Add support for offloading IP/TCP/UDP checksums to NIC hardware which

supports them.

8 files changed, 219 insertions, 102 deletions

diff --git a/sys/netinet/ip_input.c b/sys/netinet/ip_input.c
index 5d1bff3..0d9273d 100644
--- a/sys/netinet/ip_input.c
+++ b/sys/netinet/ip_input.c
@@ -324,10 +324,14 @@ ip_input(struct mbuf *m)
 		}
 		ip = mtod(m, struct ip *);
 	}
-	if (hlen == sizeof(struct ip)) {
-		sum = in_cksum_hdr(ip);
+	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
+		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
 	} else {
-		sum = in_cksum(m, hlen);
+		if (hlen == sizeof(struct ip)) {
+			sum = in_cksum_hdr(ip);
+		} else {
+			sum = in_cksum(m, hlen);
+		}
 	}
 	if (sum) {
 		ipstat.ips_badsum++;
@@ -841,6 +845,9 @@ ip_reass(m, fp, where)
 	 * our data already.  If so, drop the data from the incoming
 	 * segment.  If it provides all of our data, drop us, otherwise
 	 * stick new segment in the proper place.
+	 *
+	 * If some of the data is dropped from the the preceding
+	 * segment, then it's checksum is invalidated.
 	 */
 	if (p) {
 		i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
@@ -848,6 +855,7 @@ ip_reass(m, fp, where)
 			if (i >= ip->ip_len)
 				goto dropfrag;
 			m_adj(m, i);
+			m->m_pkthdr.csum_flags = 0;
 			ip->ip_off += i;
 			ip->ip_len -= i;
 		}
@@ -870,6 +878,7 @@ ip_reass(m, fp, where)
 			GETIP(q)->ip_len -= i;
 			GETIP(q)->ip_off += i;
 			m_adj(q, i);
+			q->m_pkthdr.csum_flags = 0;
 			break;
 		}
 		nq = q->m_nextpkt;
@@ -927,6 +936,8 @@ inserted:
 		nq = q->m_nextpkt;
 		q->m_nextpkt = NULL;
 		m_cat(m, q);
+		m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
+		m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
 	}
 
 #ifdef IPDIVERT
diff --git a/sys/netinet/ip_output.c b/sys/netinet/ip_output.c
index ee361ce..b44765f 100644
--- a/sys/netinet/ip_output.c
+++ b/sys/netinet/ip_output.c
@@ -96,6 +96,7 @@ static MALLOC_DEFINE(M_IPMOPTS, "ip_moptions", "internet multicast options");
 
 u_short ip_id;
 
+static void	in_delayed_cksum(struct mbuf *m);
 static struct mbuf *ip_insertoptions __P((struct mbuf *, struct mbuf *, int *));
 static void	ip_mloopback
 	__P((struct ifnet *, struct mbuf *, struct sockaddr_in *, int));
@@ -132,7 +133,7 @@ ip_output(m0, opt, ro, flags, imo)
 	int len, off, error = 0;
 	struct sockaddr_in *dst;
 	struct in_ifaddr *ia;
-	int isbroadcast;
+	int isbroadcast, sw_csum;
 #ifdef IPSEC
 	struct route iproute;
 	struct socket *so = NULL;
@@ -692,6 +693,15 @@ pass:
 		state.ro = ro;
 	state.dst = (struct sockaddr *)dst;
 
+	/*
+	 * XXX
+	 * delayed checksums are not currently compatible with IPsec
+	 */
+	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
+		in_delayed_cksum(m);
+		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
+	}
+
 	error = ipsec4_output(&state, sp, flags);
 
