From: deraadt <deraadt@openbsd.org>
Date: Thu, 14 Dec 1995 06:50:40 +0000 (+0000)
Subject: from netbsd:
X-Git-Url: http://artulab.com/gitweb/?a=commitdiff_plain;h=f3eb6470a8a3c103ad44fa724c6c7a9501818ea2;p=openbsd

from netbsd:
make netinet work on systems where pointers and longs are 64 bits
(like the alpha).  Biggest problem: IP headers were overlayed with
structure which included pointers, and which therefore didn't overlay
properly on 64-bit machines.  Solution: instead of threading pointers
through IP header overlays, add a "queue element" structure to do
the threading, and point it at the ip headers.
---

diff --git a/sys/netinet/ip_input.c b/sys/netinet/ip_input.c
index 100f1865298..f7457607ee8 100644
--- a/sys/netinet/ip_input.c
+++ b/sys/netinet/ip_input.c
@@ -1,4 +1,4 @@
-/*	$NetBSD: ip_input.c,v 1.24 1995/08/12 23:59:36 mycroft Exp $	*/
+/*	$NetBSD: ip_input.c,v 1.25 1995/11/21 01:07:34 cgd Exp $	*/
 
 /*
  * Copyright (c) 1982, 1986, 1988, 1993
@@ -45,7 +45,6 @@
 #include <sys/errno.h>
 #include <sys/time.h>
 #include <sys/kernel.h>
-#include <sys/syslog.h>
 
 #include <net/if.h>
 #include <net/route.h>
@@ -70,7 +69,6 @@
 #endif
 int	ipforwarding = IPFORWARDING;
 int	ipsendredirects = IPSENDREDIRECTS;
-int	ip_dosourceroute = 0;	/* no source routing unless sysctl'd to enable */
 int	ip_defttl = IPDEFTTL;
 #ifdef DIAGNOSTIC
 int	ipprintfs = 0;
@@ -83,18 +81,6 @@ int	ipqmaxlen = IFQ_MAXLEN;
 struct	in_ifaddrhead in_ifaddr;
 struct	ifqueue ipintrq;
 
-char *
-inet_ntoa(ina)
-	struct in_addr ina;
-{
-	static char buf[4*sizeof "123"];
-	unsigned char *ucp = (unsigned char *)&ina;
-
-	sprintf(buf, "%d.%d.%d.%d", ucp[0] & 0xff, ucp[1] & 0xff,
-	    ucp[2] & 0xff, ucp[3] & 0xff);
-	return (buf);
-}
-
 /*
  * We need to save the IP options in case a protocol wants to respond
  * to an incoming packet over the same route if the packet got here
@@ -131,7 +117,7 @@ ip_init()
 		if (pr->pr_domain->dom_family == PF_INET &&
 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
 			ip_protox[pr->pr_protocol] = pr - inetsw;
-	ipq.next = ipq.prev = &ipq;
+	LIST_INIT(&ipq);
 	ip_id = time.tv_sec & 0xffff;
 	ipintrq.ifq_maxlen = ipqmaxlen;
 	TAILQ_INIT(&in_ifaddr);
@@ -151,7 +137,8 @@ ipintr()
 	register struct mbuf *m;
 	register struct ipq *fp;
 	register struct in_ifaddr *ia;
-	int hlen, s;
+	struct ipqent *ipqe;
+	int hlen, mff, s;
 
 next:
 	/*
@@ -352,7 +339,7 @@ ours:
 		 * Look for queue of fragments
 		 * of this datagram.
 		 */
-		for (fp = ipq.next; fp != &ipq; fp = fp->next)
+		for (fp = ipq.lh_first; fp != NULL; fp = fp->ipq_q.le_next)
 			if (ip->ip_id == fp->ipq_id &&
 			    ip->ip_src.s_addr == fp->ipq_src.s_addr &&
 			    ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
@@ -363,21 +350,20 @@ found:
 
 		/*
 		 * Adjust ip_len to not reflect header,
-		 * set ip_mff if more fragments are expected,
+		 * set ipqe_mff if more fragments are expected,
 		 * convert offset of this to bytes.
 		 */
 		ip->ip_len -= hlen;
-		((struct ipasfrag *)ip)->ipf_mff &= ~1;
-		if (ip->ip_off & IP_MF) {
-			/*
-			 * Make sure that fragments have a data length
+		mff = (ip->ip_off & IP_MF) != 0;
+		if (mff) {
+		        /*
+		         * Make sure that fragments have a data length
 			 * that's a non-zero multiple of 8 bytes.
-			 */
+		         */
 			if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
 				ipstat.ips_badfrags++;
 				goto bad;
 			}
-			((struct ipasfrag *)ip)->ipf_mff |= 1;
 		}
 		ip->ip_off <<= 3;
 
@@ -386,9 +372,17 @@ found:
 		 * or if this is not the first fragment,
 		 * attempt reassembly; if it succeeds, proceed.
 		 */
-		if (((struct ipasfrag *)ip)->ipf_mff & 1 || ip->ip_off) {
+		if (mff || ip->ip_off) {
 			ipstat.ips_fragments++;
-			ip = ip_reass((struct ipasfrag *)ip, fp);
+			MALLOC(ipqe, struct ipqent *, sizeof (struct ipqent),
+			    M_IPQ, M_NOWAIT);
+			if (ipqe == NULL) {
+				ipstat.ips_rcvmemdrop++;
+				goto bad;
+			}
+			ipqe->ipqe_mff = mff;
+			ipqe->ipqe_ip = ip;
+			ip = ip_reass(ipqe, fp);
 			if (ip == 0)
 				goto next;
 			ipstat.ips_reassembled++;
@@ -417,14 +411,15 @@ bad:
  * is given as fp; otherwise have to make a chain.
  */
 struct ip *
-ip_reass(ip, fp)
-	register struct ipasfrag *ip;
+ip_reass(ipqe, fp)
+	register struct ipqent *ipqe;
 	register struct ipq *fp;
 {
-	register struct mbuf *m = dtom(ip);
-	register struct ipasfrag *q;
+	register struct mbuf *m = dtom(ipqe->ipqe_ip);
+	register struct ipqent *nq, *p, *q;
+	struct ip *ip;
 	struct mbuf *t;
-	int hlen = ip->ip_hl << 2;
+	int hlen = ipqe->ipqe_ip->ip_hl << 2;
 	int i, next;
 
