freebsd/sys/kern/uipc_mbuf.c

/*
 * 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	8.2 (Berkeley) 1/4/94
 */

#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");

#include "opt_mac.h"
#include "opt_param.h"
#include "opt_mbuf_stress_test.h"

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/kernel.h>
#include <sys/lock.h>
#include <sys/mac.h>
#include <sys/malloc.h>
#include <sys/mbuf.h>
#include <sys/sysctl.h>
#include <sys/domain.h>
#include <sys/protosw.h>

int	max_linkhdr;
int	max_protohdr;
int	max_hdr;
int	max_datalen;
#ifdef MBUF_STRESS_TEST
int	m_defragpackets;
int	m_defragbytes;
int	m_defraguseless;
int	m_defragfailure;
int	m_defragrandomfailures;
#endif

/*
 * sysctl(8) exported objects
 */
SYSCTL_DECL(_kern_ipc);
SYSCTL_INT(_kern_ipc, KIPC_MAX_LINKHDR, max_linkhdr, CTLFLAG_RW,
	   &max_linkhdr, 0, "");
SYSCTL_INT(_kern_ipc, KIPC_MAX_PROTOHDR, max_protohdr, CTLFLAG_RW,
	   &max_protohdr, 0, "");
SYSCTL_INT(_kern_ipc, KIPC_MAX_HDR, max_hdr, CTLFLAG_RW, &max_hdr, 0, "");
SYSCTL_INT(_kern_ipc, KIPC_MAX_DATALEN, max_datalen, CTLFLAG_RW,
	   &max_datalen, 0, "");
#ifdef MBUF_STRESS_TEST
SYSCTL_INT(_kern_ipc, OID_AUTO, m_defragpackets, CTLFLAG_RD,
	   &m_defragpackets, 0, "");
SYSCTL_INT(_kern_ipc, OID_AUTO, m_defragbytes, CTLFLAG_RD,
	   &m_defragbytes, 0, "");
SYSCTL_INT(_kern_ipc, OID_AUTO, m_defraguseless, CTLFLAG_RD,
	   &m_defraguseless, 0, "");
SYSCTL_INT(_kern_ipc, OID_AUTO, m_defragfailure, CTLFLAG_RD,
	   &m_defragfailure, 0, "");
SYSCTL_INT(_kern_ipc, OID_AUTO, m_defragrandomfailures, CTLFLAG_RW,
	   &m_defragrandomfailures, 0, "");
#endif

/*
 * "Move" mbuf pkthdr from "from" to "to".
 * "from" must have M_PKTHDR set, and "to" must be empty.
 */
void
m_move_pkthdr(struct mbuf *to, struct mbuf *from)
{

#if 0
	/* see below for why these are not enabled */
	M_ASSERTPKTHDR(to);
	/* Note: with MAC, this may not be a good assertion. */
	KASSERT(SLIST_EMPTY(&to->m_pkthdr.tags),
	    ("m_move_pkthdr: to has tags"));
#endif
	KASSERT((to->m_flags & M_EXT) == 0, ("m_move_pkthdr: to has cluster"));
#ifdef MAC
	/*
	 * XXXMAC: It could be this should also occur for non-MAC?
	 */
	if (to->m_flags & M_PKTHDR)
		m_tag_delete_chain(to, NULL);
#endif
	to->m_flags = from->m_flags & M_COPYFLAGS;
	to->m_data = to->m_pktdat;
	to->m_pkthdr = from->m_pkthdr;		/* especially tags */
	SLIST_INIT(&from->m_pkthdr.tags);	/* purge tags from src */
	from->m_flags &= ~M_PKTHDR;
}

/*
 * Duplicate "from"'s mbuf pkthdr in "to".
 * "from" must have M_PKTHDR set, and "to" must be empty.
 * In particular, this does a deep copy of the packet tags.
 */
int
m_dup_pkthdr(struct mbuf *to, struct mbuf *from, int how)
{

