ip_output.c revision 237910
162583Sitojun/*- 278064Sume * Copyright (c) 1982, 1986, 1988, 1990, 1993 362583Sitojun * The Regents of the University of California. All rights reserved. 455505Sshin * 555505Sshin * Redistribution and use in source and binary forms, with or without 655505Sshin * modification, are permitted provided that the following conditions 762583Sitojun * are met: 855505Sshin * 1. Redistributions of source code must retain the above copyright 955505Sshin * notice, this list of conditions and the following disclaimer. 1055505Sshin * 2. Redistributions in binary form must reproduce the above copyright 1155505Sshin * notice, this list of conditions and the following disclaimer in the 1255505Sshin * documentation and/or other materials provided with the distribution. 1355505Sshin * 4. Neither the name of the University nor the names of its contributors 1455505Sshin * may be used to endorse or promote products derived from this software 1555505Sshin * without specific prior written permission. 1655505Sshin * 1755505Sshin * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 1855505Sshin * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 1962583Sitojun * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 2055505Sshin * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 2155505Sshin * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 2255505Sshin * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 2355505Sshin * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 2455505Sshin * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 2555505Sshin * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 2655505Sshin * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 2755505Sshin * SUCH DAMAGE. 2855505Sshin * 2955505Sshin * @(#)ip_output.c 8.3 (Berkeley) 1/21/94 3055505Sshin */ 3155505Sshin 3255505Sshin#include <sys/cdefs.h> 3355505Sshin__FBSDID("$FreeBSD: stable/9/sys/netinet/ip_output.c 237910 2012-07-01 08:47:15Z tuexen $"); 3455505Sshin 3555505Sshin#include "opt_ipfw.h" 3655505Sshin#include "opt_ipsec.h" 3755505Sshin#include "opt_route.h" 3855505Sshin#include "opt_mbuf_stress_test.h" 3955505Sshin#include "opt_mpath.h" 4055505Sshin#include "opt_sctp.h" 4155505Sshin 4255505Sshin#include <sys/param.h> 4355505Sshin#include <sys/systm.h> 4455505Sshin#include <sys/kernel.h> 4555505Sshin#include <sys/malloc.h> 4655505Sshin#include <sys/mbuf.h> 4755505Sshin#include <sys/priv.h> 4855505Sshin#include <sys/proc.h> 4955505Sshin#include <sys/protosw.h> 5055505Sshin#include <sys/socket.h> 5155505Sshin#include <sys/socketvar.h> 5255505Sshin#include <sys/sysctl.h> 5355505Sshin#include <sys/ucred.h> 5462583Sitojun 5555505Sshin#include <net/if.h> 5662583Sitojun#include <net/if_llatbl.h> 5762583Sitojun#include <net/netisr.h> 5862583Sitojun#include <net/pfil.h> 5962583Sitojun#include <net/route.h> 6062583Sitojun#include <net/flowtable.h> 6162583Sitojun#ifdef RADIX_MPATH 6262583Sitojun#include <net/radix_mpath.h> 6362583Sitojun#endif 6462583Sitojun#include <net/vnet.h> 6562583Sitojun 6678064Sume#include <netinet/in.h> 6778064Sume#include <netinet/in_systm.h> 6862583Sitojun#include <netinet/ip.h> 6955505Sshin#include <netinet/in_pcb.h> 7055505Sshin#include <netinet/in_var.h> 7155505Sshin#include <netinet/ip_var.h> 7262583Sitojun#include <netinet/ip_options.h> 