 	/*
@@ -441,22 +436,23 @@ ip_reass(ip, fp)
 		if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL)
 			goto dropfrag;
 		fp = mtod(t, struct ipq *);
-		insque(fp, &ipq);
+		LIST_INSERT_HEAD(&ipq, fp, ipq_q);
 		fp->ipq_ttl = IPFRAGTTL;
-		fp->ipq_p = ip->ip_p;
-		fp->ipq_id = ip->ip_id;
-		fp->ipq_next = fp->ipq_prev = (struct ipasfrag *)fp;
-		fp->ipq_src = ((struct ip *)ip)->ip_src;
-		fp->ipq_dst = ((struct ip *)ip)->ip_dst;
-		q = (struct ipasfrag *)fp;
+		fp->ipq_p = ipqe->ipqe_ip->ip_p;
+		fp->ipq_id = ipqe->ipqe_ip->ip_id;
+		LIST_INIT(&fp->ipq_fragq);
+		fp->ipq_src = ipqe->ipqe_ip->ip_src;
+		fp->ipq_dst = ipqe->ipqe_ip->ip_dst;
+		p = NULL;
 		goto insert;
 	}
 
 	/*
 	 * Find a segment which begins after this one does.
 	 */
-	for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = q->ipf_next)
-		if (q->ip_off > ip->ip_off)
+	for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL;
+	    p = q, q = q->ipqe_q.le_next)
+		if (q->ipqe_ip->ip_off > ipqe->ipqe_ip->ip_off)
 			break;
 
 	/*
@@ -464,14 +460,15 @@ ip_reass(ip, fp)
 	 * our data already.  If so, drop the data from the incoming
 	 * segment.  If it provides all of our data, drop us.
 	 */
-	if (q->ipf_prev != (struct ipasfrag *)fp) {
-		i = q->ipf_prev->ip_off + q->ipf_prev->ip_len - ip->ip_off;
+	if (p != NULL) {
+		i = p->ipqe_ip->ip_off + p->ipqe_ip->ip_len -
+		    ipqe->ipqe_ip->ip_off;
 		if (i > 0) {
-			if (i >= ip->ip_len)
+			if (i >= ipqe->ipqe_ip->ip_len)
 				goto dropfrag;
-			m_adj(dtom(ip), i);
-			ip->ip_off += i;
-			ip->ip_len -= i;
+			m_adj(dtom(ipqe->ipqe_ip), i);
+			ipqe->ipqe_ip->ip_off += i;
+			ipqe->ipqe_ip->ip_len -= i;
 		}
 	}
 
@@ -479,17 +476,20 @@ ip_reass(ip, fp)
 	 * While we overlap succeeding segments trim them or,
 	 * if they are completely covered, dequeue them.
 	 */
-	while (q != (struct ipasfrag *)fp && ip->ip_off + ip->ip_len > q->ip_off) {
-		i = (ip->ip_off + ip->ip_len) - q->ip_off;
-		if (i < q->ip_len) {
-			q->ip_len -= i;
-			q->ip_off += i;
-			m_adj(dtom(q), i);
+	for (; q != NULL && ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len >
+	    q->ipqe_ip->ip_off; q = nq) {
+		i = (ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len) -
+		    q->ipqe_ip->ip_off;
+		if (i < q->ipqe_ip->ip_len) {
+			q->ipqe_ip->ip_len -= i;
+			q->ipqe_ip->ip_off += i;
+			m_adj(dtom(q->ipqe_ip), i);
 			break;
 		}
-		q = q->ipf_next;
-		m_freem(dtom(q->ipf_prev));
-		ip_deq(q->ipf_prev);
+		nq = q->ipqe_q.le_next;
+		m_freem(dtom(q->ipqe_ip));
+		LIST_REMOVE(q, ipqe_q);
+		FREE(q, M_IPQ);
 	}
 
 insert:
@@ -497,28 +497,36 @@ insert:
 	 * Stick new segment in its place;
 	 * check for complete reassembly.
 	 */
-	ip_enq(ip, q->ipf_prev);
+	if (p == NULL) {
+		LIST_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q);
+	} else {
+		LIST_INSERT_AFTER(p, ipqe, ipqe_q);
+	}
 	next = 0;
-	for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = q->ipf_next) {
-		if (q->ip_off != next)
+	for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL;
+	    p = q, q = q->ipqe_q.le_next) {
+		if (q->ipqe_ip->ip_off != next)
 			return (0);
-		next += q->ip_len;
+		next += q->ipqe_ip->ip_len;
 	}
-	if (q->ipf_prev->ipf_mff & 1)
+	if (p->ipqe_mff)
 		return (0);
 
 	/*
 	 * Reassembly is complete; concatenate fragments.
 	 */
-	q = fp->ipq_next;
-	m = dtom(q);
+	q = fp->ipq_fragq.lh_first;
+	ip = q->ipqe_ip;
+	m = dtom(q->ipqe_ip);
 	t = m->m_next;
 	m->m_next = 0;
 	m_cat(m, t);
-	q = q->ipf_next;
-	while (q != (struct ipasfrag *)fp) {
-		t = dtom(q);
-		q = q->ipf_next;
+	nq = q->ipqe_q.le_next;
+	FREE(q, M_IPQ);
+	for (q = nq; q != NULL; q = nq) {
+		t = dtom(q->ipqe_ip);
+		nq = q->ipqe_q.le_next;
+		FREE(q, M_IPQ);
 		m_cat(m, t);
 	}
 