#if 0
	/*
	 * The mbuf allocator only initializes the pkthdr
	 * when the mbuf is allocated with MGETHDR. Many users
	 * (e.g. m_copy*, m_prepend) use MGET and then
	 * smash the pkthdr as needed causing these
	 * assertions to trip.  For now just disable them.
	 */
	M_ASSERTPKTHDR(to);
	/* Note: with MAC, this may not be a good assertion. */
	KASSERT(SLIST_EMPTY(&to->m_pkthdr.tags), ("m_dup_pkthdr: to has tags"));
#endif
#ifdef MAC
	if (to->m_flags & M_PKTHDR)
		m_tag_delete_chain(to, NULL);
#endif
	to->m_flags = (from->m_flags & M_COPYFLAGS) | (to->m_flags & M_EXT);
	if ((to->m_flags & M_EXT) == 0)
		to->m_data = to->m_pktdat;
	to->m_pkthdr = from->m_pkthdr;
	SLIST_INIT(&to->m_pkthdr.tags);
	return (m_tag_copy_chain(to, from, MBTOM(how)));
}

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

	MGET(mn, how, m->m_type);
	if (mn == NULL) {
		m_freem(m);
		return (NULL);
	}
	if (m->m_flags & M_PKTHDR)
		M_MOVE_PKTHDR(mn, m);
	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_TRYWAIT/M_DONTWAIT from caller.
 * Note that the copy is read-only, because clusters are not copied,
 * only their reference counts are incremented.
 */
struct mbuf *
m_copym(struct mbuf *m, int off0, int len, int wait)
{
	struct mbuf *n, **np;
	int off = off0;
	struct mbuf *top;
	int copyhdr = 0;

	KASSERT(off >= 0, ("m_copym, negative off %d", off));
	KASSERT(len >= 0, ("m_copym, negative len %d", len));
	if (off == 0 && m->m_flags & M_PKTHDR)
		copyhdr = 1;
	while (off > 0) {
		KASSERT(m != NULL, ("m_copym, offset > size of mbuf chain"));
		if (off < m->m_len)
			break;
		off -= m->m_len;
		m = m->m_next;
	}
	np = &top;
	top = 0;
	while (len > 0) {
		if (m == NULL) {
			KASSERT(len == M_COPYALL, 
			    ("m_copym, length > size of mbuf chain"));
			break;
		}
		MGET(n, wait, m->m_type);
		*np = n;
		if (n == NULL)
			goto nospace;
		if (copyhdr) {
			if (!m_dup_pkthdr(n, m, wait))
				goto nospace;
			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) {
			n->m_data = m->m_data + off;
			n->m_ext = m->m_ext;
			n->m_flags |= M_EXT;
			MEXT_ADD_REF(m);
		} else
			bcopy(mtod(m, caddr_t)+off, mtod(n, caddr_t),
			    (u_int)n->m_len);
		if (len != M_COPYALL)
			len -= n->m_len;
		off = 0;
		m = m->m_next;
		np = &n->m_next;
	}
	if (top == NULL)
		mbstat.m_mcfail++;	/* XXX: No consistency. */

	return (top);
nospace:
	m_freem(top);
	mbstat.m_mcfail++;	/* XXX: No consistency. */
	return (NULL);
}

/*
 * Copy an entire packet, including header (which must be present).
 * An optimization of the common case `m_copym(m, 0, M_COPYALL, how)'.
 * Note that the copy is read-only, because clusters are not copied,
 * only their reference counts are incremented.
 * Preserve alignment of the first mbuf so if the creator has left
 * some room at the beginning (e.g. for inserting protocol headers)
 * the copies still have the room available.
 */
struct mbuf *
m_copypacket(struct mbuf *m, int how)
{
	struct mbuf *top, *n, *o;

