgrimes		control = m0;
5801541Srgrimes	m->m_next = control;
5811541Srgrimes	for (n = m; n; n = n->m_next)
5821541Srgrimes		sballoc(sb, n);
5833308Sphk	n = sb->sb_mb;
5843308Sphk	if (n) {
5851541Srgrimes		while (n->m_nextpkt)
5861541Srgrimes			n = n->m_nextpkt;
5871541Srgrimes		n->m_nextpkt = m;
5881541Srgrimes	} else
5891541Srgrimes		sb->sb_mb = m;
5901541Srgrimes	return (1);
5911541Srgrimes}
5921541Srgrimes
5931549Srgrimesint
5941541Srgrimessbappendcontrol(sb, m0, control)
5951541Srgrimes	struct sockbuf *sb;
5961541Srgrimes	struct mbuf *control, *m0;
5971541Srgrimes{
5981541Srgrimes	register struct mbuf *m, *n;
5991541Srgrimes	int space = 0;
6001541Srgrimes
6011541Srgrimes	if (control == 0)
6021541Srgrimes		panic("sbappendcontrol");
6031541Srgrimes	for (m = control; ; m = m->m_next) {
6041541Srgrimes		space += m->m_len;
6051541Srgrimes		if (m->m_next == 0)
6061541Srgrimes			break;
6071541Srgrimes	}
6081541Srgrimes	n = m;			/* save pointer to last control buffer */
6091541Srgrimes	for (m = m0; m; m = m->m_next)
6101541Srgrimes		space += m->m_len;
6111541Srgrimes	if (space > sbspace(sb))
6121541Srgrimes		return (0);
6131541Srgrimes	n->m_next = m0;			/* concatenate data to control */
6141541Srgrimes	for (m = control; m; m = m->m_next)
6151541Srgrimes		sballoc(sb, m);
6163308Sphk	n = sb->sb_mb;
6173308Sphk	if (n) {
6181541Srgrimes		while (n->m_nextpkt)
6191541Srgrimes			n = n->m_nextpkt;
6201541Srgrimes		n->m_nextpkt = control;
6211541Srgrimes	} else
6221541Srgrimes		sb->sb_mb = control;
6231541Srgrimes	return (1);
6241541Srgrimes}
6251541Srgrimes
6261541Srgrimes/*
6271541Srgrimes * Compress mbuf chain m into the socket
6281541Srgrimes * buffer sb following mbuf n.  If n
6291541Srgrimes * is null, the buffer is presumed empty.
6301541Srgrimes */
6311549Srgrimesvoid
6321541Srgrimessbcompress(sb, m, n)
6331541Srgrimes	register struct sockbuf *sb;
6341541Srgrimes	register struct mbuf *m, *n;
6351541Srgrimes{
6361541Srgrimes	register int eor = 0;
6371541Srgrimes	register struct mbuf *o;
6381541Srgrimes
6391541Srgrimes	while (m) {
6401541Srgrimes		eor |= m->m_flags & M_EOR;
6411541Srgrimes		if (m->m_len == 0 &&
6421541Srgrimes		    (eor == 0 ||
6431541Srgrimes		     (((o = m->m_next) || (o = n)) &&
6441541Srgrimes		      o->m_type == m->m_type))) {
6451541Srgrimes			m = m_free(m);
6461541Srgrimes			continue;
6471541Srgrimes		}
6481541Srgrimes		if (n && (n->m_flags & (M_EXT | M_EOR)) == 0 &&
6491541Srgrimes		    (n->m_data + n->m_len + m->m_len) < &n->m_dat[MLEN] &&
6501541Srgrimes		    n->m_type == m->m_type) {
6511541Srgrimes			bcopy(mtod(m, caddr_t), mtod(n, caddr_t) + n->m_len,
6521541Srgrimes			    (unsigned)m->m_len);
6531541Srgrimes			n->m_len += m->m_len;
6541541Srgrimes			sb->sb_cc += m->m_len;
6551541Srgrimes			m = m_free(m);
6561541Srgrimes			continue;
6571541Srgrimes		}
6581541Srgrimes		if (n)
6591541Srgrimes			n->m_next = m;
6601541Srgrimes		else
6611541Srgrimes			sb->sb_mb = m;
6621541Srgrimes		sballoc(sb, m);
6631541Srgrimes		n = m;
6641541Srgrimes		m->m_flags &= ~M_EOR;
6651541Srgrimes		m = m->m_next;
6661541Srgrimes		n->m_next = 0;
6671541Srgrimes	}
6681541Srgrimes	if (eor) {
6691541Srgrimes		if (n)
6701541Srgrimes			n->m_flags |= eor;
6711541Srgrimes		else
6721541Srgrimes			printf("semi-panic: sbcompress\n");
6731541Srgrimes	}
6741541Srgrimes}
6751541Srgrimes
6761541Srgrimes/*
6771541Srgrimes * Free all mbufs in a sockbuf.
6781541Srgrimes * Check that all resources are reclaimed.
6791541Srgrimes */
6801549Srgrimesvoid
6811541Srgrimessbflush(sb)
6821541Srgrimes	register struct sockbuf *sb;
6831541Srgrimes{
6841541Srgrimes
6851541Srgrimes	if (sb->sb_flags & SB_LOCK)
6861541Srgrimes		panic("sbflush");
6871541Srgrimes	while (sb->sb_mbcnt)
6881541Srgrimes		sbdrop(sb, (int)sb->sb_cc);
6891541Srgrimes	if (sb->sb_cc || sb->sb_mb)
6901541Srgrimes		panic("sbflush 2");
6911541Srgrimes}
6921541Srgrimes
6931541Srgrimes/*
6941541Srgrimes * Drop data from (the front of) a sockbuf.
6951541Srgrimes */
6961549Srgrimesvoid
6971541Srgrimessbdrop(sb, len)
6981541Srgrimes	register struct sockbuf *sb;
6991541Srgrimes	register int len;
7001541Srgrimes{
7011541Srgrimes	register struct mbuf *m, *mn;
7021541Srgrimes	struct mbuf *next;
7031541Srgrimes
7041541Srgrimes	next = (m = sb->sb_mb) ? m->m_nextpkt : 0;
7051541Srgrimes	while (len > 0) {
7061541Srgrimes		if (m == 0) {
7071541Srgrimes			if (next == 0)
7081541Srgrimes				panic("sbdrop");
7091541Srgrimes			m = next;
7101541Srgrimes			next = m->m_nextpkt;
7111541Srgrimes			continue;
7121541Srgrimes		}
7131541Srgrimes		if (m->m_len > len) {
7141541Srgrimes			m->m_len -= len;
7151541Srgrimes			m->m_data += len;
7161541Srgrimes			sb->sb_cc -= len;
7171541Srgrimes			break;
7181541Srgrimes		}
7191541Srgrimes		len -= m->m_len;
7201541Srgrimes		sbfree(sb, m);
7211541Srgrimes		MFREE(m, mn);
7221541Srgrimes		m = mn;
7231541Srgrimes	}
7241541Srgrimes	while (m && m->m_len == 0) {
7251541Srgrimes		sbfree(sb, m);
7261541Srgrimes		MFREE(m, mn);
7271541Srgrimes		m = mn;
7281541Srgrimes	}
7291541Srgrimes	if (m) {
7301541Srgrimes		sb->sb_mb = m;
7311541Srgrimes		m->m_nextpkt = next;
7321541Srgrimes	} else
7331541Srgrimes		sb->sb_mb = next;
7341541Srgrimes}
7351541Srgrimes
7361541Srgrimes/*
7371541Srgrimes * Drop a record off the front of a sockbuf
7381541Srgrimes * and move the next record to the front.
7391541Srgrimes */
7401549Srgrimesvoid
7411541Srgrimessbdroprecord(sb)
7421541Srgrimes	register struct sockbuf *sb;
7431541Srgrimes{
7441541Srgrimes	register struct mbuf *m, *mn;
7451541Srgrimes
7461541Srgrimes	m = sb->sb_mb;
7471541Srgrimes	if (m) {
7481541Srgrimes		sb->sb_mb = m->m_nextpkt;
7491541Srgrimes		do {
7501541Srgrimes			sbfree(sb, m);
7511541Srgrimes			MFREE(m, mn);
7523308Sphk			m = mn;
7533308Sphk		} while (m);
7541541Srgrimes	}
7551541Srgrimes}
75617047Swollman
75717047Swollman#ifdef PRU_OLDSTYLE
75817047Swollman/*
75917047Swollman * The following routines mediate between the old-style `pr_usrreq'
76017047Swollman * protocol implementations and the new-style `struct pr_usrreqs'
76117047Swollman * calling convention.
76217047Swollman */
76317047Swollman
76417047Swollman/* syntactic sugar */
76517047Swollman#define	nomb	(struct mbuf *)0
76617047Swollman
76717047Swollmanstatic int
76817047Swollmanold_abort(struct socket *so)
76917047Swollman{
77017096Swollman	return so->so_proto->pr_ousrreq(so, PRU_ABORT, nomb, nomb, nomb);
77117047Swollman}
77217047Swollman
77317047Swollmanstatic int
77417047Swollmanold_accept(struct socket *so, struct mbuf *nam)
77517047Swollman{
77617096Swollman	return so->so_proto->pr_ousrreq(so, PRU_ACCEPT, nomb,  nam, nomb);
77717047Swollman}
77817047Swollman
77917047Swollmanstatic int
78017047Swollmanold_attach(struct socket *so, int proto)
78117047Swollman{
78217096Swollman	return so->so_proto->pr_ousrreq(so, PRU_ATTACH, nomb,
78317047Swollman				       (struct mbuf *)proto, /* XXX */
78417047Swollman				       nomb);
78517047Swollman}
78617047Swollman
78717047Swollmanstatic int
78817047Swollmanold_bind(struct socket *so, struct mbuf *nam)
78917047Swollman{
79017096Swollman	return so->so_proto->pr_ousrreq(so, PRU_BIND, nomb, nam, nomb);
79117047Swollman}
79217047Swollman
79317047Swollmanstatic int
79417047Swollmanold_connect(struct socket *so, struct mbuf *nam)
79517047Swollman{
79617096Swollman	return so->so_proto->pr_ousrreq(so, PRU_CONNECT, nomb, nam, nomb);
79717047Swollman}
79817047Swollman
79917047Swollmanstatic int
80017047Swollmanold_connect2(struct socket *so1, struct socket *so2)
80117047Swollman{
80217096Swollman	return so1->so_proto->pr_ousrreq(so1, PRU_CONNECT2, nomb,
80317047Swollman				       (struct mbuf *)so2, nomb);
80417047Swollman}
80517047Swollman
80617047Swollmanstatic int
80717096Swollmanold_control(struct socket *so, int cmd, caddr_t data, struct ifnet *ifp)
80817047Swollman{
80917096Swollman	return so->so_proto->pr_ousrreq(so, PRU_CONTROL, (struct mbuf *)cmd,
81017096Swollman				       (struct mbuf *)data,
81117096Swollman				       (struct mbuf *)ifp);
81217047Swollman}
81317047Swollman
81417047Swollmanstatic int
81517047Swollmanold_detach(struct socket *so)
81617047Swollman{
81717096Swollman	return so->so_proto->pr_ousrreq(so, PRU_DETACH, nomb, nomb, nomb);
81817047Swollman}
81917047Swollman
82017047Swollmanstatic int
82117047Swollmanold_disconnect(struct socket *so)
82217047Swollman{
82317096Swollman	return so->so_proto->pr_ousrreq(so, PRU_DISCONNECT, nomb, nomb, nomb);
82417047Swollman}
82517047Swollman
82617047Swollmanstatic int
82717047Swollmanold_listen(struct socket *so)
82817047Swollman{
82917096Swollman	return so->so_proto->pr_ousrreq(so, PRU_LISTEN, nomb, nomb, nomb);
83017047Swollman}
83117047Swollman
83217047Swollmanstatic int
83317047Swollmanold_peeraddr(struct socket *so, struct mbuf *nam)
83417047Swollman{
83517096Swollman	return so->so_proto->pr_ousrreq(so, PRU_PEERADDR, nomb, nam, nomb);
83617047Swollman}
83717047Swollman
83817047Swollmanstatic int
83917047Swollmanold_rcvd(struct socket *so, int flags)
84017047Swollman{
84117096Swollman	return so->so_proto->pr_ousrreq(so, PRU_RCVD, nomb,
84217047Swollman				       (struct mbuf *)flags, /* XXX */
84317047Swollman				       nomb);
84417047Swollman}
84517047Swollman
84617047Swollmanstatic int
84717047Swollmanold_rcvoob(struct socket *so, struct mbuf *m, int flags)
84817047Swollman{
84917096Swollman	return so->so_proto->pr_ousrreq(so, PRU_RCVOOB, m,
85017047Swollman				       (struct mbuf *)flags, /* XXX */
85117047Swollman				       nomb);
85217047Swollman}
85317047Swollman
85417047Swollmanstatic int
85517047Swollmanold_send(struct socket *so, int flags, struct mbuf *m, struct mbuf *addr,
85617047Swollman	 struct mbuf *control)
85717047Swollman{
85817047Swollman	int req;
85917047Swollman
86017047Swollman	if (flags & PRUS_OOB) {
86117047Swollman		req = PRU_SENDOOB;
86217047Swollman	} else if(flags & PRUS_EOF) {
86317047Swollman		req = PRU_SEND_EOF;
86417047Swollman	} else {
86517047Swollman		req = PRU_SEND;
86617047Swollman	}
86717096Swollman	return so->so_proto->pr_ousrreq(so, req, m, addr, control);
86817047Swollman}
86917047Swollman
87017047Swollmanstatic int
87117047Swollmanold_sense(struct socket *so, struct stat *sb)
87217047Swollman{
87317096Swollman	return so->so_proto->pr_ousrreq(so, PRU_SENSE, (struct mbuf *)sb,
87417047Swollman				       nomb, nomb);
87517047Swollman}
87617047Swollman
87717047Swollmanstatic int
87817047Swollmanold_shutdown(struct socket *so)
87917047Swollman{
88017096Swollman	return so->so_proto->pr_ousrreq(so, PRU_SHUTDOWN, nomb, nomb, nomb);
88117047Swollman}
88217047Swollman
88317047Swollmanstatic int
88417047Swollmanold_sockaddr(struct socket *so, struct mbuf *nam)
88517047Swollman{
88617096Swollman	return so->so_proto->pr_ousrreq(so, PRU_SOCKADDR, nomb, nam, nomb);
88717047Swollman}
88817047Swollman
88917047Swollmanstruct pr_usrreqs pru_oldstyle = {
89017047Swollman	old_abort, old_accept, old_attach, old_bind, old_connect,
89117047Swollman	old_connect2, old_control, old_detach, old_disconnect,
89217047Swollman	old_listen, old_peeraddr, old_rcvd, old_rcvoob, old_send,
89317047Swollman	old_sense, old_shutdown, old_sockaddr
89417047Swollman};
89517047Swollman
89617096Swollman#endif /* PRU_OLDSTYLE */
89717096Swollman
89817047Swollman/*
89917096Swollman * Some routines that return EOPNOTSUPP for entry points that are not
90017096Swollman * supported by a protocol.  Fill in as needed.
90117047Swollman */
90217047Swollmanint
90317096Swollmanpru_connect2_notsupp(struct socket *so1, struct socket *so2)
90417047Swollman{
90517096Swollman	return EOPNOTSUPP;
90617047Swollman}
907