xref: /src/sys/kern/uipc_mbuf.c

/*	$NetBSD: uipc_mbuf.c,v 1.50 2000/08/18 16:19:22 itojun Exp $	*/

/*-
 * Copyright (c) 1999 The NetBSD Foundation, Inc.
 * All rights reserved.
 *
 * This code is derived from software contributed to The NetBSD Foundation
 * by Jason R. Thorpe of the Numerical Aerospace Simulation Facility,
 * NASA Ames Research Center.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 * 3. All advertising materials mentioning features or use of this software
 *    must display the following acknowledgement:
 *	This product includes software developed by the NetBSD
 *	Foundation, Inc. and its contributors.
 * 4. Neither the name of The NetBSD Foundation nor the names of its
 *    contributors may be used to endorse or promote products derived
 *    from this software without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
 * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
 * POSSIBILITY OF SUCH DAMAGE.
 */

/*
 * Copyright (c) 1982, 1986, 1988, 1991, 1993
 *	The Regents of the University of California.  All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 * 3. All advertising materials mentioning features or use of this software
 *    must display the following acknowledgement:
 *	This product includes software developed by the University of
 *	California, Berkeley and its contributors.
 * 4. Neither the name of the University nor the names of its contributors
 *    may be used to endorse or promote products derived from this software
 *    without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 * SUCH DAMAGE.
 *
 *	@(#)uipc_mbuf.c	8.4 (Berkeley) 2/14/95
 */

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/proc.h>
#include <sys/malloc.h>
#include <sys/map.h>
#define MBTYPES
#include <sys/mbuf.h>
#include <sys/kernel.h>
#include <sys/syslog.h>
#include <sys/domain.h>
#include <sys/protosw.h>
#include <sys/pool.h>
#include <sys/socket.h>
#include <net/if.h>

#include <uvm/uvm_extern.h>

#include <sys/sysctl.h>

struct	pool mbpool;		/* mbuf pool */
struct	pool mclpool;		/* mbuf cluster pool */

struct mbstat mbstat;
int	max_linkhdr;
int	max_protohdr;
int	max_hdr;
int	max_datalen;

void	*mclpool_alloc __P((unsigned long, int, int));
void	mclpool_release __P((void *, unsigned long, int));
static struct mbuf *m_copym0 __P((struct mbuf *, int, int, int, int));

const char *mclpool_warnmsg =
    "WARNING: mclpool limit reached; increase NMBCLUSTERS";

/*
 * Initialize the mbuf allcator.
 */
void
mbinit()
{

	pool_init(&mbpool, msize, 0, 0, 0, "mbpl", 0, NULL, NULL, 0);
	pool_init(&mclpool, mclbytes, 0, 0, 0, "mclpl", 0, mclpool_alloc,
	    mclpool_release, 0);

	/*
	 * Set the hard limit on the mclpool to the number of
	 * mbuf clusters the kernel is to support.  Log the limit
	 * reached message max once a minute.
	 */
	pool_sethardlimit(&mclpool, nmbclusters, mclpool_warnmsg, 60);

	/*
	 * Set a low water mark for both mbufs and clusters.  This should
	 * help ensure that they can be allocated in a memory starvation
	 * situation.  This is important for e.g. diskless systems which
	 * must allocate mbufs in order for the pagedaemon to clean pages.
	 */
	pool_setlowat(&mbpool, mblowat);
	pool_setlowat(&mclpool, mcllowat);
}

int
sysctl_dombuf(name, namelen, oldp, oldlenp, newp, newlen)
	int *name;
	u_int namelen;
	void *oldp;
	size_t *oldlenp;
	void *newp;
	size_t newlen;
{
	int error, newval;