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

	if (!m_dup_pkthdr(n, m, how))
		goto nospace;
	n->m_len = m->m_len;
	if (m->m_flags & M_EXT) {
		n->m_data = m->m_data;
		n->m_ext = m->m_ext;
		n->m_flags |= M_EXT;
		MEXT_ADD_REF(m);
	} else {
		n->m_data = n->m_pktdat + (m->m_data - m->m_pktdat );
		bcopy(mtod(m, char *), mtod(n, char *), n->m_len);
	}

	m = m->m_next;
	while (m) {
		MGET(o, how, m->m_type);
		if (o == NULL)
			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;
			n->m_flags |= M_EXT;
			MEXT_ADD_REF(m);
		} else {
			bcopy(mtod(m, char *), mtod(n, char *), n->m_len);
		}

		m = m->m_next;
	}
	return top;
nospace:
	m_freem(top);
	mbstat.m_mcfail++;	/* XXX: No consistency. */ 
	return (NULL);
}

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

	KASSERT(off >= 0, ("m_copydata, negative off %d", off));
	KASSERT(len >= 0, ("m_copydata, negative len %d", len));
	while (off > 0) {
		KASSERT(m != NULL, ("m_copydata, offset > size of mbuf chain"));
		if (off < m->m_len)
			break;
		off -= m->m_len;
		m = m->m_next;
	}
	while (len > 0) {
		KASSERT(m != NULL, ("m_copydata, length > size of mbuf chain"));
		count = min(m->m_len - off, len);
		bcopy(mtod(m, caddr_t) + off, cp, count);
		len -= count;
		cp += count;
		off = 0;
		m = m->m_next;
	}
}

/*
 * Copy a packet header mbuf chain into a completely new chain, including
 * copying any mbuf clusters.  Use this instead of m_copypacket() when
 * you need a writable copy of an mbuf chain.
 */
struct mbuf *
m_dup(struct mbuf *m, int how)
{
	struct mbuf **p, *top = NULL;
	int remain, moff, nsize;

	/* Sanity check */
	if (m == NULL)
		return (NULL);
	M_ASSERTPKTHDR(m);

	/* While there's more data, get a new mbuf, tack it on, and fill it */
	remain = m->m_pkthdr.len;
	moff = 0;
	p = &top;
	while (remain > 0 || top == NULL) {	/* allow m->m_pkthdr.len == 0 */
		struct mbuf *n;

		/* Get the next new mbuf */
		MGET(n, how, m->m_type);
		if (n == NULL)
			goto nospace;
		if (top == NULL) {		/* first one, must be PKTHDR */
			if (!m_dup_pkthdr(n, m, how))
				goto nospace;
			nsize = MHLEN;
		} else				/* not the first one */
			nsize = MLEN;
		if (remain >= MINCLSIZE) {
			MCLGET(n, how);
			if ((n->m_flags & M_EXT) == 0) {
				(void)m_free(n);
				goto nospace;
			}
			nsize = MCLBYTES;
		}
		n->m_len = 0;

		/* Link it into the new chain */
		*p = n;
		p = &n->m_next;

		/* Copy data from original mbuf(s) into new mbuf */
		while (n->m_len < nsize && m != NULL) {
			int chunk = min(nsize - n->m_len, m->m_len - moff);

			bcopy(m->m_data + moff, n->m_data + n->m_len, chunk);
			moff += chunk;
			n->m_len += chunk;
			remain -= chunk;
			if (moff == m->m_len) {
				m = m->m_next;
				moff = 0;
			}
		}

		/* Check correct total mbuf length */
		KASSERT((remain > 0 && m != NULL) || (remain == 0 && m == NULL),
		    	("%s: bogus m_pkthdr.len", __func__));
	}
	return (top);

nospace:
	m_freem(top);
	mbstat.m_mcfail++;	/* XXX: No consistency. */
	return (NULL);
}

/*
 * 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(struct mbuf *m, struct mbuf *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 */
		bcopy(mtod(n, caddr_t), mtod(m, caddr_t) + m->m_len,
		    (u_int)n->m_len);
		m->m_len += n->m_len;
		n = m_free(n);
	}
}

void
m_adj(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.
 */
struct mbuf *
m_pullup(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 == NULL)
			goto bad;
		m->m_len = 0;
		if (n->m_flags & M_PKTHDR)
			M_MOVE_PKTHDR(m, n);
	}
	space = &m->m_dat[MLEN] - (m->m_data + m->m_len);
	do {
		count = min(min(max(len, max_protohdr), space), n->m_len);
		bcopy(mtod(n, caddr_t), mtod(m, caddr_t) + m->m_len,
		  (u_int)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);
	mbstat.m_mpfail++;	/* XXX: No consistency. */
	return (NULL);
}

/*
 * 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.
 *
 * Note that the resulting mbufs might be read-only, because the new
 * mbuf can end up sharing an mbuf cluster with the original mbuf if
 * the "breaking point" happens to lie within a cluster mbuf. Use the
 * M_WRITABLE() macro to check for this case.
 */
struct mbuf *
m_split(struct mbuf *m0, int len0, int wait)
{
	struct mbuf *m, *n;
	u_int len = len0, remain;

	for (m = m0; m && len > m->m_len; m = m->m_next)
		len -= m->m_len;
	if (m == NULL)
		return (NULL);
	remain = m->m_len - len;
	if (m0->m_flags & M_PKTHDR) {
		MGETHDR(n, wait, m0->m_type);
		if (n == NULL)
			return (NULL);
		n->m_pkthdr.rcvif = m0->m_pkthdr.rcvif;
		n->m_pkthdr.len = m0->m_pkthdr.len - len0;
		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 == NULL) {
				(void) m_free(n);
				return (NULL);
			} else {
				n->m_len = 0;
				return (n);
			}
		} else
			MH_ALIGN(n, remain);
	} else if (remain == 0) {
		n = m->m_next;
		m->m_next = NULL;
		return (n);
	} else {
		MGET(n, wait, m->m_type);
		if (n == NULL)
			return (NULL);
		M_ALIGN(n, remain);
	}
extpacket:
	if (m->m_flags & M_EXT) {
		n->m_flags |= M_EXT;
		n->m_ext = m->m_ext;
		MEXT_ADD_REF(m);
		n->m_data = m->m_data + len;
	} else {
		bcopy(mtod(m, caddr_t) + len, mtod(n, caddr_t), remain);
	}
	n->m_len = remain;
	m->m_len = len;
	n->m_next = m->m_next;
	m->m_next = NULL;
	return (n);
}
/*
 * Routine to copy from device local memory into mbufs.
 * Note that `off' argument is offset into first mbuf of target chain from
 * which to begin copying the data to.
 */
struct mbuf *
m_devget(char *buf, int totlen, int off, struct ifnet *ifp,
	 void (*copy)(char *from, caddr_t to, u_int len))
{
	struct mbuf *m;
	struct mbuf *top = 0, **mp = &top;
	int len;

	if (off < 0 || off > MHLEN)
		return (NULL);

	MGETHDR(m, M_DONTWAIT, MT_DATA);
	if (m == NULL)
		return (NULL);
	m->m_pkthdr.rcvif = ifp;
	m->m_pkthdr.len = totlen;
	len = MHLEN;

	while (totlen > 0) {
		if (top) {
			MGET(m, M_DONTWAIT, MT_DATA);
			if (m == NULL) {
				m_freem(top);
				return (NULL);
			}
			len = MLEN;
		}
		if (totlen + off >= MINCLSIZE) {
			MCLGET(m, M_DONTWAIT);
			if (m->m_flags & M_EXT)
				len = MCLBYTES;
		} else {
			/*
			 * Place initial small packet/header at end of mbuf.
			 */
			if (top == NULL && totlen + off + max_linkhdr <= len) {
				m->m_data += max_linkhdr;
				len -= max_linkhdr;
			}
		}
		if (off) {
			m->m_data += off;
			len -= off;
			off = 0;
		}
		m->m_len = len = min(totlen, len);
		if (copy)
			copy(buf, mtod(m, caddr_t), (u_int)len);
		else
			bcopy(buf, mtod(m, caddr_t), (u_int)len);
		buf += len;
		*mp = m;
		mp = &m->m_next;
		totlen -= len;
	}
	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(struct mbuf *m0, int off, int len, caddr_t cp)
{
	int mlen;
	struct mbuf *m = m0, *n;
	int totlen = 0;