 	m = state.m;
@@ -754,17 +764,29 @@ pass:
 skip_ipsec:
 #endif /*IPSEC*/
 
+	sw_csum = m->m_pkthdr.csum_flags | CSUM_IP;
+	m->m_pkthdr.csum_flags = sw_csum & ifp->if_hwassist;
+	sw_csum &= ~ifp->if_hwassist;
+	if (sw_csum & CSUM_DELAY_DATA) {
+		in_delayed_cksum(m);
+		sw_csum &= ~CSUM_DELAY_DATA;
+	}
+
 	/*
-	 * If small enough for interface, can just send directly.
+	 * If small enough for interface, or the interface will take
+	 * care of the fragmentation for us, can just send directly.
 	 */
-	if ((u_short)ip->ip_len <= ifp->if_mtu) {
+	if ((u_short)ip->ip_len <= ifp->if_mtu ||
+	    ifp->if_hwassist & CSUM_FRAGMENT) {
 		ip->ip_len = htons((u_short)ip->ip_len);
 		ip->ip_off = htons((u_short)ip->ip_off);
 		ip->ip_sum = 0;
-		if (ip->ip_vhl == IP_VHL_BORING) {
-			ip->ip_sum = in_cksum_hdr(ip);
-		} else {
-			ip->ip_sum = in_cksum(m, hlen);
+		if (sw_csum & CSUM_DELAY_IP) {
+			if (ip->ip_vhl == IP_VHL_BORING) {
+				ip->ip_sum = in_cksum_hdr(ip);
+			} else {
+				ip->ip_sum = in_cksum(m, hlen);
+			}
 		}
 		error = (*ifp->if_output)(ifp, m,
 				(struct sockaddr *)dst, ro->ro_rt);
@@ -797,9 +819,20 @@ skip_ipsec:
 		goto bad;
 	}
 
+	/*
+	 * if the interface will not calculate checksums on
+	 * fragmented packets, then do it here.
+	 */
+	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;
+	}
+
     {
 	int mhlen, firstlen = len;
 	struct mbuf **mnext = &m->m_nextpkt;
+	int nfrags = 1;
 
 	/*
 	 * Loop through length of segment after first fragment,
@@ -814,7 +847,7 @@ skip_ipsec:
 			ipstat.ips_odropped++;
 			goto sendorfree;
 		}
-		m->m_flags |= (m0->m_flags & M_MCAST);
+		m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG;
 		m->m_data += max_linkhdr;
 		mhip = mtod(m, struct ip *);
 		*mhip = *ip;
@@ -840,17 +873,27 @@ skip_ipsec:
 		}
 		m->m_pkthdr.len = mhlen + len;
 		m->m_pkthdr.rcvif = (struct ifnet *)0;
+		m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
 		mhip->ip_off = htons((u_short)mhip->ip_off);
 		mhip->ip_sum = 0;
-		if (mhip->ip_vhl == IP_VHL_BORING) {
-			mhip->ip_sum = in_cksum_hdr(mhip);
-		} else {
-			mhip->ip_sum = in_cksum(m, mhlen);
+		if (sw_csum & CSUM_DELAY_IP) {
+			if (mhip->ip_vhl == IP_VHL_BORING) {
+				mhip->ip_sum = in_cksum_hdr(mhip);
+			} else {
+				mhip->ip_sum = in_cksum(m, mhlen);
+			}
 		}
 		*mnext = m;
 		mnext = &m->m_nextpkt;
-		ipstat.ips_ofragments++;
+		nfrags++;
 	}
+	ipstat.ips_ofragments += nfrags;
+
+	/* 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's been copied out
 	 * and updating header, then send each fragment (in order).
@@ -861,10 +904,12 @@ skip_ipsec:
 	ip->ip_len = htons((u_short)m->m_pkthdr.len);
 	ip->ip_off = htons((u_short)(ip->ip_off | IP_MF));
 	ip->ip_sum = 0;
-	if (ip->ip_vhl == IP_VHL_BORING) {
-		ip->ip_sum = in_cksum_hdr(ip);
-	} else {
-		ip->ip_sum = in_cksum(m, hlen);
+	if (sw_csum & CSUM_DELAY_IP) {
+		if (ip->ip_vhl == IP_VHL_BORING) {
+			ip->ip_sum = in_cksum_hdr(ip);
+		} else {
+			ip->ip_sum = in_cksum(m, hlen);
+		}
 	}
 sendorfree:
 	for (m = m0; m; m = m0) {
@@ -898,6 +943,31 @@ bad:
 	goto done;
 }
 
+static void
+in_delayed_cksum(struct mbuf *m)
+{
+	struct ip *ip;
+	u_short csum, offset;
+
+	ip = mtod(m, struct ip *);
+	offset = IP_VHL_HL(ip->ip_vhl) << 2 ;
+	csum = in_cksum_skip(m, ip->ip_len, offset);
+	offset += m->m_pkthdr.csum_data;	/* checksum offset */
+
+	if (offset + sizeof(u_short) > m->m_len) {
+		printf("delayed m_pullup, m->len: %d  off: %d  p: %d\n",
+		    m->m_len, offset, ip->ip_p);
+		/*
+		 * XXX
+		 * this shouldn't happen, but if it does, the
+		 * correct behavior may be to insert the checksum
+		 * in the existing chain instead of rearranging it.
+		 */
+		m = m_pullup(m, offset + sizeof(u_short));
+	}
+	*(u_short *)(m->m_data + offset) = csum;
+}
+
 /*
  * Insert IP options into preformed packet.
  * Adjust IP destination as required for IP source routing,
diff --git a/sys/netinet/tcp_input.c b/sys/netinet/tcp_input.c
index 1d2d8c3..dc8ba3c 100644
--- a/sys/netinet/tcp_input.c
+++ b/sys/netinet/tcp_input.c
@@ -100,6 +100,8 @@ struct tcphdr tcp_savetcp;
 #include <netkey/key.h>
 #endif /*IPSEC*/
 
+#include <machine/in_cksum.h>
+
 MALLOC_DEFINE(M_TSEGQ, "tseg_qent", "TCP segment queue entry");
 
 static int	tcprexmtthresh = 3;
@@ -425,17 +427,27 @@ tcp_input(m, off0, proto)
 	}
 	ip = mtod(m, struct ip *);
 	ipov = (struct ipovly *)ip;
-
-	/*
-	 * Checksum extended TCP header and data.
-	 */
-	tlen = ip->ip_len;
-	len = sizeof (struct ip) + tlen;
-	bzero(ipov->ih_x1, sizeof(ipov->ih_x1));
-	ipov->ih_len = (u_short)tlen;
-	HTONS(ipov->ih_len);
 	th = (struct tcphdr *)((caddr_t)ip + off0);
-	th->th_sum = in_cksum(m, len);
+	tlen = ip->ip_len;
+
+	if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
+		if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
+                	th->th_sum = m->m_pkthdr.csum_data;
+		else
+	                th->th_sum = in_pseudo(ip->ip_src.s_addr,
+			    ip->ip_dst.s_addr, htonl(m->m_pkthdr.csum_data +
+			    ip->ip_len + IPPROTO_TCP));
+		th->th_sum ^= 0xffff;
+	} else {
+		/*
+		 * Checksum extended TCP header and data.
+		 */
+		len = sizeof (struct ip) + tlen;
+		bzero(ipov->ih_x1, sizeof(ipov->ih_x1));
+		ipov->ih_len = (u_short)tlen;
+		HTONS(ipov->ih_len);
+		th->th_sum = in_cksum(m, len);
+	}
 	if (th->th_sum) {
 		tcpstat.tcps_rcvbadsum++;
 		goto drop;
diff --git a/sys/netinet/tcp_output.c b/sys/netinet/tcp_output.c
index ae85e84..238d25a 100644
--- a/sys/netinet/tcp_output.c
+++ b/sys/netinet/tcp_output.c
@@ -80,6 +80,8 @@
 #include <netinet6/ipsec.h>
 #endif /*IPSEC*/
 
+#include <machine/in_cksum.h>
+
 #ifdef notyet
 extern struct mbuf *m_copypack();
 #endif
@@ -645,6 +647,7 @@ send:
 	ip = mtod(m, struct ip *);
 	ipov = (struct ipovly *)ip;
 	th = (struct tcphdr *)(ip + 1);
+	/* this picks up the pseudo header (w/o the length) */
 	bcopy((caddr_t)tp->t_template->tt_ipgen, (caddr_t)ip,
 	      sizeof(struct ip));
 	bcopy((caddr_t)&tp->t_template->tt_t, (caddr_t)th,
@@ -722,15 +725,15 @@ send:
 	else
 #endif /* INET6 */
       {
+	m->m_pkthdr.csum_flags = CSUM_TCP;
+	m->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum);
 	if (len + optlen)
-		ipov->ih_len = htons((u_short)(sizeof (struct tcphdr) +
-		    optlen + len));
-	th->th_sum = in_cksum(m, (int)(hdrlen + len));
-#ifdef INET6
-	/* Re-initialization for later version check */
-	ip->ip_v = IPVERSION;
-	
-#endif /* INET6 */
+		th->th_sum = in_addword(th->th_sum, 
+		    htons((u_short)(optlen + len)));
+
+	/* IP version must be set here for ipv4/ipv6 checking later */
+	KASSERT(ip->ip_v == IPVERSION,
+	    ("%s: IP version incorrect: %d", __FUNCTION__, ip->ip_v));
       }
 
 	/*
diff --git a/sys/netinet/tcp_reass.c b/sys/netinet/tcp_reass.c
index 1d2d8c3..dc8ba3c 100644
--- a/sys/netinet/tcp_reass.c
+++ b/sys/netinet/tcp_reass.c
@@ -100,6 +100,8 @@ struct tcphdr tcp_savetcp;
 #include <netkey/key.h>
 #endif /*IPSEC*/
 
+#include <machine/in_cksum.h>
+
 MALLOC_DEFINE(M_TSEGQ, "tseg_qent", "TCP segment queue entry");
 
 static int	tcprexmtthresh = 3;
@@ -425,17 +427,27 @@ tcp_input(m, off0, proto)
 	}
 	ip = mtod(m, struct ip *);
 	ipov = (struct ipovly *)ip;
-
-	/*
-	 * Checksum extended TCP header and data.
-	 */
-	tlen = ip->ip_len;
-	len = sizeof (struct ip) + tlen;
-	bzero(ipov->ih_x1, sizeof(ipov->ih_x1));
-	ipov->ih_len = (u_short)tlen;
-	HTONS(ipov->ih_len);
 	th = (struct tcphdr *)((caddr_t)ip + off0);
-	th->th_sum = in_cksum(m, len);
+	tlen = ip->ip_len;
+
+	if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
+		if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
+                	th->th_sum = m->m_pkthdr.csum_data;
+		else
+	                th->th_sum = in_pseudo(ip->ip_src.s_addr,
+			    ip->ip_dst.s_addr, htonl(m->m_pkthdr.csum_data +
+			    ip->ip_len + IPPROTO_TCP));
+		th->th_sum ^= 0xffff;
+	} else {
+		/*
+		 * Checksum extended TCP header and data.
+		 */
+		len = sizeof (struct ip) + tlen;
+		bzero(ipov->ih_x1, sizeof(ipov->ih_x1));
+		ipov->ih_len = (u_short)tlen;
+		HTONS(ipov->ih_len);
+		th->th_sum = in_cksum(m, len);
+	}
 	if (th->th_sum) {
 		tcpstat.tcps_rcvbadsum++;
 		goto drop;
diff --git a/sys/netinet/tcp_subr.c b/sys/netinet/tcp_subr.c
index ee62998..d4c552f 100644
--- a/sys/netinet/tcp_subr.c
+++ b/sys/netinet/tcp_subr.c
@@ -39,6 +39,7 @@
 #include "opt_ipsec.h"
 #include "opt_tcpdebug.h"
 
+#include <stddef.h>
 #include <sys/param.h>
 #include <sys/systm.h>
 #include <sys/callout.h>
@@ -93,6 +94,8 @@
 #include <netinet6/ipsec.h>
 #endif /*IPSEC*/
 
+#include <machine/in_cksum.h>
+
 int 	tcp_mssdflt = TCP_MSS;
 SYSCTL_INT(_net_inet_tcp, TCPCTL_MSSDFLT, mssdflt, CTLFLAG_RW, 
     &tcp_mssdflt , 0, "Default TCP Maximum Segment Size");
@@ -242,17 +245,19 @@ tcp_template(tp)
 		ip6->ip6_plen = sizeof(struct tcphdr);
 		ip6->ip6_src = inp->in6p_laddr;
 		ip6->ip6_dst = inp->in6p_faddr;
+		n->tt_t.th_sum = 0;
 	} else
 #endif
       {
-	register struct ipovly *ipov;
-
-	ipov = (struct ipovly *)n->tt_ipgen;
-	bzero(ipov->ih_x1, sizeof(ipov->ih_x1));
-	ipov->ih_pr = IPPROTO_TCP;
-	ipov->ih_len = htons(sizeof (struct tcpiphdr) - sizeof (struct ip));
-	ipov->ih_src = inp->inp_laddr;
-	ipov->ih_dst = inp->inp_faddr;
+	struct ip *ip = (struct ip *)n->tt_ipgen;
+
+	bzero(ip, sizeof(struct ip));		/* XXX overkill? */
+	ip->ip_vhl = IP_VHL_BORING;
+	ip->ip_p = IPPROTO_TCP;
+	ip->ip_src = inp->inp_laddr;
+	ip->ip_dst = inp->inp_faddr;
+	n->tt_t.th_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr,
+	    htons(sizeof(struct tcphdr) + IPPROTO_TCP));
       }
 	n->tt_t.th_sport = inp->inp_lport;
 	n->tt_t.th_dport = inp->inp_fport;
@@ -262,7 +267,6 @@ tcp_template(tp)
 	n->tt_t.th_off = 5;
 	n->tt_t.th_flags = 0;
 	n->tt_t.th_win = 0;
-	n->tt_t.th_sum = 0;
 	n->tt_t.th_urp = 0;
 	return (n);
 }
@@ -296,7 +300,6 @@ tcp_respond(tp, ipgen, th, m, ack, seq, flags)
 	struct route *ro = 0;
 	struct route sro;
 	struct ip *ip;
-	struct ipovly *ipov;
 	struct tcphdr *nth;
 #ifdef INET6
 	struct route_in6 *ro6 = 0;
@@ -311,7 +314,6 @@ tcp_respond(tp, ipgen, th, m, ack, seq, flags)
 	ip6 = ipgen;
 #endif /* INET6 */
 	ip = ipgen;
-	ipov = ipgen;
 
 	if (tp) {
 		if (!(flags & TH_RST)) {
@@ -358,7 +360,6 @@ tcp_respond(tp, ipgen, th, m, ack, seq, flags)
 	      {
 		bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
 		ip = mtod(m, struct ip *);
-		ipov = mtod(m, struct ipovly *);
 		nth = (struct tcphdr *)(ip + 1);
 	      }
 		bcopy((caddr_t)th, (caddr_t)nth, sizeof(struct tcphdr));
@@ -400,8 +401,9 @@ tcp_respond(tp, ipgen, th, m, ack, seq, flags)
 	} else
 #endif
       {
-	ipov->ih_len = htons((u_short)(sizeof (struct tcphdr) + tlen));
 	tlen += sizeof (struct tcpiphdr);
+	ip->ip_len = tlen;
+	ip->ip_ttl = ip_defttl;
       }
 	m->m_len = tlen;
 	m->m_pkthdr.len = tlen;
@@ -416,7 +418,6 @@ tcp_respond(tp, ipgen, th, m, ack, seq, flags)
 	else
 		nth->th_win = htons((u_short)win);
 	nth->th_urp = 0;
-	nth->th_sum = 0;
 #ifdef INET6
 	if (isipv6) {
 		nth->th_sum = in6_cksum(m, IPPROTO_TCP,
@@ -429,14 +430,10 @@ tcp_respond(tp, ipgen, th, m, ack, seq, flags)
 	} else
 #endif /* INET6 */
       {
-	bzero(ipov->ih_x1, sizeof(ipov->ih_x1));
-	nth->th_sum = in_cksum(m, tlen);
-#ifdef INET6
-	/* Re-initialization for later version check */
-	ip->ip_vhl = IP_MAKE_VHL(IPVERSION, 0);
-#endif /* INET6 */
-	ip->ip_len = tlen;
-	ip->ip_ttl = ip_defttl;
+        nth->th_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr,
+	    htons((u_short)(tlen - sizeof(struct ip) + ip->ip_p)));
+        m->m_pkthdr.csum_flags = CSUM_TCP;
+        m->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum);
       }
 #ifdef TCPDEBUG
 	if (tp == NULL || (tp->t_inpcb->inp_socket->so_options & SO_DEBUG))
diff --git a/sys/netinet/tcp_timewait.c b/sys/netinet/tcp_timewait.c
index ee62998..d4c552f 100644
--- a/sys/netinet/tcp_timewait.c
+++ b/sys/netinet/tcp_timewait.c
@@ -39,6 +39,7 @@
 #include "opt_ipsec.h"
 #include "opt_tcpdebug.h"
 
+#include <stddef.h>
 #include <sys/param.h>
 #include <sys/systm.h>
 #include <sys/callout.h>
@@ -93,6 +94,8 @@
 #include <netinet6/ipsec.h>
 #endif /*IPSEC*/
 
+#include <machine/in_cksum.h>
+
 int 	tcp_mssdflt = TCP_MSS;
 SYSCTL_INT(_net_inet_tcp, TCPCTL_MSSDFLT, mssdflt, CTLFLAG_RW, 
     &tcp_mssdflt , 0, "Default TCP Maximum Segment Size");
@@ -242,17 +245,19 @@ tcp_template(tp)
 		ip6->ip6_plen = sizeof(struct tcphdr);
 		ip6->ip6_src = inp->in6p_laddr;
 		ip6->ip6_dst = inp->in6p_faddr;
+		n->tt_t.th_sum = 0;
 	} else
 #endif
       {
-	register struct ipovly *ipov;
-
-	ipov = (struct ipovly *)n->tt_ipgen;
-	bzero(ipov->ih_x1, sizeof(ipov->ih_x1));
-	ipov->ih_pr = IPPROTO_TCP;
-	ipov->ih_len = htons(sizeof (struct tcpiphdr) - sizeof (struct ip));
-	ipov->ih_src = inp->inp_laddr;
-	ipov->ih_dst = inp->inp_faddr;
+	struct ip *ip = (struct ip *)n->tt_ipgen;
+
+	bzero(ip, sizeof(struct ip));		/* XXX overkill? */
+	ip->ip_vhl = IP_VHL_BORING;
+	ip->ip_p = IPPROTO_TCP;
+	ip->ip_src = inp->inp_laddr;
+	ip->ip_dst = inp->inp_faddr;
+	n->tt_t.th_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr,
+	    htons(sizeof(struct tcphdr) + IPPROTO_TCP));
       }
 	n->tt_t.th_sport = inp->inp_lport;
 	n->tt_t.th_dport = inp->inp_fport;
@@ -262,7 +267,6 @@ tcp_template(tp)
 	n->tt_t.th_off = 5;
 	n->tt_t.th_flags = 0;
 	n->tt_t.th_win = 0;
-	n->tt_t.th_sum = 0;
 	n->tt_t.th_urp = 0;
 	return (n);
 }
@@ -296,7 +300,6 @@ tcp_respond(tp, ipgen, th, m, ack, seq, flags)
 	struct route *ro = 0;
 	struct route sro;
 	struct ip *ip;
-	struct ipovly *ipov;
 	struct tcphdr *nth;
 #ifdef INET6
 	struct route_in6 *ro6 = 0;
@@ -311,7 +314,6 @@ tcp_respond(tp, ipgen, th, m, ack, seq, flags)
 	ip6 = ipgen;
 #endif /* INET6 */
 	ip = ipgen;
-	ipov = ipgen;
 
 	if (tp) {
 		if (!(flags & TH_RST)) {
@@ -358,7 +360,6 @@ tcp_respond(tp, ipgen, th, m, ack, seq, flags)
 	      {
 		bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
 		ip = mtod(m, struct ip *);
-		ipov = mtod(m, struct ipovly *);
 		nth = (struct tcphdr *)(ip + 1);
 	      }
 		bcopy((caddr_t)th, (caddr_t)nth, sizeof(struct tcphdr));
@@ -400,8 +401,9 @@ tcp_respond(tp, ipgen, th, m, ack, seq, flags)
 	} else
 #endif
       {
-	ipov->ih_len = htons((u_short)(sizeof (struct tcphdr) + tlen));
 	tlen += sizeof (struct tcpiphdr);
+	ip->ip_len = tlen;
+	ip->ip_ttl = ip_defttl;
       }
 	m->m_len = tlen;
 	m->m_pkthdr.len = tlen;
@@ -416,7 +418,6 @@ tcp_respond(tp, ipgen, th, m, ack, seq, flags)
 	else
 		nth->th_win = htons((u_short)win);
 	nth->th_urp = 0;
-	nth->th_sum = 0;
 #ifdef INET6
 	if (isipv6) {
 		nth->th_sum = in6_cksum(m, IPPROTO_TCP,
@@ -429,14 +430,10 @@ tcp_respond(tp, ipgen, th, m, ack, seq, flags)
 	} else
 #endif /* INET6 */
       {
-	bzero(ipov->ih_x1, sizeof(ipov->ih_x1));
-	nth->th_sum = in_cksum(m, tlen);
-#ifdef INET6
-	/* Re-initialization for later version check */
-	ip->ip_vhl = IP_MAKE_VHL(IPVERSION, 0);
-#endif /* INET6 */
-	ip->ip_len = tlen;
-	ip->ip_ttl = ip_defttl;
+        nth->th_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr,
+	    htons((u_short)(tlen - sizeof(struct ip) + ip->ip_p)));
+        m->m_pkthdr.csum_flags = CSUM_TCP;
+        m->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum);
       }
 #ifdef TCPDEBUG
 	if (tp == NULL || (tp->t_inpcb->inp_socket->so_options & SO_DEBUG))
diff --git a/sys/netinet/udp_usrreq.c b/sys/netinet/udp_usrreq.c
index 2ed9a36..b8700de 100644
--- a/sys/netinet/udp_usrreq.c
+++ b/sys/netinet/udp_usrreq.c
@@ -37,6 +37,7 @@
 #include "opt_ipsec.h"
 #include "opt_inet6.h"
 
+#include <stddef.h>
 #include <sys/param.h>
 #include <sys/systm.h>
 #include <sys/kernel.h>
@@ -76,6 +77,8 @@
 #include <netinet6/ipsec.h>
 #endif /*IPSEC*/
 
+#include <machine/in_cksum.h>
+
 /*
  * UDP protocol implementation.
  * Per RFC 768, August, 1980.
@@ -208,9 +211,19 @@ udp_input(m, off, proto)
 	 * Checksum extended UDP header and data.
 	 */
 	if (uh->uh_sum) {
-		bzero(((struct ipovly *)ip)->ih_x1, 9);
-		((struct ipovly *)ip)->ih_len = uh->uh_ulen;
-		uh->uh_sum = in_cksum(m, len + sizeof (struct ip));
+		if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
+			if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
+				uh->uh_sum = m->m_pkthdr.csum_data;
+			else
+	                	uh->uh_sum = in_pseudo(ip->ip_src.s_addr,
+				    ip->ip_dst.s_addr, htonl(ip->ip_len +
+				    m->m_pkthdr.csum_data + IPPROTO_UDP));
+			uh->uh_sum ^= 0xffff;
+		} else {
+			bzero(((struct ipovly *)ip)->ih_x1, 9);
+			((struct ipovly *)ip)->ih_len = uh->uh_ulen;
+			uh->uh_sum = in_cksum(m, len + sizeof (struct ip));
+		}
 		if (uh->uh_sum) {
 			udpstat.udps_badsum++;
 			m_freem(m);
@@ -679,22 +692,24 @@ udp_output(inp, m, addr, control, p)
 	 * and addresses and length put into network format.
 	 */
 	ui = mtod(m, struct udpiphdr *);
-	bzero(ui->ui_x1, sizeof(ui->ui_x1));
+	bzero(ui->ui_x1, sizeof(ui->ui_x1));	/* XXX still needed? */
 	ui->ui_pr = IPPROTO_UDP;
-	ui->ui_len = htons((u_short)len + sizeof (struct udphdr));
 	ui->ui_src = inp->inp_laddr;
 	ui->ui_dst = inp->inp_faddr;
 	ui->ui_sport = inp->inp_lport;
 	ui->ui_dport = inp->inp_fport;
-	ui->ui_ulen = ui->ui_len;
+	ui->ui_ulen = htons((u_short)len + sizeof(struct udphdr));
 
 	/*
-	 * Stuff checksum and output datagram.
+	 * Set up checksum and output datagram.
 	 */
-	ui->ui_sum = 0;
 	if (udpcksum) {
-	    if ((ui->ui_sum = in_cksum(m, sizeof (struct udpiphdr) + len)) == 0)
-		ui->ui_sum = 0xffff;
+        	ui->ui_sum = in_pseudo(ui->ui_src.s_addr, ui->ui_dst.s_addr,
+		    htons((u_short)len + sizeof(struct udphdr) + IPPROTO_UDP));
+		m->m_pkthdr.csum_flags = CSUM_UDP;
+		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
+	} else {
+		ui->ui_sum = 0;
 	}
 	((struct ip *)ui)->ip_len = sizeof (struct udpiphdr) + len;
 	((struct ip *)ui)->ip_ttl = inp->inp_ip_ttl;	/* XXX */