Back out the ip_fragment() code -- it is not urgent to have it in now,

I will put it back in in a better form after 5.0 is out.

Requested by: sam, rwatson, luigi (on second thought)
Approved by: re

2 files changed, 143 insertions, 170 deletions

diff --git a/sys/netinet/ip_output.c b/sys/netinet/ip_output.c
index 1ed5459..4c19090 100644
--- a/sys/netinet/ip_output.c
+++ b/sys/netinet/ip_output.c
@@ -116,10 +116,15 @@ extern	struct protosw inetsw[];
  * The mbuf opt, if present, will not be freed.
  */
 int
-ip_output(struct mbuf *m0, struct mbuf *opt, struct route *ro,
-	int flags, struct ip_moptions *imo, struct inpcb *inp)
+ip_output(m0, opt, ro, flags, imo, inp)
+	struct mbuf *m0;
+	struct mbuf *opt;
+	struct route *ro;
+	int flags;
+	struct ip_moptions *imo;
+	struct inpcb *inp;
 {
-	struct ip *ip;
+	struct ip *ip, *mhip;
 	struct ifnet *ifp = NULL;	/* keep compiler happy */
 	struct mbuf *m;
 	int hlen = sizeof (struct ip);
@@ -453,7 +458,7 @@ ip_output(struct mbuf *m0, struct mbuf *opt, struct route *ro,
 			goto bad;
 		}
 		/* don't allow broadcast messages to be fragmented */
-		if (ip->ip_len > ifp->if_mtu) {
+		if ((u_short)ip->ip_len > ifp->if_mtu) {
 			error = EMSGSIZE;
 			goto bad;
 		}
@@ -980,7 +985,8 @@ pass:
 	 * If small enough for interface, or the interface will take
 	 * care of the fragmentation for us, can just send directly.
 	 */
-	if (ip->ip_len <= ifp->if_mtu || ifp->if_hwassist & CSUM_FRAGMENT) {
+	if ((u_short)ip->ip_len <= ifp->if_mtu ||
+	    ifp->if_hwassist & CSUM_FRAGMENT) {
 		ip->ip_len = htons(ip->ip_len);
 		ip->ip_off = htons(ip->ip_off);
 		ip->ip_sum = 0;
@@ -1002,6 +1008,10 @@ pass:
 				(struct sockaddr *)dst, ro->ro_rt);
 		goto done;
 	}
+	/*
+	 * Too large for interface; fragment if possible.
+	 * Must be able to put at least 8 bytes per fragment.
+	 */
 	if (ip->ip_off & IP_DF) {
 		error = EMSGSIZE;
 		/*
@@ -1019,122 +1029,38 @@ pass:
 		ipstat.ips_cantfrag++;
 		goto bad;
 	}
-	/*
-	 * Too large for interface; fragment if possible. If successful,
-	 * on return m will point to a list of packets to be sent.
-	 */
-	error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist, sw_csum);
-	if (error)
+	len = (ifp->if_mtu - hlen) &~ 7;
+	if (len < 8) {
+		error = EMSGSIZE;
 		goto bad;
-	for (; m; m = m0) {
-		m0 = m->m_nextpkt;
-		m->m_nextpkt = NULL;
-#ifdef IPSEC
-		/* clean ipsec history once it goes out of the node */
-		ipsec_delaux(m);
-#endif
-		if (error == 0) {
-			/* Record statistics for this interface address. */
-			if (ia != NULL) {
-				ia->ia_ifa.if_opackets++;
-				ia->ia_ifa.if_obytes += m->m_pkthdr.len;
-			}
-			
-			error = (*ifp->if_output)(ifp, m,
-			    (struct sockaddr *)dst, ro->ro_rt);
-		} else
-			m_freem(m);
-	}
-
-	if (error == 0)
-		ipstat.ips_fragmented++;
-
-done:
-#ifdef IPSEC
-	if (ro == &iproute && ro->ro_rt) {
-		RTFREE(ro->ro_rt);
-		ro->ro_rt = NULL;
 	}
-	if (sp != NULL) {
-		KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
-			printf("DP ip_output call free SP:%p\n", sp));
-		key_freesp(sp);
-	}
-#endif /* IPSEC */
-#ifdef FAST_IPSEC
-	if (ro == &iproute && ro->ro_rt) {
-		RTFREE(ro->ro_rt);
-		ro->ro_rt = NULL;
-	}
-	if (sp != NULL)
-		KEY_FREESP(&sp);
-#endif /* FAST_IPSEC */
-	return (error);
-bad:
-	m_freem(m);
-	goto done;
-}
-
-/*
- * Create a chain of fragments which fit the given mtu. m_frag points to the
- * mbuf to be fragmented; on return it points to the chain with the fragments.
- * Return 0 if no error.
- *
- * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
- * sw_csum contains the delayed checksums flags (e.g., CSUM_DELAY_IP).
- */
-int
-ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
-	u_long if_hwassist_flags, int sw_csum)
-{
-	int error = 0;
-	int hlen = ip->ip_hl << 2;
-	int len = (mtu - hlen) &~ 7;	/* size of payload in each fragment */
-	int off;
-	struct mbuf *m0 = *m_frag;	/* the original packet		*/
-	int firstlen;
-	struct mbuf **mnext;
-	int nfrags;
-
-	if (ip->ip_off & IP_DF) {	/* Fragmentation not allowed */
-		ipstat.ips_cantfrag++;
-		return EMSGSIZE;
-	}
-
-	/*
-	 * Must be able to put at least 8 bytes per fragment.
-	 */
-	if (len < 8)
-		return EMSGSIZE;
 