@@ -528,14 +536,11 @@ insert:
 	 * dequeue and discard fragment reassembly header.
 	 * Make header visible.
 	 */
-	ip = fp->ipq_next;
 	ip->ip_len = next;
-	ip->ipf_mff &= ~1;
-	((struct ip *)ip)->ip_src = fp->ipq_src;
-	((struct ip *)ip)->ip_dst = fp->ipq_dst;
-	remque(fp);
+	ip->ip_src = fp->ipq_src;
+	ip->ip_dst = fp->ipq_dst;
+	LIST_REMOVE(fp, ipq_q);
 	(void) m_free(dtom(fp));
-	m = dtom(ip);
 	m->m_len += (ip->ip_hl << 2);
 	m->m_data -= (ip->ip_hl << 2);
 	/* some debugging cruft by sklower, below, will go away soon */
@@ -545,11 +550,12 @@ insert:
 			plen += m->m_len;
 		t->m_pkthdr.len = plen;
 	}
-	return ((struct ip *)ip);
+	return (ip);
 
 dropfrag:
 	ipstat.ips_fragdropped++;
 	m_freem(m);
+	FREE(ipqe, M_IPQ);
 	return (0);
 }
 
@@ -561,44 +567,18 @@ void
 ip_freef(fp)
 	struct ipq *fp;
 {
-	register struct ipasfrag *q, *p;
+	register struct ipqent *q, *p;
 
-	for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = p) {
-		p = q->ipf_next;
-		ip_deq(q);
-		m_freem(dtom(q));
+	for (q = fp->ipq_fragq.lh_first; q != NULL; q = p) {
+		p = q->ipqe_q.le_next;
+		m_freem(dtom(q->ipqe_ip));
+		LIST_REMOVE(q, ipqe_q);
+		FREE(q, M_IPQ);
 	}
-	remque(fp);
+	LIST_REMOVE(fp, ipq_q);
 	(void) m_free(dtom(fp));
 }
 
-/*
- * Put an ip fragment on a reassembly chain.
- * Like insque, but pointers in middle of structure.
- */
-void
-ip_enq(p, prev)
-	register struct ipasfrag *p, *prev;
-{
-
-	p->ipf_prev = prev;
-	p->ipf_next = prev->ipf_next;
-	prev->ipf_next->ipf_prev = p;
-	prev->ipf_next = p;
-}
-
-/*
- * To ip_enq as remque is to insque.
- */
-void
-ip_deq(p)
-	register struct ipasfrag *p;
-{
-
-	p->ipf_prev->ipf_next = p->ipf_next;
-	p->ipf_next->ipf_prev = p->ipf_prev;
-}
-
 /*
  * IP timer processing;
  * if a timer expires on a reassembly
@@ -607,20 +587,14 @@ ip_deq(p)
 void
 ip_slowtimo()
 {
-	register struct ipq *fp;
+	register struct ipq *fp, *nfp;
 	int s = splsoftnet();
 
-	fp = ipq.next;
-	if (fp == 0) {
-		splx(s);
-		return;
-	}
-	while (fp != &ipq) {
-		--fp->ipq_ttl;
-		fp = fp->next;
-		if (fp->prev->ipq_ttl == 0) {
+	for (fp = ipq.lh_first; fp != NULL; fp = nfp) {
+		nfp = fp->ipq_q.le_next;
+		if (--fp->ipq_ttl == 0) {
 			ipstat.ips_fragtimeout++;
-			ip_freef(fp->prev);
+			ip_freef(fp);
 		}
 	}
 	splx(s);
@@ -633,9 +607,9 @@ void
 ip_drain()
 {
 
-	while (ipq.next != &ipq) {
+	while (ipq.lh_first != NULL) {
 		ipstat.ips_fragdropped++;
-		ip_freef(ipq.next);
+		ip_freef(ipq.lh_first);
 	}
 }
 
@@ -716,19 +690,6 @@ ip_dooptions(m)
 				save_rte(cp, ip->ip_src);
 				break;
 			}
-
-			if (!ip_dosourceroute) {
-				char buf[4*sizeof "123"];
-
-				strcpy(buf, inet_ntoa(ip->ip_dst));
-				log(LOG_WARNING,
-				    "attempted source route from %s to %s\n",
-				    inet_ntoa(ip->ip_src), buf);
-				type = ICMP_UNREACH;
-				code = ICMP_UNREACH_SRCFAIL;
-				goto bad;
-			}
-
 			/*
 			 * locate outgoing interface
 			 */
@@ -1032,8 +993,8 @@ ip_forward(m, srcrt)
 	dest = 0;
 #ifdef DIAGNOSTIC
 	if (ipprintfs)
-		printf("forward: src %lx dst %x ttl %x\n", ip->ip_src.s_addr,
-			ip->ip_dst.s_addr, ip->ip_ttl);
+		printf("forward: src %x dst %x ttl %x\n", ip->ip_src,
+			ip->ip_dst, ip->ip_ttl);
 #endif
 	if (m->m_flags & M_BCAST || in_canforward(ip->ip_dst) == 0) {
 		ipstat.ips_cantforward++;
@@ -1096,7 +1057,7 @@ ip_forward(m, srcrt)
 		    code = ICMP_REDIRECT_HOST;
 #ifdef DIAGNOSTIC
 		    if (ipprintfs)
-			printf("redirect (%d) to %lx\n", code, (u_int32_t)dest);
+		        printf("redirect (%d) to %lx\n", code, (u_int32_t)dest);
 #endif
 		}
 	}
@@ -1178,8 +1139,6 @@ ip_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
 	case IPCTL_DEFMTU:
 		return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_mtu));
 #endif
-	case IPCTL_SOURCEROUTE:
-		return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_dosourceroute));
 	default:
 		return (EOPNOTSUPP);
 	}
diff --git a/sys/netinet/ip_var.h b/sys/netinet/ip_var.h
index 58707e805e0..c78365d93f8 100644
--- a/sys/netinet/ip_var.h
+++ b/sys/netinet/ip_var.h
@@ -1,4 +1,4 @@
-/*	$NetBSD: ip_var.h,v 1.14 1995/06/12 00:47:47 mycroft Exp $	*/
+/*	$NetBSD: ip_var.h,v 1.15 1995/11/21 01:07:38 cgd Exp $	*/
 