7355505Sshin#ifdef SCTP 7455505Sshin#include <netinet/sctp.h> 7555505Sshin#include <netinet/sctp_crc32.h> 7662583Sitojun#endif 7762583Sitojun 7862583Sitojun#ifdef IPSEC 7962583Sitojun#include <netinet/ip_ipsec.h> 8062583Sitojun#include <netipsec/ipsec.h> 8162583Sitojun#endif /* IPSEC*/ 8262583Sitojun 8362583Sitojun#include <machine/in_cksum.h> 8478064Sume 8562583Sitojun#include <security/mac/mac_framework.h> 8655505Sshin 8762583SitojunVNET_DEFINE(u_short, ip_id); 8862583Sitojun 8955505Sshin#ifdef MBUF_STRESS_TEST 9078064Sumestatic int mbuf_frag_size = 0; 9178064SumeSYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW, 9262583Sitojun &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size"); 9355505Sshin#endif 9462583Sitojun 9555505Sshinstatic void ip_mloopback 9655505Sshin (struct ifnet *, struct mbuf *, struct sockaddr_in *, int); 9755505Sshin 9855505Sshin 9955505Sshinextern int in_mcast_loop; 10055505Sshinextern struct protosw inetsw[]; 10155505Sshin 10255505Sshin/* 10355505Sshin * IP output. The packet in mbuf chain m contains a skeletal IP 10455505Sshin * header (with len, off, ttl, proto, tos, src, dst). 10562583Sitojun * ip_len and ip_off are in host format. 10655505Sshin * The mbuf chain containing the packet will be freed. 10755505Sshin * The mbuf opt, if present, will not be freed. 10855505Sshin * In the IP forwarding case, the packet will arrive with options already 10955505Sshin * inserted, so must have a NULL opt pointer. 11055505Sshin */ 11155505Sshinint 11255505Sshinip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags, 11355505Sshin struct ip_moptions *imo, struct inpcb *inp) 11455505Sshin{ 11555505Sshin struct ip *ip; 11655505Sshin struct ifnet *ifp = NULL; /* keep compiler happy */ 11755505Sshin struct mbuf *m0; 11855505Sshin int hlen = sizeof (struct ip); 11955505Sshin int mtu; 12078064Sume int n; /* scratchpad */ 12178064Sume int error = 0; 12278064Sume int nortfree = 0; 12355505Sshin struct sockaddr_in *dst; 12455505Sshin struct in_ifaddr *ia = NULL; 12555505Sshin int isbroadcast, sw_csum; 12655505Sshin struct route iproute; 12755505Sshin struct rtentry *rte; /* cache for ro->ro_rt */ 12855505Sshin struct in_addr odst; 12955505Sshin#ifdef IPFIREWALL_FORWARD 13055505Sshin struct m_tag *fwd_tag = NULL; 13155505Sshin#endif 13255505Sshin#ifdef IPSEC 13355505Sshin int no_route_but_check_spd = 0; 13455505Sshin#endif 13555505Sshin M_ASSERTPKTHDR(m); 13655505Sshin 13755505Sshin if (inp != NULL) { 13855505Sshin INP_LOCK_ASSERT(inp); 13955505Sshin M_SETFIB(m, inp->inp_inc.inc_fibnum); 14055505Sshin if (inp->inp_flags & (INP_HW_FLOWID|INP_SW_FLOWID)) { 14155505Sshin m->m_pkthdr.flowid = inp->inp_flowid; 14255505Sshin m->m_flags |= M_FLOWID; 14355505Sshin } 14455505Sshin } 14555505Sshin 14655505Sshin if (ro == NULL) { 14755505Sshin ro = &iproute; 14855505Sshin bzero(ro, sizeof (*ro)); 14955505Sshin 15055505Sshin#ifdef FLOWTABLE 15162583Sitojun { 15278064Sume struct flentry *fle; 15362583Sitojun 15455505Sshin /* 15555505Sshin * The flow table returns route entries valid for up to 30 15655505Sshin * seconds; we rely on the remainder of ip_output() taking no 15755505Sshin * longer than that long for the stability of ro_rt. The 15855505Sshin * flow ID assignment must have happened before this point. 