 	/*
 	 * if the interface will not calculate checksums on
 	 * fragmented packets, then do it here.
 	 */
-	if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA &&
-	    (if_hwassist_flags & CSUM_IP_FRAGS) == 0) {
-		in_delayed_cksum(m0);
-		m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
+	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA &&
+	    (ifp->if_hwassist & CSUM_IP_FRAGS) == 0) {
+		in_delayed_cksum(m);
+		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
 	}
 
 	if (len > PAGE_SIZE) {
 		/* 
-		 * Fragment large datagrams such that each segment 
+		 * Fragement large datagrams such that each segment 
 		 * contains a multiple of PAGE_SIZE amount of data, 
 		 * plus headers. This enables a receiver to perform 
 		 * page-flipping zero-copy optimizations.
-		 *
-		 * XXX When does this help given that sender and receiver
-		 * could have different page sizes, and also mtu could
-		 * be less than the receiver's page size ?
 		 */
 
 		int newlen;
-		struct mbuf *m;
+		struct mbuf *mtmp;
 
-		for (m = m0, off = 0; m && ((off + m->m_len) <= mtu);
-		     m = m->m_next)
-			off += m->m_len;
+		for (mtmp = m, off = 0; 
+		     mtmp && ((off + mtmp->m_len) <= ifp->if_mtu);
+		     mtmp = mtmp->m_next) {
+			off += mtmp->m_len;
+		}
 		/*
 		 * firstlen (off - hlen) must be aligned on an 
 		 * 8-byte boundary
@@ -1142,46 +1068,44 @@ ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
 		if (off < hlen)
 			goto smart_frag_failure;
 		off = ((off - hlen) & ~7) + hlen;
-		newlen = (~PAGE_MASK) & mtu;
-		if ((newlen + sizeof (struct ip)) > mtu) {
+		newlen = (~PAGE_MASK) & ifp->if_mtu;
+		if ((newlen + sizeof (struct ip)) > ifp->if_mtu) {
 			/* we failed, go back the default */
 smart_frag_failure:
 			newlen = len;
 			off = hlen + len;
 		}
+
+/*		printf("ipfrag: len = %d, hlen = %d, mhlen = %d, newlen = %d, off = %d\n",
+		len, hlen, sizeof (struct ip), newlen, off);*/
+
 		len = newlen;
 