 /*
  * Copyright (c) 1982, 1986, 1993
@@ -35,18 +35,47 @@
  *	@(#)ip_var.h	8.1 (Berkeley) 6/10/93
  */
 
+#include <sys/queue.h>
+
 /*
  * Overlay for ip header used by other protocols (tcp, udp).
  */
 struct ipovly {
-	caddr_t  ih_next, ih_prev;	/* for protocol sequence q's */
-	u_int8_t ih_x1;			/* (unused) */
+	u_int8_t ih_x1[9];		/* (unused) */
 	u_int8_t ih_pr;			/* protocol */
 	int16_t	 ih_len;		/* protocol length */
 	struct	 in_addr ih_src;	/* source internet address */
 	struct	 in_addr ih_dst;	/* destination internet address */
 };
 
+/*
+ * Ip (reassembly or sequence) queue structures.
+ *
+ * XXX -- The following explains why the ipqe_m field is here, for TCP's use:
+ * We want to avoid doing m_pullup on incoming packets but that
+ * means avoiding dtom on the tcp reassembly code.  That in turn means
+ * keeping an mbuf pointer in the reassembly queue (since we might
+ * have a cluster).  As a quick hack, the source & destination
+ * port numbers (which are no longer needed once we've located the
+ * tcpcb) are overlayed with an mbuf pointer.
+ */
+LIST_HEAD(ipqehead, ipqent);
+struct ipqent {
+	LIST_ENTRY(ipqent) ipqe_q;
+	union {
+		struct ip	*_ip;
+		struct tcpiphdr *_tcp;
+	} _ipqe_u1;
+	union {
+		u_int8_t	_mff;	/* for IP fragmentation */
+		struct mbuf	*_m;	/* XXX for TCP; see above */
+	} _ipqe_u2;
+};
+#define	ipqe_ip		_ipqe_u1._ip
+#define	ipqe_tcp	_ipqe_u1._tcp
+#define	ipqe_mff	_ipqe_u2._mff
+#define	ipqe_m		_ipqe_u2._m
+
 /*
  * Ip reassembly queue structure.  Each fragment
  * being reassembled is attached to one of these structures.
@@ -54,42 +83,14 @@ struct ipovly {
  * be reclaimed if memory becomes tight.
  */
 struct ipq {
-	struct	  ipq *next, *prev;	/* to other reass headers */
+	LIST_ENTRY(ipq) ipq_q;		/* to other reass headers */
 	u_int8_t  ipq_ttl;		/* time for reass q to live */
 	u_int8_t  ipq_p;		/* protocol of this fragment */
 	u_int16_t ipq_id;		/* sequence id for reassembly */
-	struct	  ipasfrag *ipq_next, *ipq_prev;
-					/* to ip headers of fragments */
+	struct	  ipqehead ipq_fragq;	/* to ip fragment queue */
 	struct	  in_addr ipq_src, ipq_dst;
 };
 
-/*
- * Ip header, when holding a fragment.
- *
- * Note: ipf_next must be at same offset as ipq_next above
- */
-struct	ipasfrag {
-#if BYTE_ORDER == LITTLE_ENDIAN
-	u_int8_t  ip_hl:4,
-		  ip_v:4;
-#endif
-#if BYTE_ORDER == BIG_ENDIAN
-	u_int8_t  ip_v:4,
-		  ip_hl:4;
-#endif
-	u_int8_t  ipf_mff;		/* XXX overlays ip_tos: use low bit
-					 * to avoid destroying tos;
-					 * copied from (ip_off&IP_MF) */
-	int16_t	  ip_len;
-	u_int16_t ip_id;
-	int16_t	  ip_off;
-	u_int8_t  ip_ttl;
-	u_int8_t  ip_p;
-	u_int16_t ip_sum;
-	struct	  ipasfrag *ipf_next, *ipf_prev;
-					/* list of fragments */
-};
-
 /*
  * Structure stored in mbuf in inpcb.ip_options
  * and passed to ip_output when ip options are in use.
@@ -141,6 +142,7 @@ struct	ipstat {
 	u_long	ips_badvers;		/* ip version != 4 */
 	u_long	ips_rawout;		/* total raw ip packets generated */
 	u_long	ips_badfrags;		/* malformed fragments (bad length) */
+	u_long	ips_rcvmemdrop;		/* frags dropped for lack of memory */
 };
 