	if (m0 == NULL)
		return;
	while (off > (mlen = m->m_len)) {
		off -= mlen;
		totlen += mlen;
		if (m->m_next == NULL) {
			n = m_get_clrd(M_DONTWAIT, m->m_type);
			if (n == NULL)
				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);
		bcopy(cp, off + mtod(m, caddr_t), (u_int)mlen);
		cp += mlen;
		len -= mlen;
		mlen += off;
		off = 0;
		totlen += mlen;
		if (len == 0)
			break;
		if (m->m_next == NULL) {
			n = m_get(M_DONTWAIT, m->m_type);
			if (n == NULL)
				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;
}

void
m_print(const struct mbuf *m)
{
	int len;
	const struct mbuf *m2;

	len = m->m_pkthdr.len;
	m2 = m;
	while (len) {
		printf("%p %*D\n", m2, m2->m_len, (u_char *)m2->m_data, "-");
		len -= m2->m_len;
		m2 = m2->m_next;
	}
	return;
}

u_int
m_fixhdr(struct mbuf *m0)
{
	u_int len;

	len = m_length(m0, NULL);
	m0->m_pkthdr.len = len;
	return (len);
}

u_int
m_length(struct mbuf *m0, struct mbuf **last)
{
	struct mbuf *m;
	u_int len;

	len = 0;
	for (m = m0; m != NULL; m = m->m_next) {
		len += m->m_len;
		if (m->m_next == NULL)
			break;
	}
	if (last != NULL)
		*last = m;
	return (len);
}

/*
 * Defragment a mbuf chain, returning the shortest possible
 * chain of mbufs and clusters.  If allocation fails and
 * this cannot be completed, NULL will be returned, but
 * the passed in chain will be unchanged.  Upon success,
 * the original chain will be freed, and the new chain
 * will be returned.
 *
 * If a non-packet header is passed in, the original
 * mbuf (chain?) will be returned unharmed.
 */
struct mbuf *
m_defrag(struct mbuf *m0, int how)
{
	struct mbuf	*m_new = NULL, *m_final = NULL;
	int		progress = 0, length;

	if (!(m0->m_flags & M_PKTHDR))
		return (m0);

#ifdef MBUF_STRESS_TEST
	if (m_defragrandomfailures) {
		int temp = arc4random() & 0xff;
		if (temp == 0xba)
			goto nospace;
	}
#endif
	
	if (m0->m_pkthdr.len > MHLEN)
		m_final = m_getcl(how, MT_DATA, M_PKTHDR);
	else
		m_final = m_gethdr(how, MT_DATA);

	if (m_final == NULL)
		goto nospace;

	if (m_dup_pkthdr(m_final, m0, how) == NULL)
		goto nospace;

	m_new = m_final;

	while (progress < m0->m_pkthdr.len) {
		length = m0->m_pkthdr.len - progress;
		if (length > MCLBYTES)
			length = MCLBYTES;

		if (m_new == NULL) {
			if (length > MLEN)
				m_new = m_getcl(how, MT_DATA, 0);
			else
				m_new = m_get(how, MT_DATA);
			if (m_new == NULL)
				goto nospace;
		}

		m_copydata(m0, progress, length, mtod(m_new, caddr_t));
		progress += length;
		m_new->m_len = length;
		if (m_new != m_final)
			m_cat(m_final, m_new);
		m_new = NULL;
	}
#ifdef MBUF_STRESS_TEST
	if (m0->m_next == NULL)
		m_defraguseless++;
#endif
	m_freem(m0);
	m0 = m_final;
#ifdef MBUF_STRESS_TEST
	m_defragpackets++;
	m_defragbytes += m0->m_pkthdr.len;
#endif
	return (m0);
nospace:
#ifdef MBUF_STRESS_TEST
	m_defragfailure++;
#endif
	if (m_new)
		m_free(m_new);
	if (m_final)
		m_freem(m_final);
	return (NULL);
}