 	} else {
 		off = hlen + len;
 	}
 
-	firstlen = off - hlen;
-	mnext = &m0->m_nextpkt;		/* pointer to next packet */
+
+
+    {
+	int mhlen, firstlen = off - hlen;
+	struct mbuf **mnext = &m->m_nextpkt;
+	int nfrags = 1;
 
 	/*
 	 * Loop through length of segment after first fragment,
 	 * make new header and copy data of each part and link onto chain.
-	 * Here, m0 is the original packet, m is the fragment being created.
-	 * The fragments are linked off the m_nextpkt of the original
-	 * packet, which after processing serves as the first fragment.
 	 */
-	for (nfrags=1; off < ip->ip_len; off += len, nfrags++) {
-		struct ip *mhip;	/* ip header on the fragment */
-		struct mbuf *m;
-		int mhlen = sizeof (struct ip);
-
+	m0 = m;
+	mhlen = sizeof (struct ip);
+	for (; off < (u_short)ip->ip_len; off += len) {
 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
 		if (m == 0) {
 			error = ENOBUFS;
 			ipstat.ips_odropped++;
-			goto done;
+			goto sendorfree;
 		}
 		m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG;
-		/*
-		 * In the first mbuf, leave room for the link header, then
-		 * copy the original IP header including options. The payload
-		 * goes into an additional mbuf chain returned by m_copy().
-		 */
 		m->m_data += max_linkhdr;
 		mhip = mtod(m, struct ip *);
 		*mhip = *ip;
@@ -1191,20 +1115,18 @@ smart_frag_failure:
 			mhip->ip_hl = mhlen >> 2;
 		}
 		m->m_len = mhlen;
-		/* XXX do we need to add ip->ip_off below ? */
 		mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off;
-		if (off + len >= ip->ip_len) { /* last fragment */
-			len = ip->ip_len - off;
-			m->m_flags |= M_LASTFRAG;
-		} else
+		if (off + len >= (u_short)ip->ip_len)
+			len = (u_short)ip->ip_len - off;
+		else
 			mhip->ip_off |= IP_MF;
 		mhip->ip_len = htons((u_short)(len + mhlen));
 		m->m_next = m_copy(m0, off, len);
-		if (m->m_next == 0) {		/* copy failed */
-			m_free(m);
+		if (m->m_next == 0) {
+			(void) m_free(m);
 			error = ENOBUFS;	/* ??? */
 			ipstat.ips_odropped++;
-			goto done;
+			goto sendorfree;
 		}
 		m->m_pkthdr.len = mhlen + len;
 		m->m_pkthdr.rcvif = (struct ifnet *)0;
@@ -1218,42 +1140,76 @@ smart_frag_failure:
 			mhip->ip_sum = in_cksum(m, mhlen);
 		*mnext = m;
 		mnext = &m->m_nextpkt;
+		nfrags++;
 	}
 	ipstat.ips_ofragments += nfrags;
 
-	/* set first markers for fragment chain */
+	/* set first/last markers for fragment chain */
+	m->m_flags |= M_LASTFRAG;
 	m0->m_flags |= M_FIRSTFRAG | M_FRAG;
 	m0->m_pkthdr.csum_data = nfrags;
 