 #ifdef _KERNEL
@@ -151,15 +153,13 @@ struct	ipstat {
 #define	IP_ALLOWBROADCAST	SO_BROADCAST	/* can send broadcast packets */
 
 struct	  ipstat ipstat;
-struct	  ipq	 ipq;			/* ip reass. queue */
-u_int16_t ip_id;				/* ip packet ctr, for ids */
+LIST_HEAD(ipqhead, ipq)	ipq;		/* ip reass. queue */
+u_int16_t ip_id;			/* ip packet ctr, for ids */
 int	  ip_defttl;			/* default IP ttl */
 
 int	 ip_ctloutput __P((int, struct socket *, int, int, struct mbuf **));
-void	 ip_deq __P((struct ipasfrag *));
 int	 ip_dooptions __P((struct mbuf *));
 void	 ip_drain __P((void));
-void	 ip_enq __P((struct ipasfrag *, struct ipasfrag *));
 void	 ip_forward __P((struct mbuf *, int));
 void	 ip_freef __P((struct ipq *));
 void	 ip_freemoptions __P((struct ip_moptions *));
@@ -171,7 +171,7 @@ int	 ip_output __P((struct mbuf *,
 	    struct mbuf *, struct route *, int, struct ip_moptions *));
 int	 ip_pcbopts __P((struct mbuf **, struct mbuf *));
 struct ip *
-	 ip_reass __P((struct ipasfrag *, struct ipq *));
+	 ip_reass __P((struct ipqent *, struct ipq *));
 struct in_ifaddr *
 	 ip_rtaddr __P((struct in_addr));
 int	 ip_setmoptions __P((int, struct ip_moptions **, struct mbuf *));
diff --git a/sys/netinet/raw_ip.c b/sys/netinet/raw_ip.c
index 88fc13da5fb..785e4b309cd 100644
--- a/sys/netinet/raw_ip.c
+++ b/sys/netinet/raw_ip.c
@@ -1,4 +1,4 @@
-/*	$NetBSD: raw_ip.c,v 1.21 1995/06/18 20:01:15 cgd Exp $	*/
+/*	$NetBSD: raw_ip.c,v 1.22 1995/11/30 16:42:18 pk Exp $	*/
 
 /*
  * Copyright (c) 1982, 1986, 1988, 1993
@@ -255,6 +255,15 @@ rip_usrreq(so, req, m, nam, control)
 #ifdef MROUTING
 	extern struct socket *ip_mrouter;
 #endif
+	if (req == PRU_CONTROL)
+		return (in_control(so, (long)m, (caddr_t)nam,
+			(struct ifnet *)control));
+
+	if (inp == NULL && req != PRU_ATTACH) {
+		error = EINVAL;
+		goto release;
+	}
+
 	switch (req) {
 
 	case PRU_ATTACH:
@@ -396,6 +405,7 @@ rip_usrreq(so, req, m, nam, control)
 	default:
 		panic("rip_usrreq");
 	}
+release:
 	if (m != NULL)
 		m_freem(m);
 	return (error);
diff --git a/sys/netinet/tcp_input.c b/sys/netinet/tcp_input.c
index 8354c364742..b49c0e9defe 100644
--- a/sys/netinet/tcp_input.c
+++ b/sys/netinet/tcp_input.c
@@ -1,4 +1,4 @@
-/*	$NetBSD: tcp_input.c,v 1.19 1995/08/04 01:12:23 mycroft Exp $	*/
+/*	$NetBSD: tcp_input.c,v 1.20 1995/11/21 01:07:39 cgd Exp $	*/
 
 /*
  * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1994
@@ -87,7 +87,7 @@ extern u_long sb_max;
  */
 #define	TCP_REASS(tp, ti, m, so, flags) { \
 	if ((ti)->ti_seq == (tp)->rcv_nxt && \
-	    (tp)->seg_next == (struct tcpiphdr *)(tp) && \
+	    (tp)->segq.lh_first == NULL && \
 	    (tp)->t_state == TCPS_ESTABLISHED) { \
 		if ((ti)->ti_flags & TH_PUSH) \
 			tp->t_flags |= TF_ACKNOW; \
@@ -112,7 +112,7 @@ tcp_reass(tp, ti, m)
 	register struct tcpiphdr *ti;
 	struct mbuf *m;
 {
-	register struct tcpiphdr *q;
+	register struct ipqent *p, *q, *nq, *tiqe;
 	struct socket *so = tp->t_inpcb->inp_socket;
 	int flags;
 
@@ -123,12 +123,23 @@ tcp_reass(tp, ti, m)
 	if (ti == 0)
 		goto present;
 
+	/*
+	 * Allocate a new queue entry, before we throw away any data.
+	 * If we can't, just drop the packet.  XXX
+	 */
+	MALLOC(tiqe, struct ipqent *, sizeof (struct ipqent), M_IPQ, M_NOWAIT);
+	if (tiqe == NULL) {
+		tcpstat.tcps_rcvmemdrop++;
+		m_freem(m);
+		return (0);
+	}
+
 	/*
 	 * Find a segment which begins after this one does.
 	 */
-	for (q = tp->seg_next; q != (struct tcpiphdr *)tp;
-	    q = (struct tcpiphdr *)q->ti_next)
-		if (SEQ_GT(q->ti_seq, ti->ti_seq))
+	for (p = NULL, q = tp->segq.lh_first; q != NULL;
+	    p = q, q = q->ipqe_q.le_next)
+		if (SEQ_GT(q->ipqe_tcp->ti_seq, ti->ti_seq))
 			break;
 
 	/*
@@ -136,52 +147,58 @@ tcp_reass(tp, ti, m)
 	 * our data already.  If so, drop the data from the incoming
 	 * segment.  If it provides all of our data, drop us.
 	 */
-	if ((struct tcpiphdr *)q->ti_prev != (struct tcpiphdr *)tp) {
+	if (p != NULL) {
+		register struct tcpiphdr *phdr = p->ipqe_tcp;
 		register int i;
-		q = (struct tcpiphdr *)q->ti_prev;
+
 		/* conversion to int (in i) handles seq wraparound */
-		i = q->ti_seq + q->ti_len - ti->ti_seq;
+		i = phdr->ti_seq + phdr->ti_len - ti->ti_seq;
 		if (i > 0) {
 			if (i >= ti->ti_len) {
 				tcpstat.tcps_rcvduppack++;
 				tcpstat.tcps_rcvdupbyte += ti->ti_len;
 				m_freem(m);
+				FREE(tiqe, M_IPQ);
 				return (0);
 			}
 			m_adj(m, i);
 			ti->ti_len -= i;
 			ti->ti_seq += i;
 		}
-		q = (struct tcpiphdr *)(q->ti_next);
 	}
 	tcpstat.tcps_rcvoopack++;
 	tcpstat.tcps_rcvoobyte += ti->ti_len;
-	REASS_MBUF(ti) = m;		/* XXX */
 
 	/*
 	 * While we overlap succeeding segments trim them or,
 	 * if they are completely covered, dequeue them.
 	 */
-	while (q != (struct tcpiphdr *)tp) {
-		register int i = (ti->ti_seq + ti->ti_len) - q->ti_seq;
+	for (; q != NULL; q = nq) {
+		register struct tcpiphdr *qhdr = q->ipqe_tcp;
+		register int i = (ti->ti_seq + ti->ti_len) - qhdr->ti_seq;
+
 		if (i <= 0)
 			break;
-		if (i < q->ti_len) {
-			q->ti_seq += i;
-			q->ti_len -= i;
-			m_adj(REASS_MBUF(q), i);
+		if (i < qhdr->ti_len) {
+			qhdr->ti_seq += i;
+			qhdr->ti_len -= i;
+			m_adj(q->ipqe_m, i);
 			break;
 		}
-		q = (struct tcpiphdr *)q->ti_next;
-		m = REASS_MBUF((struct tcpiphdr *)q->ti_prev);
-		remque(q->ti_prev);
-		m_freem(m);
+		nq = q->ipqe_q.le_next;
+		m_freem(q->ipqe_m);
+		LIST_REMOVE(q, ipqe_q);
+		FREE(q, M_IPQ);
 	}
 
-	/*
-	 * Stick new segment in its place.
-	 */
-	insque(ti, q->ti_prev);
+	/* Insert the new fragment queue entry into place. */
+	tiqe->ipqe_m = m;
+	tiqe->ipqe_tcp = ti;
+	if (p == NULL) {
+		LIST_INSERT_HEAD(&tp->segq, tiqe, ipqe_q);
+	} else {
+		LIST_INSERT_AFTER(p, tiqe, ipqe_q);
+	}
 
 present:
 	/*
@@ -190,22 +207,24 @@ present:
 	 */
 	if (TCPS_HAVEESTABLISHED(tp->t_state) == 0)
 		return (0);
-	ti = tp->seg_next;
-	if (ti == (struct tcpiphdr *)tp || ti->ti_seq != tp->rcv_nxt)
+	q = tp->segq.lh_first;
+	if (q == NULL || q->ipqe_tcp->ti_seq != tp->rcv_nxt)
 		return (0);
-	if (tp->t_state == TCPS_SYN_RECEIVED && ti->ti_len)
+	if (tp->t_state == TCPS_SYN_RECEIVED && q->ipqe_tcp->ti_len)
 		return (0);
 	do {
-		tp->rcv_nxt += ti->ti_len;
-		flags = ti->ti_flags & TH_FIN;
-		remque(ti);
-		m = REASS_MBUF(ti);
-		ti = (struct tcpiphdr *)ti->ti_next;
+		tp->rcv_nxt += q->ipqe_tcp->ti_len;
+		flags = q->ipqe_tcp->ti_flags & TH_FIN;
+
+		nq = q->ipqe_q.le_next;
+		LIST_REMOVE(q, ipqe_q);
 		if (so->so_state & SS_CANTRCVMORE)
-			m_freem(m);
+			m_freem(q->ipqe_m);
 		else
-			sbappend(&so->so_rcv, m);
-	} while (ti != (struct tcpiphdr *)tp && ti->ti_seq == tp->rcv_nxt);
+			sbappend(&so->so_rcv, q->ipqe_m);
+		FREE(q, M_IPQ);
+		q = nq;
+	} while (q != NULL && q->ipqe_tcp->ti_seq == tp->rcv_nxt);
 	sorwakeup(so);
 	return (flags);
 }
@@ -257,8 +276,7 @@ tcp_input(m, iphlen)
 	 */
 	tlen = ((struct ip *)ti)->ip_len;
 	len = sizeof (struct ip) + tlen;
-	ti->ti_next = ti->ti_prev = 0;
-	ti->ti_x1 = 0;
+	bzero(ti->ti_x1, sizeof ti->ti_x1);
 	ti->ti_len = (u_int16_t)tlen;
 	HTONS(ti->ti_len);
 	if (ti->ti_sum = in_cksum(m, len)) {
@@ -479,7 +497,7 @@ findpcb:
 				return;
 			}
 		} else if (ti->ti_ack == tp->snd_una &&
-		    tp->seg_next == (struct tcpiphdr *)tp &&
+		    tp->segq.lh_first == NULL &&
 		    ti->ti_len <= sbspace(&so->so_rcv)) {
 			/*
 			 * this is a pure, in-sequence data packet
diff --git a/sys/netinet/tcp_subr.c b/sys/netinet/tcp_subr.c
index 76f8c9d631b..c64a868767a 100644
--- a/sys/netinet/tcp_subr.c
+++ b/sys/netinet/tcp_subr.c
@@ -1,4 +1,4 @@
-/*	$NetBSD: tcp_subr.c,v 1.19 1995/06/12 06:48:54 mycroft Exp $	*/
+/*	$NetBSD: tcp_subr.c,v 1.20 1995/11/21 01:07:41 cgd Exp $	*/
 
 /*
  * Copyright (c) 1982, 1986, 1988, 1990, 1993
@@ -104,8 +104,7 @@ tcp_template(tp)
 		m->m_len = sizeof (struct tcpiphdr);
 		n = mtod(m, struct tcpiphdr *);
 	}
-	n->ti_next = n->ti_prev = 0;
-	n->ti_x1 = 0;
+	bzero(n->ti_x1, sizeof n->ti_x1);
 	n->ti_pr = IPPROTO_TCP;
 	n->ti_len = htons(sizeof (struct tcpiphdr) - sizeof (struct ip));
 	n->ti_src = inp->inp_laddr;
@@ -176,13 +175,12 @@ tcp_respond(tp, ti, m, ack, seq, flags)
 		xchg(ti->ti_dport, ti->ti_sport, u_int16_t);
 #undef xchg
 	}
-	ti->ti_len = htons((u_short)(sizeof (struct tcphdr) + tlen));
+	ti->ti_len = htons((u_int16_t)(sizeof (struct tcphdr) + tlen));
 	tlen += sizeof (struct tcpiphdr);
 	m->m_len = tlen;
 	m->m_pkthdr.len = tlen;
 	m->m_pkthdr.rcvif = (struct ifnet *) 0;
-	ti->ti_next = ti->ti_prev = 0;
-	ti->ti_x1 = 0;
+	bzero(ti->ti_x1, sizeof ti->ti_x1);
 	ti->ti_seq = htonl(seq);
 	ti->ti_ack = htonl(ack);
 	ti->ti_x2 = 0;
@@ -215,7 +213,7 @@ tcp_newtcpcb(inp)
 	if (tp == NULL)
 		return ((struct tcpcb *)0);
 	bzero((char *) tp, sizeof(struct tcpcb));
-	tp->seg_next = tp->seg_prev = (struct tcpiphdr *)tp;
+	LIST_INIT(&tp->segq);
 	tp->t_maxseg = tcp_mssdflt;
 
 	tp->t_flags = tcp_do_rfc1323 ? (TF_REQ_SCALE|TF_REQ_TSTMP) : 0;
@@ -271,7 +269,7 @@ struct tcpcb *
 tcp_close(tp)
 	register struct tcpcb *tp;
 {
-	register struct tcpiphdr *t;
+	register struct ipqent *qe;
 	struct inpcb *inp = tp->t_inpcb;
 	struct socket *so = inp->inp_socket;
 	register struct mbuf *m;
@@ -346,12 +344,10 @@ tcp_close(tp)
 	}
 #endif /* RTV_RTT */
 	/* free the reassembly queue, if any */
-	t = tp->seg_next;
-	while (t != (struct tcpiphdr *)tp) {
-		t = (struct tcpiphdr *)t->ti_next;
-		m = REASS_MBUF((struct tcpiphdr *)t->ti_prev);
-		remque(t->ti_prev);
-		m_freem(m);
+	while ((qe = tp->segq.lh_first) != NULL) {
+		LIST_REMOVE(qe, ipqe_q);
+		m_freem(qe->ipqe_m);
+		FREE(qe, M_IPQ);
 	}
 	if (tp->t_template)
 		(void) m_free(dtom(tp->t_template));
diff --git a/sys/netinet/tcp_var.h b/sys/netinet/tcp_var.h
index c97afca58a9..99ab071627c 100644
--- a/sys/netinet/tcp_var.h
+++ b/sys/netinet/tcp_var.h
@@ -1,4 +1,4 @@
-/*	$NetBSD: tcp_var.h,v 1.14 1995/09/30 07:02:08 thorpej Exp $	*/
+/*	$NetBSD: tcp_var.h,v 1.15 1995/11/21 01:07:43 cgd Exp $	*/
 
 /*
  * Copyright (c) 1982, 1986, 1993, 1994
@@ -43,8 +43,7 @@
  * Tcp control block, one per tcp; fields:
  */
 struct tcpcb {
-	struct	tcpiphdr *seg_next, *seg_prev;
-					/* list of control blocks */
+	struct ipqehead segq;		/* sequencing queue */
 	short	t_state;		/* state of this connection */
 	short	t_timer[TCPT_NTIMERS];	/* tcp timers */
 	short	t_rxtshift;		/* log(2) of rexmt exp. backoff */
@@ -163,16 +162,6 @@ struct tcpcb {
 #define	TCP_REXMTVAL(tp) \
 	((((tp)->t_srtt >> TCP_RTT_SHIFT) + (tp)->t_rttvar) >> 2)
 
-/* XXX
- * We want to avoid doing m_pullup on incoming packets but that
- * means avoiding dtom on the tcp reassembly code.  That in turn means
- * keeping an mbuf pointer in the reassembly queue (since we might
- * have a cluster).  As a quick hack, the source & destination
- * port numbers (which are no longer needed once we've located the
- * tcpcb) are overlayed with an mbuf pointer.
- */
-#define REASS_MBUF(ti) (*(struct mbuf **)&((ti)->ti_t))
-
 /*
  * TCP statistics.
  * Many of these should be kept per connection,
@@ -211,6 +200,7 @@ struct	tcpstat {
 	u_long	tcps_rcvbyte;		/* bytes received in sequence */
 	u_long	tcps_rcvbadsum;		/* packets received with ccksum errs */
 	u_long	tcps_rcvbadoff;		/* packets received with bad offset */
+	u_long	tcps_rcvmemdrop;	/* packets dropped for lack of memory */
 	u_long	tcps_rcvshort;		/* packets received too short */
 	u_long	tcps_rcvduppack;	/* duplicate-only packets received */
 	u_long	tcps_rcvdupbyte;	/* duplicate-only bytes received */
diff --git a/sys/netinet/tcpip.h b/sys/netinet/tcpip.h
index 9fbfb2614ee..ba09485816c 100644
--- a/sys/netinet/tcpip.h
+++ b/sys/netinet/tcpip.h
@@ -1,4 +1,4 @@
-/*	$NetBSD: tcpip.h,v 1.5 1994/06/29 06:38:55 cgd Exp $	*/
+/*	$NetBSD: tcpip.h,v 1.6 1995/11/21 01:07:44 cgd Exp $	*/
 