	/* All sysctl names at this level are terminal. */
	if (namelen != 1)
		return (ENOTDIR);		/* overloaded */

	switch (name[0]) {
	case MBUF_MSIZE:
		return (sysctl_rdint(oldp, oldlenp, newp, msize));
	case MBUF_MCLBYTES:
		return (sysctl_rdint(oldp, oldlenp, newp, mclbytes));
	case MBUF_NMBCLUSTERS:
		/*
		 * If we have direct-mapped pool pages, we can adjust this
		 * number on the fly.  If not, we're limited by the size
		 * of mb_map, and cannot change this value.
		 *
		 * Note: we only allow the value to be increased, never
		 * decreased.
		 */
		if (mb_map == NULL) {
			newval = nmbclusters;
			error = sysctl_int(oldp, oldlenp, newp, newlen,
			    &newval);
			if (error != 0)
				return (error);
			if (newp != NULL) {
				if (newval >= nmbclusters) {
					nmbclusters = newval;
					pool_sethardlimit(&mclpool,
					    nmbclusters, mclpool_warnmsg, 60);
				} else
					error = EINVAL;
			}
			return (error);
		} else
			return (sysctl_rdint(oldp, oldlenp, newp, nmbclusters));
	case MBUF_MBLOWAT:
	case MBUF_MCLLOWAT:
		/* New value must be >= 0. */
		newval = (name[0] == MBUF_MBLOWAT) ? mblowat : mcllowat;
		error = sysctl_int(oldp, oldlenp, newp, newlen, &newval);
		if (error != 0)
			return (error);
		if (newp != NULL) {
			if (newval >= 0) {
				if (name[0] == MBUF_MBLOWAT) {
					mblowat = newval;
					pool_setlowat(&mbpool, newval);
				} else {
					mcllowat = newval;
					pool_setlowat(&mclpool, newval);
				}
			} else
				error = EINVAL;
		}
		return (error);
	default:
		return (EOPNOTSUPP);
	}
	/* NOTREACHED */
}

void *
mclpool_alloc(sz, flags, mtype)
	unsigned long sz;
	int flags;
	int mtype;
{
	boolean_t waitok = (flags & PR_WAITOK) ? TRUE : FALSE;

	return ((void *)uvm_km_alloc_poolpage1(mb_map, uvmexp.mb_object,
	    waitok));
}

void
mclpool_release(v, sz, mtype)
	void *v;
	unsigned long sz;
	int mtype;
{

	uvm_km_free_poolpage1(mb_map, (vaddr_t)v);
}

/*
 * When MGET failes, ask protocols to free space when short of memory,
 * then re-attempt to allocate an mbuf.
 */
struct mbuf *
m_retry(i, t)
	int i, t;
{
	struct mbuf *m;

	m_reclaim(i);
#define m_retry(i, t)	(struct mbuf *)0
	MGET(m, i, t);
#undef m_retry
	if (m != NULL)
		mbstat.m_wait++;
	else
		mbstat.m_drops++;
	return (m);
}

/*
 * As above; retry an MGETHDR.
 */
struct mbuf *
m_retryhdr(i, t)
	int i, t;
{
	struct mbuf *m;

	m_reclaim(i);
#define m_retryhdr(i, t) (struct mbuf *)0
	MGETHDR(m, i, t);
#undef m_retryhdr
	if (m != NULL)
		mbstat.m_wait++;
	else
		mbstat.m_drops++;
	return (m);
}

void
m_reclaim(how)
	int how;
{
	struct domain *dp;
	struct protosw *pr;
	struct ifnet *ifp;
	int s = splimp();

	for (dp = domains; dp; dp = dp->dom_next)
		for (pr = dp->dom_protosw;
		     pr < dp->dom_protoswNPROTOSW; pr++)
			if (pr->pr_drain)
				(*pr->pr_drain)();
	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
		if (ifp->if_drain)
			(*ifp->if_drain)(ifp);
	splx(s);
	mbstat.m_drain++;
}

/*
 * Space allocation routines.
 * These are also available as macros
 * for critical paths.
 */
struct mbuf *
m_get(nowait, type)
	int nowait, type;
{
	struct mbuf *m;

	MGET(m, nowait, type);
	return (m);
}

struct mbuf *
m_gethdr(nowait, type)
	int nowait, type;
{
	struct mbuf *m;

	MGETHDR(m, nowait, type);
	return (m);
}

struct mbuf *
m_getclr(nowait, type)
	int nowait, type;
{
	struct mbuf *m;

	MGET(m, nowait, type);
	if (m == 0)
		return (0);
	memset(mtod(m, caddr_t), 0, MLEN);
	return (m);
}

struct mbuf *
m_free(m)
	struct mbuf *m;
{
	struct mbuf *n;

	MFREE(m, n);
	return (n);
}

void
m_freem(m)
	struct mbuf *m;
{
	struct mbuf *n;

	if (m == NULL)
		return;
	if ((m->m_flags & M_PKTHDR) != 0 && m->m_pkthdr.aux) {
		m_freem(m->m_pkthdr.aux);
		m->m_pkthdr.aux = NULL;
	}
	do {
		MFREE(m, n);
		m = n;
	} while (m);
}

/*
 * Mbuffer utility routines.
 */

/*
 * Lesser-used path for M_PREPEND:
 * allocate new mbuf to prepend to chain,
 * copy junk along.
 */
struct mbuf *
m_prepend(m, len, how)
	struct mbuf *m;
	int len, how;
{
	struct mbuf *mn;

	MGET(mn, how, m->m_type);
	if (mn == (struct mbuf *)NULL) {
		m_freem(m);
		return ((struct mbuf *)NULL);
	}
	if (m->m_flags & M_PKTHDR) {
		M_COPY_PKTHDR(mn, m);
		m->m_flags &= ~M_PKTHDR;
	}
	mn->m_next = m;
	m = mn;
	if (len < MHLEN)
		MH_ALIGN(m, len);
	m->m_len = len;
	return (m);
}

/*
 * Make a copy of an mbuf chain starting "off0" bytes from the beginning,
 * continuing for "len" bytes.  If len is M_COPYALL, copy to end of mbuf.
 * The wait parameter is a choice of M_WAIT/M_DONTWAIT from caller.
 */
int MCFail;

struct mbuf *
m_copym(m, off0, len, wait)
	struct mbuf *m;
	int off0, wait;
	int len;
{
	return m_copym0(m, off0, len, wait, 0);	/* shallow copy on M_EXT */
}

struct mbuf *
m_dup(m, off0, len, wait)
	struct mbuf *m;
	int off0, wait;
	int len;
{
	return m_copym0(m, off0, len, wait, 1);	/* deep copy */
}

static struct mbuf *
m_copym0(m, off0, len, wait, deep)
	struct mbuf *m;
	int off0, wait;
	int len;
	int deep;	/* deep copy */
{
	struct mbuf *n, **np;
	int off = off0;
	struct mbuf *top;
	int copyhdr = 0;

	if (off < 0 || len < 0)
		panic("m_copym: off %d, len %d", off, len);
	if (off == 0 && m->m_flags & M_PKTHDR)
		copyhdr = 1;
	while (off > 0) {
		if (m == 0)
			panic("m_copym: m == 0");
		if (off < m->m_len)
			break;
		off -= m->m_len;
		m = m->m_next;
	}
	np = &top;
	top = 0;
	while (len > 0) {
		if (m == 0) {
			if (len != M_COPYALL)
				panic("m_copym: m == 0 and not COPYALL");
			break;
		}
		MGET(n, wait, m->m_type);
		*np = n;
		if (n == 0)
			goto nospace;
		if (copyhdr) {
			M_COPY_PKTHDR(n, m);
			if (len == M_COPYALL)
				n->m_pkthdr.len -= off0;
			else
				n->m_pkthdr.len = len;
			copyhdr = 0;
		}
		n->m_len = min(len, m->m_len - off);
		if (m->m_flags & M_EXT) {
			if (!deep) {
				n->m_data = m->m_data + off;
				n->m_ext = m->m_ext;
				MCLADDREFERENCE(m, n);
			} else {
				/*
				 * we are unsure about the way m was allocated.
				 * copy into multiple MCLBYTES cluster mbufs.
				 */
				MCLGET(n, wait);
				n->m_len = 0;
				n->m_len = M_TRAILINGSPACE(n);
				n->m_len = min(n->m_len, len);
				n->m_len = min(n->m_len, m->m_len - off);
				memcpy(mtod(n, caddr_t), mtod(m, caddr_t) + off,
				    (unsigned)n->m_len);
			}
		} else
			memcpy(mtod(n, caddr_t), mtod(m, caddr_t)+off,
			    (unsigned)n->m_len);
		if (len != M_COPYALL)
			len -= n->m_len;
		off += n->m_len;
#ifdef DIAGNOSTIC
		if (off > m->m_len)
			panic("m_copym0 overrun");
#endif
		if (off == m->m_len) {
			m = m->m_next;
			off = 0;
		}
		np = &n->m_next;
	}
	if (top == 0)
		MCFail++;
	return (top);
nospace:
	m_freem(top);
	MCFail++;
	return (0);
}