 	/*
-	 * Update first fragment by trimming what has been copied out
-	 * and updating header.
+	 * Update first fragment by trimming what's been copied out
+	 * and updating header, then send each fragment (in order).
 	 */
-	m_adj(m0, hlen + firstlen - ip->ip_len);
-	m0->m_pkthdr.len = hlen + firstlen;
-	ip->ip_len = htons((u_short)m0->m_pkthdr.len);
+	m = m0;
+	m_adj(m, hlen + firstlen - (u_short)ip->ip_len);
+	m->m_pkthdr.len = hlen + firstlen;
+	ip->ip_len = htons((u_short)m->m_pkthdr.len);
 	ip->ip_off |= IP_MF;
 	ip->ip_off = htons(ip->ip_off);
 	ip->ip_sum = 0;
 	if (sw_csum & CSUM_DELAY_IP)
-		ip->ip_sum = in_cksum(m0, hlen);
-	*m_frag = m0;
-
-done:
-	if (error) {
-		struct mbuf *m;
-
-		for (m = m0; m; m = m0) {
-			m0 = m->m_nextpkt;
-			m->m_nextpkt = 0;
+		ip->ip_sum = in_cksum(m, hlen);
+sendorfree:
+	for (m = m0; m; m = m0) {
+		m0 = m->m_nextpkt;
+		m->m_nextpkt = 0;
 #ifdef IPSEC
-			/* clean ipsec history */
-			ipsec_delaux(m);
+		/* clean ipsec history once it goes out of the node */
+		ipsec_delaux(m);
 #endif
+		if (error == 0) {
+			/* Record statistics for this interface address. */
+			if (ia != NULL) {
+				ia->ia_ifa.if_opackets++;
+				ia->ia_ifa.if_obytes += m->m_pkthdr.len;
+			}
+			
+			error = (*ifp->if_output)(ifp, m,
+			    (struct sockaddr *)dst, ro->ro_rt);
+		} else
 			m_freem(m);
-		}
 	}
-	return error;
+
+	if (error == 0)
+		ipstat.ips_fragmented++;
+    }
+done:
+#ifdef IPSEC
+	if (ro == &iproute && ro->ro_rt) {
+		RTFREE(ro->ro_rt);
+		ro->ro_rt = NULL;
+	}
+	if (sp != NULL) {
+		KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
+			printf("DP ip_output call free SP:%p\n", sp));
+		key_freesp(sp);
+	}
+#endif /* IPSEC */
+#ifdef FAST_IPSEC
+	if (ro == &iproute && ro->ro_rt) {
+		RTFREE(ro->ro_rt);
+		ro->ro_rt = NULL;
+	}
+	if (sp != NULL)
+		KEY_FREESP(&sp);
+#endif /* FAST_IPSEC */
+	return (error);
+bad:
+	m_freem(m);
+	goto done;
 }
 
 void
@@ -1291,15 +1247,18 @@ in_delayed_cksum(struct mbuf *m)
  * XXX This routine assumes that the packet has no options in place.
  */
 static struct mbuf *
-ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen)
+ip_insertoptions(m, opt, phlen)
+	register struct mbuf *m;
+	struct mbuf *opt;
+	int *phlen;
 {
-	struct ipoption *p = mtod(opt, struct ipoption *);
+	register struct ipoption *p = mtod(opt, struct ipoption *);
 	struct mbuf *n;
-	struct ip *ip = mtod(m, struct ip *);
+	register struct ip *ip = mtod(m, struct ip *);
 	unsigned optlen;
 
 	optlen = opt->m_len - sizeof(p->ipopt_dst);
-	if (optlen + ip->ip_len > IP_MAXPACKET) {
+	if (optlen + (u_short)ip->ip_len > IP_MAXPACKET) {
 		*phlen = 0;
 		return (m);		/* XXX should fail */
 	}
@@ -1343,9 +1302,10 @@ ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen)
  * omitting those not copied during fragmentation.
  */
 int
-ip_optcopy(struct ip *ip, struct ip *jp)
+ip_optcopy(ip, jp)
+	struct ip *ip, *jp;
 {
-	u_char *cp, *dp;
+	register u_char *cp, *dp;
 	int opt, optlen, cnt;
 