 /*
  * Copyright (c) 1982, 1986, 1993
@@ -42,8 +42,6 @@ struct tcpiphdr {
 	struct 	ipovly ti_i;		/* overlaid ip structure */
 	struct	tcphdr ti_t;		/* tcp header */
 };
-#define	ti_next		ti_i.ih_next
-#define	ti_prev		ti_i.ih_prev
 #define	ti_x1		ti_i.ih_x1
 #define	ti_pr		ti_i.ih_pr
 #define	ti_len		ti_i.ih_len
diff --git a/sys/netinet/udp_usrreq.c b/sys/netinet/udp_usrreq.c
index 3742a68973d..b7647352803 100644
--- a/sys/netinet/udp_usrreq.c
+++ b/sys/netinet/udp_usrreq.c
@@ -1,4 +1,4 @@
-/*	$NetBSD: udp_usrreq.c,v 1.24 1995/08/12 23:59:42 mycroft Exp $	*/
+/*	$NetBSD: udp_usrreq.c,v 1.25 1995/11/21 01:07:46 cgd Exp $	*/
 
 /*
  * Copyright (c) 1982, 1986, 1988, 1990, 1993
@@ -142,9 +142,8 @@ udp_input(m, iphlen)
 	 * Checksum extended UDP header and data.
 	 */
 	if (udpcksum && uh->uh_sum) {
-		((struct ipovly *)ip)->ih_next = 0;
-		((struct ipovly *)ip)->ih_prev = 0;
-		((struct ipovly *)ip)->ih_x1 = 0;
+		bzero(((struct ipovly *)ip)->ih_x1,
+		    sizeof ((struct ipovly *)ip)->ih_x1);
 		((struct ipovly *)ip)->ih_len = uh->uh_ulen;
 		if (uh->uh_sum = in_cksum(m, len + sizeof (struct ip))) {
 			udpstat.udps_badsum++;
@@ -437,8 +436,7 @@ udp_output(inp, m, addr, control)
 	 * and addresses and length put into network format.
 	 */
 	ui = mtod(m, struct udpiphdr *);
-	ui->ui_next = ui->ui_prev = 0;
-	ui->ui_x1 = 0;
+	bzero(ui->ui_x1, sizeof ui->ui_x1);
 	ui->ui_pr = IPPROTO_UDP;
 	ui->ui_len = htons((u_int16_t)len + sizeof (struct udphdr));
 	ui->ui_src = inp->inp_laddr;
diff --git a/sys/netinet/udp_var.h b/sys/netinet/udp_var.h
index 6dadf4f1425..c58dbce9616 100644
--- a/sys/netinet/udp_var.h
+++ b/sys/netinet/udp_var.h
@@ -1,4 +1,4 @@
-/*	$NetBSD: udp_var.h,v 1.9 1995/06/12 00:48:09 mycroft Exp $	*/
+/*	$NetBSD: udp_var.h,v 1.10 1995/11/21 01:07:48 cgd Exp $	*/
 
 /*
  * Copyright (c) 1982, 1986, 1989, 1993
@@ -42,8 +42,6 @@ struct	udpiphdr {
 	struct 	ipovly ui_i;		/* overlaid ip structure */
 	struct	udphdr ui_u;		/* udp header */
 };
-#define	ui_next		ui_i.ih_next
-#define	ui_prev		ui_i.ih_prev
 #define	ui_x1		ui_i.ih_x1
 #define	ui_pr		ui_i.ih_pr
 #define	ui_len		ui_i.ih_len
diff --git a/sys/netiso/tuba_subr.c b/sys/netiso/tuba_subr.c
index 465af6293db..927fea34a12 100644
--- a/sys/netiso/tuba_subr.c
+++ b/sys/netiso/tuba_subr.c
@@ -1,4 +1,4 @@
-/*	$NetBSD: tuba_subr.c,v 1.4 1995/06/13 07:58:24 mycroft Exp $	*/
+/*	$NetBSD: tuba_subr.c,v 1.5 1995/11/21 01:07:54 cgd Exp $	*/
 
 /*
  * Copyright (c) 1992, 1993
@@ -326,8 +326,8 @@ tuba_tcpinput(m, src, dst)
 		ti->ti_src.s_addr = tuba_table[findex]->tc_sum;
 	else
 		ti->ti_src.s_addr = tuba_table[findex]->tc_ssum;
-	ti->ti_prev = ti->ti_next = 0;
-	ti->ti_x1 = 0; ti->ti_pr = ISOPROTO_TCP;
+	bzero(ti->ti_x1, sizeof ti->ti_x1);
+	ti->ti_pr = ISOPROTO_TCP;
 	ti->ti_len = htons((u_short)tlen);
 	if (ti->ti_sum = in_cksum(m, m->m_pkthdr.len)) {
 		tcpstat.tcps_rcvbadsum++;
diff --git a/sys/netiso/tuba_table.c b/sys/netiso/tuba_table.c
index 1f87ef60ad4..e9663a13539 100644
--- a/sys/netiso/tuba_table.c
+++ b/sys/netiso/tuba_table.c
@@ -1,4 +1,4 @@
-/*	$NetBSD: tuba_table.c,v 1.4 1995/08/12 23:59:53 mycroft Exp $	*/
+/*	$NetBSD: tuba_table.c,v 1.5 1995/10/16 05:28:58 mycroft Exp $	*/
 
 /*
  * Copyright (c) 1992, 1993
@@ -91,7 +91,7 @@ tuba_lookup(siso, wait)
 	struct tuba_cache **new;
 	int dupentry = 0, sum_a = 0, sum_b = 0, old_size, i;
 
-	if ((rn = rn_match((caddr_t)&siso->siso_addr, tuba_tree->rnh_treetop))
+	if ((rn = rn_match((caddr_t)&siso->siso_addr, tuba_tree))
 	     && ((rn->rn_flags & RNF_ROOT) == 0)) {
 		tc = (struct tuba_cache *)rn;
 		tc->tc_time = time.tv_sec;