/*
 * Copy an entire packet, including header (which must be present).
 * An optimization of the common case `m_copym(m, 0, M_COPYALL, how)'.
 */
struct mbuf *
m_copypacket(m, how)
	struct mbuf *m;
	int how;
{
	struct mbuf *top, *n, *o;

	MGET(n, how, m->m_type);
	top = n;
	if (!n)
		goto nospace;

	M_COPY_PKTHDR(n, m);
	n->m_len = m->m_len;
	if (m->m_flags & M_EXT) {
		n->m_data = m->m_data;
		n->m_ext = m->m_ext;
		MCLADDREFERENCE(m, n);
	} else {
		memcpy(mtod(n, char *), mtod(m, char *), n->m_len);
	}

	m = m->m_next;
	while (m) {
		MGET(o, how, m->m_type);
		if (!o)
			goto nospace;

		n->m_next = o;
		n = n->m_next;

		n->m_len = m->m_len;
		if (m->m_flags & M_EXT) {
			n->m_data = m->m_data;
			n->m_ext = m->m_ext;
			MCLADDREFERENCE(m, n);
		} else {
			memcpy(mtod(n, char *), mtod(m, char *), n->m_len);
		}

		m = m->m_next;
	}
	return top;
nospace:
	m_freem(top);
	MCFail++;
	return 0;
}

/*
 * Copy data from an mbuf chain starting "off" bytes from the beginning,
 * continuing for "len" bytes, into the indicated buffer.
 */
void
m_copydata(m, off, len, cp)
	struct mbuf *m;
	int off;
	int len;
	caddr_t cp;
{
	unsigned count;

	if (off < 0 || len < 0)
		panic("m_copydata");
	while (off > 0) {
		if (m == 0)
			panic("m_copydata");
		if (off < m->m_len)
			break;
		off -= m->m_len;
		m = m->m_next;
	}
	while (len > 0) {
		if (m == 0)
			panic("m_copydata");
		count = min(m->m_len - off, len);
		memcpy(cp, mtod(m, caddr_t) + off, count);
		len -= count;
		cp += count;
		off = 0;
		m = m->m_next;
	}
}

/*
 * Concatenate mbuf chain n to m.
 * Both chains must be of the same type (e.g. MT_DATA).
 * Any m_pkthdr is not updated.
 */
void
m_cat(m, n)
	struct mbuf *m, *n;
{
	while (m->m_next)
		m = m->m_next;
	while (n) {
		if (m->m_flags & M_EXT ||
		    m->m_data + m->m_len + n->m_len >= &m->m_dat[MLEN]) {
			/* just join the two chains */
			m->m_next = n;
			return;
		}
		/* splat the data from one into the other */
		memcpy(mtod(m, caddr_t) + m->m_len, mtod(n, caddr_t),
		    (u_int)n->m_len);
		m->m_len += n->m_len;
		n = m_free(n);
	}
}

void
m_adj(mp, req_len)
	struct mbuf *mp;
	int req_len;
{
	int len = req_len;
	struct mbuf *m;
	int count;

	if ((m = mp) == NULL)
		return;
	if (len >= 0) {
		/*
		 * Trim from head.
		 */
		while (m != NULL && len > 0) {
			if (m->m_len <= len) {
				len -= m->m_len;
				m->m_len = 0;
				m = m->m_next;
			} else {
				m->m_len -= len;
				m->m_data += len;
				len = 0;
			}
		}
		m = mp;
		if (mp->m_flags & M_PKTHDR)
			m->m_pkthdr.len -= (req_len - len);
	} else {
		/*
		 * Trim from tail.  Scan the mbuf chain,
		 * calculating its length and finding the last mbuf.
		 * If the adjustment only affects this mbuf, then just
		 * adjust and return.  Otherwise, rescan and truncate
		 * after the remaining size.
		 */
		len = -len;
		count = 0;
		for (;;) {
			count += m->m_len;
			if (m->m_next == (struct mbuf *)0)
				break;
			m = m->m_next;
		}
		if (m->m_len >= len) {
			m->m_len -= len;
			if (mp->m_flags & M_PKTHDR)
				mp->m_pkthdr.len -= len;
			return;
		}
		count -= len;
		if (count < 0)
			count = 0;
		/*
		 * Correct length for chain is "count".
		 * Find the mbuf with last data, adjust its length,
		 * and toss data from remaining mbufs on chain.
		 */
		m = mp;
		if (m->m_flags & M_PKTHDR)
			m->m_pkthdr.len = count;
		for (; m; m = m->m_next) {
			if (m->m_len >= count) {
				m->m_len = count;
				break;
			}
			count -= m->m_len;
		}
		while (m->m_next)
			(m = m->m_next) ->m_len = 0;
	}
}

/*
 * Rearange an mbuf chain so that len bytes are contiguous
 * and in the data area of an mbuf (so that mtod and dtom
 * will work for a structure of size len).  Returns the resulting
 * mbuf chain on success, frees it and returns null on failure.
 * If there is room, it will add up to max_protohdr-len extra bytes to the
 * contiguous region in an attempt to avoid being called next time.
 */
int MPFail;

struct mbuf *
m_pullup(n, len)
	struct mbuf *n;
	int len;
{
	struct mbuf *m;
	int count;
	int space;

	/*
	 * If first mbuf has no cluster, and has room for len bytes
	 * without shifting current data, pullup into it,
	 * otherwise allocate a new mbuf to prepend to the chain.
	 */
	if ((n->m_flags & M_EXT) == 0 &&
	    n->m_data + len < &n->m_dat[MLEN] && n->m_next) {
		if (n->m_len >= len)
			return (n);
		m = n;
		n = n->m_next;
		len -= m->m_len;
	} else {
		if (len > MHLEN)
			goto bad;
		MGET(m, M_DONTWAIT, n->m_type);
		if (m == 0)
			goto bad;
		m->m_len = 0;
		if (n->m_flags & M_PKTHDR) {
			M_COPY_PKTHDR(m, n);
			n->m_flags &= ~M_PKTHDR;
		}
	}
	space = &m->m_dat[MLEN] - (m->m_data + m->m_len);
	do {
		count = min(min(max(len, max_protohdr), space), n->m_len);
		memcpy(mtod(m, caddr_t) + m->m_len, mtod(n, caddr_t),
		  (unsigned)count);
		len -= count;
		m->m_len += count;
		n->m_len -= count;
		space -= count;
		if (n->m_len)
			n->m_data += count;
		else
			n = m_free(n);
	} while (len > 0 && n);
	if (len > 0) {
		(void) m_free(m);
		goto bad;
	}
	m->m_next = n;
	return (m);
bad:
	m_freem(n);
	MPFail++;
	return (0);
}

/*
 * Partition an mbuf chain in two pieces, returning the tail --
 * all but the first len0 bytes.  In case of failure, it returns NULL and
 * attempts to restore the chain to its original state.
 */
struct mbuf *
m_split(m0, len0, wait)
	struct mbuf *m0;
	int len0, wait;
{
	struct mbuf *m, *n;
	unsigned len = len0, remain, len_save;

	for (m = m0; m && len > m->m_len; m = m->m_next)
		len -= m->m_len;
	if (m == 0)
		return (0);
	remain = m->m_len - len;
	if (m0->m_flags & M_PKTHDR) {
		MGETHDR(n, wait, m0->m_type);
		if (n == 0)
			return (0);
		n->m_pkthdr.rcvif = m0->m_pkthdr.rcvif;
		n->m_pkthdr.len = m0->m_pkthdr.len - len0;
		len_save = m0->m_pkthdr.len;
		m0->m_pkthdr.len = len0;
		if (m->m_flags & M_EXT)
			goto extpacket;
		if (remain > MHLEN) {
			/* m can't be the lead packet */
			MH_ALIGN(n, 0);
			n->m_next = m_split(m, len, wait);
			if (n->m_next == 0) {
				(void) m_free(n);
				m0->m_pkthdr.len = len_save;
				return (0);
			} else
				return (n);
		} else
			MH_ALIGN(n, remain);
	} else if (remain == 0) {
		n = m->m_next;
		m->m_next = 0;
		return (n);
	} else {
		MGET(n, wait, m->m_type);
		if (n == 0)
			return (0);
		M_ALIGN(n, remain);
	}
extpacket:
	if (m->m_flags & M_EXT) {
		n->m_ext = m->m_ext;
		MCLADDREFERENCE(m, n);
		n->m_data = m->m_data + len;
	} else {
		memcpy(mtod(n, caddr_t), mtod(m, caddr_t) + len, remain);
	}
	n->m_len = remain;
	m->m_len = len;
	n->m_next = m->m_next;
	m->m_next = 0;
	return (n);
}
/*
 * Routine to copy from device local memory into mbufs.
 */
struct mbuf *
m_devget(buf, totlen, off0, ifp, copy)
	char *buf;
	int totlen, off0;
	struct ifnet *ifp;
	void (*copy) __P((const void *from, void *to, size_t len));
{
	struct mbuf *m;
	struct mbuf *top = 0, **mp = &top;
	int off = off0, len;
	char *cp;
	char *epkt;

	cp = buf;
	epkt = cp + totlen;
	if (off) {
		/*
		 * If 'off' is non-zero, packet is trailer-encapsulated,
		 * so we have to skip the type and length fields.
		 */
		cp += off + 2 * sizeof(u_int16_t);
		totlen -= 2 * sizeof(u_int16_t);
	}
	MGETHDR(m, M_DONTWAIT, MT_DATA);
	if (m == 0)
		return (0);
	m->m_pkthdr.rcvif = ifp;
	m->m_pkthdr.len = totlen;
	m->m_len = MHLEN;

	while (totlen > 0) {
		if (top) {
			MGET(m, M_DONTWAIT, MT_DATA);
			if (m == 0) {
				m_freem(top);
				return (0);
			}
			m->m_len = MLEN;
		}
		len = min(totlen, epkt - cp);
		if (len >= MINCLSIZE) {
			MCLGET(m, M_DONTWAIT);
			if ((m->m_flags & M_EXT) == 0) {
				m_free(m);
				m_freem(top);
				return (0);
			}
			m->m_len = len = min(len, MCLBYTES);
		} else {
			/*
			 * Place initial small packet/header at end of mbuf.
			 */
			if (len < m->m_len) {
				if (top == 0 && len + max_linkhdr <= m->m_len)
					m->m_data += max_linkhdr;
				m->m_len = len;
			} else
				len = m->m_len;
		}
		if (copy)
			copy(cp, mtod(m, caddr_t), (size_t)len);
		else
			memcpy(mtod(m, caddr_t), cp, (size_t)len);
		cp += len;
		*mp = m;
		mp = &m->m_next;
		totlen -= len;
		if (cp == epkt)
			cp = buf;
	}
	return (top);
}

/*
 * Copy data from a buffer back into the indicated mbuf chain,
 * starting "off" bytes from the beginning, extending the mbuf
 * chain if necessary.
 */
void
m_copyback(m0, off, len, cp)
	struct	mbuf *m0;
	int off;
	int len;
	caddr_t cp;
{
	int mlen;
	struct mbuf *m = m0, *n;
	int totlen = 0;

	if (m0 == 0)
		return;
	while (off > (mlen = m->m_len)) {
		off -= mlen;
		totlen += mlen;
		if (m->m_next == 0) {
			n = m_getclr(M_DONTWAIT, m->m_type);
			if (n == 0)
				goto out;
			n->m_len = min(MLEN, len + off);
			m->m_next = n;
		}
		m = m->m_next;
	}
	while (len > 0) {
		mlen = min (m->m_len - off, len);
		memcpy(mtod(m, caddr_t) + off, cp, (unsigned)mlen);
		cp += mlen;
		len -= mlen;
		mlen += off;
		off = 0;
		totlen += mlen;
		if (len == 0)
			break;
		if (m->m_next == 0) {
			n = m_get(M_DONTWAIT, m->m_type);
			if (n == 0)
				break;
			n->m_len = min(MLEN, len);
			m->m_next = n;
		}
		m = m->m_next;
	}
out:	if (((m = m0)->m_flags & M_PKTHDR) && (m->m_pkthdr.len < totlen))
		m->m_pkthdr.len = totlen;
}