 	cp = (u_char *)(ip + 1);
@@ -1385,7 +1345,9 @@ ip_optcopy(struct ip *ip, struct ip *jp)
  * IP socket option processing.
  */
 int
-ip_ctloutput(struct socket *so, struct sockopt *sopt)
+ip_ctloutput(so, sopt)
+	struct socket *so;
+	struct sockopt *sopt;
 {
 	struct	inpcb *inp = sotoinpcb(so);
 	int	error, optval;
@@ -1646,10 +1608,13 @@ ip_ctloutput(struct socket *so, struct sockopt *sopt)
  * with destination address if source routed.
  */
 static int
-ip_pcbopts(int optname, struct mbuf **pcbopt, struct mbuf *m)
+ip_pcbopts(optname, pcbopt, m)
+	int optname;
+	struct mbuf **pcbopt;
+	register struct mbuf *m;
 {
-	int cnt, optlen;
-	u_char *cp;
+	register int cnt, optlen;
+	register u_char *cp;
 	u_char opt;
 
 	/* turn off any old options */
@@ -1752,7 +1717,9 @@ bad:
  * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index.
  */
 static struct ifnet *
-ip_multicast_if(struct in_addr *a, int *ifindexp)
+ip_multicast_if(a, ifindexp)
+	struct in_addr *a;
+	int *ifindexp;
 {
 	int ifindex;
 	struct ifnet *ifp;
@@ -1776,7 +1743,9 @@ ip_multicast_if(struct in_addr *a, int *ifindexp)
  * Set the IP multicast options in response to user setsockopt().
  */
 static int
-ip_setmoptions(struct sockopt *sopt, struct ip_moptions **imop)
+ip_setmoptions(sopt, imop)
+	struct sockopt *sopt;
+	struct ip_moptions **imop;
 {
 	int error = 0;
 	int i;
@@ -2072,7 +2041,9 @@ ip_setmoptions(struct sockopt *sopt, struct ip_moptions **imop)
  * Return the IP multicast options in response to user getsockopt().
  */
 static int
-ip_getmoptions(struct sockopt *sopt, struct ip_moptions *imo)
+ip_getmoptions(sopt, imo)
+	struct sockopt *sopt;
+	register struct ip_moptions *imo;
 {
 	struct in_addr addr;
 	struct in_ifaddr *ia;
@@ -2136,9 +2107,10 @@ ip_getmoptions(struct sockopt *sopt, struct ip_moptions *imo)
  * Discard the IP multicast options.
  */
 void
-ip_freemoptions(struct ip_moptions *imo)
+ip_freemoptions(imo)
+	register struct ip_moptions *imo;
 {
-	int i;
+	register int i;
 
 	if (imo != NULL) {
 		for (i = 0; i < imo->imo_num_memberships; ++i)
@@ -2155,13 +2127,16 @@ ip_freemoptions(struct ip_moptions *imo)
  * replicating that code here.
  */
 static void
-ip_mloopback(struct ifnet *ifp, struct mbuf *m,
-	struct sockaddr_in *dst, int hlen)
+ip_mloopback(ifp, m, dst, hlen)
+	struct ifnet *ifp;
+	register struct mbuf *m;
+	register struct sockaddr_in *dst;
+	int hlen;
 {
-	struct ip *ip;
+	register struct ip *ip;
 	struct mbuf *copym;
 
-	copym = m_copypacket(m, M_DONTWAIT);
+	copym = m_copy(m, 0, M_COPYALL);
 	if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen))
 		copym = m_pullup(copym, hlen);
 	if (copym != NULL) {
diff --git a/sys/netinet/ip_var.h b/sys/netinet/ip_var.h
index 642580f..c03c446 100644
--- a/sys/netinet/ip_var.h
+++ b/sys/netinet/ip_var.h
@@ -163,8 +163,6 @@ extern struct	pr_usrreqs rip_usrreqs;
 
 int	 ip_ctloutput(struct socket *, struct sockopt *sopt);
 void	 ip_drain(void);
-int	 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
-		u_long if_hwassist_flags, int sw_csum);
 void	 ip_freemoptions(struct ip_moptions *);
 void	 ip_init(void);
 extern int	 (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *,