Make SOCK_RAW sockets to be truly raw, not modifying received and sent

packets at all. Swapping byte order on SOCK_RAW was actually a bug, an
artifact from the BSD network stack, that used to convert a packet to
native byte order once it is received by kernel.

Other operating systems didn't follow this, and later other BSD
descendants fixed this, leaving us alone with the bug. Now it is
clear that we should fix the bug.

In collaboration with:	Olivier Cochard-Labbé <olivier cochard.me>
See also:		https://wiki.freebsd.org/SOCK_RAW
Sponsored by:		Nginx, Inc.

share/man/man4/ip.4

@@ -28,7 +28,7 @@
 .\"     @(#)ip.4	8.2 (Berkeley) 11/30/93
 .\" $FreeBSD$
 .\"
-.Dd October 12, 2012
+.Dd September 1, 2014
 .Dt IP 4
 .Os
 .Sh NAME
@@ -755,13 +755,11 @@ number the socket is created with),
 unless the
 .Dv IP_HDRINCL
 option has been set.
-Incoming packets are received with
+Unlike in previous
+.Bx
+releases, incoming packets are received with
 .Tn IP
-header and options intact, except for
-.Va ip_len
-and
-.Va ip_off
-fields converted to host byte order.
+header and options intact, leaving all fields in network byte order.
 .Pp
 .Dv IP_HDRINCL
 indicates the complete IP header is included with the data
@@ -784,17 +782,16 @@ the fields of the IP header, including the following:
 ip->ip_v = IPVERSION;
 ip->ip_hl = hlen >> 2;
 ip->ip_id = 0;  /* 0 means kernel set appropriate value */
-ip->ip_off = offset;
+ip->ip_off = htons(offset);
+ip->ip_len = htons(len);
 .Ed
 .Pp
-The
+The packet should be provided as is to be sent over wire.
+This implies all fields, including
 .Va ip_len
 and
 .Va ip_off
-fields
-.Em must
-be provided in host byte order.
-All other fields must be provided in network byte order.
+to be in network byte order.
 See
 .Xr byteorder 3
 for more information on network byte order.
@@ -891,3 +888,16 @@ packets received on raw IP sockets had the
 subtracted from the
 .Va ip_len
 field.
+.Pp
+Before
+.Fx 11.0
+packets received on raw IP sockets had the
+.Va ip_len
+and
+.Va ip_off
+fields converted to host byte order.
+Packets written to raw IP sockets were expected to have
+.Va ip_len
+and
+.Va ip_off
+in host byte order.

sys/netinet/raw_ip.c

@@ -290,11 +290,6 @@ rip_input(struct mbuf **mp, int *offp, int proto)
 	last = NULL;
 
 	ifp = m->m_pkthdr.rcvif;
-	/*
-	 * Applications on raw sockets expect host byte order.
-	 */
-	ip->ip_len = ntohs(ip->ip_len);
-	ip->ip_off = ntohs(ip->ip_off);
 
 	hash = INP_PCBHASH_RAW(proto, ip->ip_src.s_addr,
 	    ip->ip_dst.s_addr, V_ripcbinfo.ipi_hashmask);
@@ -504,8 +499,8 @@ rip_output(struct mbuf *m, struct socket *so, ...)
 		 * and don't allow packet length sizes that will crash.
 		 */
 		if (((ip->ip_hl != (sizeof (*ip) >> 2)) && inp->inp_options)
-		    || (ip->ip_len > m->m_pkthdr.len)
-		    || (ip->ip_len < (ip->ip_hl << 2))) {
+		    || (ntohs(ip->ip_len) > m->m_pkthdr.len)
+		    || (ntohs(ip->ip_len) < (ip->ip_hl << 2))) {
 			INP_RUNLOCK(inp);
 			m_freem(m);
 			return (EINVAL);
@@ -513,13 +508,6 @@ rip_output(struct mbuf *m, struct socket *so, ...)
 		if (ip->ip_id == 0)
 			ip->ip_id = ip_newid();
 
-		/*
-		 * Applications on raw sockets pass us packets
-		 * in host byte order.
-		 */
-		ip->ip_len = htons(ip->ip_len);
-		ip->ip_off = htons(ip->ip_off);
-
 		/*
 		 * XXX prevent ip_output from overwriting header fields.
 		 */

sys/sys/param.h

@@ -58,7 +58,7 @@
  *		in the range 5 to 9.
  */
 #undef __FreeBSD_version
-#define __FreeBSD_version 1100029	/* Master, propagated to newvers */
+#define __FreeBSD_version 1100030	/* Master, propagated to newvers */
 
 /*
  * __FreeBSD_kernel__ indicates that this system uses the kernel of FreeBSD,

usr.sbin/traceroute/Makefile

@@ -13,7 +13,7 @@ CLEANFILES=	version.c
 CFLAGS+= -DHAVE_SYS_SELECT_H=1 -DHAVE_SYS_SOCKIO_H=1 \
 	 -DHAVE_NET_ROUTE_H=1 -DHAVE_NET_IF_DL_H=1 \
 	 -DHAVE_STRERROR=1 -DHAVE_USLEEP=1 \
-	 -DHAVE_SYS_SYSCTL_H=1 \
+	 -DHAVE_SYS_SYSCTL_H=1 -DBYTESWAP_IP_HDR=1 \
 	 -DHAVE_SETLINEBUF=1 -DHAVE_RAW_OPTIONS=1 \
 	 -DHAVE_SOCKADDR_SA_LEN=1 -DHAVE_ICMP_NEXTMTU=1
 .if !defined(TRACEROUTE_NO_IPSEC)