15955505Sshin */ 16055505Sshin if ((fle = flowtable_lookup_mbuf(V_ip_ft, m, AF_INET)) != NULL) { 16155505Sshin flow_to_route(fle, ro); 16255505Sshin nortfree = 1; 16355505Sshin } 16455505Sshin } 16555505Sshin#endif 16655505Sshin } 16755505Sshin 16855505Sshin if (opt) { 16955505Sshin int len = 0; 17055505Sshin m = ip_insertoptions(m, opt, &len); 17155505Sshin if (len != 0) 17255505Sshin hlen = len; /* ip->ip_hl is updated above */ 17355505Sshin } 17455505Sshin ip = mtod(m, struct ip *); 17555505Sshin 17655505Sshin /* 17755505Sshin * Fill in IP header. If we are not allowing fragmentation, 17855505Sshin * then the ip_id field is meaningless, but we don't set it 17955505Sshin * to zero. Doing so causes various problems when devices along 18055505Sshin * the path (routers, load balancers, firewalls, etc.) illegally 18155505Sshin * disable DF on our packet. Note that a 16-bit counter 18262583Sitojun * will wrap around in less than 10 seconds at 100 Mbit/s on a 18362583Sitojun * medium with MTU 1500. See Steven M. Bellovin, "A Technique 18455505Sshin * for Counting NATted Hosts", Proc. IMW'02, available at 18562583Sitojun * <http://www.cs.columbia.edu/~smb/papers/fnat.pdf>. 18662583Sitojun */ 18762583Sitojun if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 18855505Sshin ip->ip_v = IPVERSION; 18955505Sshin ip->ip_hl = hlen >> 2; 19062583Sitojun ip->ip_id = ip_newid(); 19155505Sshin IPSTAT_INC(ips_localout); 19255505Sshin } else { 19355505Sshin /* Header already set, fetch hlen from there */ 19455505Sshin hlen = ip->ip_hl << 2; 19555505Sshin } 19655505Sshin 19755505Sshin dst = (struct sockaddr_in *)&ro->ro_dst; 19855505Sshinagain: 19955505Sshin /* 20055505Sshin * If there is a cached route, 20155505Sshin * check that it is to the same destination 20255505Sshin * and is still up. If not, free it and try again. 20355505Sshin * The address family should also be checked in case of sharing the 20455505Sshin * cache with IPv6. 20555505Sshin */ 20655505Sshin rte = ro->ro_rt; 20755505Sshin if (rte && ((rte->rt_flags & RTF_UP) == 0 || 20855505Sshin rte->rt_ifp == NULL || 20955505Sshin !RT_LINK_IS_UP(rte->rt_ifp) || 21078064Sume dst->sin_family != AF_INET || 21155505Sshin dst->sin_addr.s_addr != ip->ip_dst.s_addr)) { 21255505Sshin if (!nortfree) 21355505Sshin RTFREE(rte); 21455505Sshin rte = ro->ro_rt = (struct rtentry *)NULL; 21555505Sshin ro->ro_lle = (struct llentry *)NULL; 21655505Sshin } 21755505Sshin#ifdef IPFIREWALL_FORWARD 21855505Sshin if (rte == NULL && fwd_tag == NULL) { 21955505Sshin#else 22055505Sshin if (rte == NULL) { 22155505Sshin#endif 22255505Sshin bzero(dst, sizeof(*dst)); 22355505Sshin dst->sin_family = AF_INET; 22455505Sshin dst->sin_len = sizeof(*dst); 22555505Sshin dst->sin_addr = ip->ip_dst; 22655505Sshin } 22755505Sshin /* 22855505Sshin * If routing to interface only, short circuit routing lookup. 22955505Sshin * The use of an all-ones broadcast address implies this; an 23055505Sshin * interface is specified by the broadcast address of an interface, 23155505Sshin * or the destination address of a ptp interface. 23255505Sshin */ 23355505Sshin if (flags & IP_SENDONES) { 23455505Sshin if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst)))) == NULL && 23555505Sshin (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL) { 23655505Sshin IPSTAT_INC(ips_noroute); 23755505Sshin error = ENETUNREACH; 23855505Sshin goto bad; 23955505Sshin } 24055505Sshin ip->ip_dst.s_addr = INADDR_BROADCAST; 24155505Sshin dst->sin_addr = ip->ip_dst; 24255505Sshin ifp = ia->ia_ifp; 24355505Sshin ip->ip_ttl = 1; 24455505Sshin isbroadcast = 1; 24555505Sshin } else if (flags & IP_ROUTETOIF) { 24655505Sshin if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL && 24755505Sshin (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0))) == NULL) { 24855505Sshin IPSTAT_INC(ips_noroute); 24955505Sshin error = ENETUNREACH; 25055505Sshin goto bad; 25155505Sshin } 25255505Sshin ifp = ia->ia_ifp; 25355505Sshin ip->ip_ttl = 1; 25455505Sshin isbroadcast = in_broadcast(dst->sin_addr, ifp); 25555505Sshin } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) && 25655505Sshin imo != NULL && imo->imo_multicast_ifp != NULL) { 25755505Sshin /* 25855505Sshin * Bypass the normal routing lookup for multicast 25955505Sshin * packets if the interface is specified. 26055505Sshin */ 26155505Sshin ifp = imo->imo_multicast_ifp; 26255505Sshin IFP_TO_IA(ifp, ia); 26355505Sshin isbroadcast = 0; /* fool gcc */ 26455505Sshin } else { 26555505Sshin /* 26655505Sshin * We want to do any cloning requested by the link layer, 26755505Sshin * as this is probably required in all cases for correct 26855505Sshin * operation (as it is for ARP). 26955505Sshin */ 27055505Sshin if (rte == NULL) { 27155505Sshin#ifdef RADIX_MPATH 27255505Sshin rtalloc_mpath_fib(ro, 27355505Sshin ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr), 27455505Sshin inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m)); 27555505Sshin#else 27655505Sshin in_rtalloc_ign(ro, 0, 27755505Sshin inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m)); 27855505Sshin#endif 27955505Sshin rte = ro->ro_rt; 28055505Sshin } 28155505Sshin if (rte == NULL || 28255505Sshin rte->rt_ifp == NULL || 28355505Sshin !RT_LINK_IS_UP(rte->rt_ifp)) { 28455505Sshin#ifdef IPSEC 28555505Sshin /* 28678064Sume * There is no route for this packet, but it is 28755505Sshin * possible that a matching SPD entry exists. 28855505Sshin */ 28955505Sshin no_route_but_check_spd = 1; 29055505Sshin mtu = 0; /* Silence GCC warning. */ 29155505Sshin goto sendit; 29255505Sshin#endif 29355505Sshin IPSTAT_INC(ips_noroute); 29455505Sshin error = EHOSTUNREACH; 29555505Sshin goto bad; 29655505Sshin } 29755505Sshin ia = ifatoia(rte->rt_ifa); 29855505Sshin ifa_ref(&ia->ia_ifa); 29955505Sshin ifp = rte->rt_ifp; 30055505Sshin rte->rt_rmx.rmx_pksent++; 30155505Sshin if (rte->rt_flags & RTF_GATEWAY) 30255505Sshin dst = (struct sockaddr_in *)rte->rt_gateway; 30355505Sshin if (rte->rt_flags & RTF_HOST) 30455505Sshin isbroadcast = (rte->rt_flags & RTF_BROADCAST); 30555505Sshin else 30655505Sshin isbroadcast = in_broadcast(dst->sin_addr, ifp); 30755505Sshin } 30855505Sshin /* 30955505Sshin * Calculate MTU. If we have a route that is up, use that, 31055505Sshin * otherwise use the interface's MTU. 31155505Sshin */ 31255505Sshin if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST))) { 31355505Sshin /* 31455505Sshin * This case can happen if the user changed the MTU 31555505Sshin * of an interface after enabling IP on it. Because 31655505Sshin * most netifs don't keep track of routes pointing to 31755505Sshin * them, there is no way for one to update all its 31855505Sshin * routes when the MTU is changed. 31955505Sshin */ 32055505Sshin if (rte->rt_rmx.rmx_mtu > ifp->if_mtu) 32155505Sshin rte->rt_rmx.rmx_mtu = ifp->if_mtu; 32255505Sshin mtu = rte->rt_rmx.rmx_mtu; 32355505Sshin } else { 32455505Sshin mtu = ifp->if_mtu; 32555505Sshin } 32655505Sshin /* Catch a possible divide by zero later. */ 32755505Sshin KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p", 32855505Sshin __func__, mtu, rte, (rte != NULL) ? rte->rt_flags : 0, ifp)); 32955505Sshin if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 33055505Sshin m->m_flags |= M_MCAST; 33155505Sshin /* 33255505Sshin * IP destination address is multicast. Make sure "dst" 33355505Sshin * still points to the address in "ro". (It may have been 33455505Sshin * changed to point to a gateway address, above.) 33555505Sshin */ 33655505Sshin dst = (struct sockaddr_in *)&ro->ro_dst; 33755505Sshin /* 33855505Sshin * See if the caller provided any multicast options 33955505Sshin */ 34055505Sshin if (imo != NULL) { 34162583Sitojun ip->ip_ttl = imo->imo_multicast_ttl; 34255505Sshin if (imo->imo_multicast_vif != -1) 34362583Sitojun ip->ip_src.s_addr = 34462583Sitojun ip_mcast_src ? 34555505Sshin ip_mcast_src(imo->imo_multicast_vif) : 34655505Sshin INADDR_ANY; 34755505Sshin } else 34855505Sshin ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 34955505Sshin /* 35055505Sshin * Confirm that the outgoing interface supports multicast. 35155505Sshin */ 35255505Sshin if ((imo == NULL) || (imo->imo_multicast_vif == -1)) { 35355505Sshin if ((ifp->if_flags & IFF_MULTICAST) == 0) { 35455505Sshin IPSTAT_INC(ips_noroute); 35555505Sshin error = ENETUNREACH; 35655505Sshin goto bad; 35755505Sshin } 35855505Sshin } 35955505Sshin /* 36055505Sshin * If source address not specified yet, use address 36155505Sshin * of outgoing interface. 36255505Sshin */ 36362583Sitojun if (ip->ip_src.s_addr == INADDR_ANY) { 36455505Sshin /* Interface may have no addresses. */ 36562583Sitojun if (ia != NULL) 36662583Sitojun ip->ip_src = IA_SIN(ia)->sin_addr; 36755505Sshin } 36855505Sshin 36955505Sshin if ((imo == NULL && in_mcast_loop) || 37055505Sshin (imo && imo->imo_multicast_loop)) { 37155505Sshin /* 37255505Sshin * Loop back multicast datagram if not expressly 37355505Sshin * forbidden to do so, even if we are not a member 37455505Sshin * of the group; ip_input() will filter it later, 37555505Sshin * thus deferring a hash lookup and mutex acquisition 37655505Sshin * at the expense of a cheap copy using m_copym(). 37755505Sshin */ 37855505Sshin ip_mloopback(ifp, m, dst, hlen); 37955505Sshin } else { 38055505Sshin /* 38155505Sshin * If we are acting as a multicast router, perform 38255505Sshin * multicast forwarding as if the packet had just 38355505Sshin * arrived on the interface to which we are about 38455505Sshin * to send. The multicast forwarding function 38555505Sshin * recursively calls this function, using the 38655505Sshin * IP_FORWARDING flag to prevent infinite recursion. 38755505Sshin * 38855505Sshin * Multicasts that are looped back by ip_mloopback(), 38955505Sshin * above, will be forwarded by the ip_input() routine, 39055505Sshin * if necessary. 39155505Sshin */ 39255505Sshin if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) { 39355505Sshin /* 39455505Sshin * If rsvp daemon is not running, do not 39555505Sshin * set ip_moptions. This ensures that the packet 39655505Sshin * is multicast and not just sent down one link 39755505Sshin * as prescribed by rsvpd. 39855505Sshin */ 39955505Sshin if (!V_rsvp_on) 40055505Sshin imo = NULL; 40155505Sshin if (ip_mforward && 40255505Sshin ip_mforward(ip, ifp, m, imo) != 0) { 40355505Sshin m_freem(m); 40455505Sshin goto done; 40555505Sshin } 40655505Sshin } 40755505Sshin } 40855505Sshin 40955505Sshin /* 41055505Sshin * Multicasts with a time-to-live of zero may be looped- 41155505Sshin * back, above, but must not be transmitted on a network. 41255505Sshin * Also, multicasts addressed to the loopback interface 41355505Sshin * are not sent -- the above call to ip_mloopback() will 41455505Sshin * loop back a copy. ip_input() will drop the copy if 41555505Sshin * this host does not belong to the destination group on 41655505Sshin * the loopback interface. 41755505Sshin */ 41855505Sshin if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) { 41955505Sshin m_freem(m); 42055505Sshin goto done; 42155505Sshin } 42255505Sshin 42355505Sshin goto sendit; 42455505Sshin } 42555505Sshin 42655505Sshin /* 42755505Sshin * If the source address is not specified yet, use the address 42855505Sshin * of the outoing interface. 42955505Sshin */ 43055505Sshin if (ip->ip_src.s_addr == INADDR_ANY) { 43155505Sshin /* Interface may have no addresses. */ 43255505Sshin if (ia != NULL) { 43355505Sshin ip->ip_src = IA_SIN(ia)->sin_addr; 43455505Sshin } 43555505Sshin } 43655505Sshin 43755505Sshin /* 43855505Sshin * Verify that we have any chance at all of being able to queue the 43955505Sshin * packet or packet fragments, unless ALTQ is enabled on the given 44055505Sshin * interface in which case packetdrop should be done by queueing. 44155505Sshin */ 44255505Sshin n = ip->ip_len / mtu + 1; /* how many fragments ? */ 44362583Sitojun if ( 44455505Sshin#ifdef ALTQ 44562583Sitojun (!ALTQ_IS_ENABLED(&ifp->if_snd)) && 44662583Sitojun#endif /* ALTQ */ 44755505Sshin (ifp->if_snd.ifq_len + n) >= ifp->if_snd.ifq_maxlen ) { 44855505Sshin error = ENOBUFS; 44955505Sshin IPSTAT_INC(ips_odropped); 45055505Sshin ifp->if_snd.ifq_drops += n; 45155505Sshin goto bad; 45255505Sshin } 45355505Sshin 45462583Sitojun /* 45555505Sshin * Look for broadcast address and 45662583Sitojun * verify user is allowed to send 45762583Sitojun * such a packet. 45855505Sshin */ 45955505Sshin if (isbroadcast) { 46055505Sshin if ((ifp->if_flags & IFF_BROADCAST) == 0) { 46155505Sshin error = EADDRNOTAVAIL; 46255505Sshin goto bad; 46355505Sshin } 46455505Sshin if ((flags & IP_ALLOWBROADCAST) == 0) { 46555505Sshin error = EACCES; 46655505Sshin goto bad; 46755505Sshin } 46855505Sshin /* don't allow broadcast messages to be fragmented */ 46955505Sshin if (ip->ip_len > mtu) { 47055505Sshin error = EMSGSIZE; 47155505Sshin goto bad; 47255505Sshin } 47355505Sshin m->m_flags |= M_BCAST; 47455505Sshin } else { 47555505Sshin m->m_flags &= ~M_BCAST; 47655505Sshin } 47755505Sshin 47855505Sshinsendit: 47955505Sshin#ifdef IPSEC 48055505Sshin switch(ip_ipsec_output(&m, inp, &flags, &error)) { 48155505Sshin case 1: 48255505Sshin goto bad; 48355505Sshin case -1: 48455505Sshin goto done; 48555505Sshin case 0: 48655505Sshin default: 48755505Sshin break; /* Continue with packet processing. */ 48855505Sshin } 48955505Sshin /* 49055505Sshin * Check if there was a route for this packet; return error if not. 49155505Sshin */ 49255505Sshin if (no_route_but_check_spd) { 49355505Sshin IPSTAT_INC(ips_noroute); 49455505Sshin error = EHOSTUNREACH; 49555505Sshin goto bad; 49655505Sshin } 49755505Sshin /* Update variables that are affected by ipsec4_output(). */ 49855505Sshin ip = mtod(m, struct ip *); 49955505Sshin hlen = ip->ip_hl << 2; 50055505Sshin#endif /* IPSEC */ 50155505Sshin 50255505Sshin /* Jump over all PFIL processing if hooks are not active. */ 50355505Sshin if (!PFIL_HOOKED(&V_inet_pfil_hook)) 50455505Sshin goto passout; 50555505Sshin 50655505Sshin /* Run through list of hooks for output packets. */ 50755505Sshin odst.s_addr = ip->ip_dst.s_addr; 50855505Sshin error = pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_OUT, inp); 50955505Sshin if (error != 0 || m == NULL) 51055505Sshin goto done; 51155505Sshin 51255505Sshin ip = mtod(m, struct ip *); 51355505Sshin 51455505Sshin /* See if destination IP address was changed by packet filter. */ 51555505Sshin if (odst.s_addr != ip->ip_dst.s_addr) { 51655505Sshin m->m_flags |= M_SKIP_FIREWALL; 51755505Sshin /* If destination is now ourself drop to ip_input(). */ 51855505Sshin if (in_localip(ip->ip_dst)) { 51955505Sshin m->m_flags |= M_FASTFWD_OURS; 52055505Sshin if (m->m_pkthdr.rcvif == NULL) 52155505Sshin m->m_pkthdr.rcvif = V_loif; 52255505Sshin if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 52355505Sshin m->m_pkthdr.csum_flags |= 52455505Sshin CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 52555505Sshin m->m_pkthdr.csum_data = 0xffff; 52655505Sshin } 52755505Sshin m->m_pkthdr.csum_flags |= 52855505Sshin CSUM_IP_CHECKED | CSUM_IP_VALID; 52955505Sshin#ifdef SCTP 53055505Sshin if (m->m_pkthdr.csum_flags & CSUM_SCTP) 53155505Sshin m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID; 53255505Sshin#endif 53355505Sshin error = netisr_queue(NETISR_IP, m); 53455505Sshin goto done; 53555505Sshin } else 53655505Sshin goto again; /* Redo the routing table lookup. */ 53755505Sshin } 53855505Sshin 53955505Sshin#ifdef IPFIREWALL_FORWARD 54055505Sshin /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */ 54155505Sshin if (m->m_flags & M_FASTFWD_OURS) { 54255505Sshin if (m->m_pkthdr.rcvif == NULL) 54355505Sshin m->m_pkthdr.rcvif = V_loif; 54455505Sshin if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 54555505Sshin m->m_pkthdr.csum_flags |= 54655505Sshin CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 54755505Sshin m->m_pkthdr.csum_data = 0xffff; 54855505Sshin } 54955505Sshin#ifdef SCTP 55055505Sshin if (m->m_pkthdr.csum_flags & CSUM_SCTP) 55155505Sshin m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID; 55255505Sshin#endif 55378064Sume m->m_pkthdr.csum_flags |= 55455505Sshin CSUM_IP_CHECKED | CSUM_IP_VALID; 55555505Sshin 55655505Sshin error = netisr_queue(NETISR_IP, m); 55755505Sshin goto done; 55855505Sshin } 55955505Sshin /* Or forward to some other address? */ 56055505Sshin fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL); 56155505Sshin if (fwd_tag) { 56255505Sshin dst = (struct sockaddr_in *)&ro->ro_dst; 56355505Sshin bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in)); 56455505Sshin m->m_flags |= M_SKIP_FIREWALL; 56555505Sshin m_tag_delete(m, fwd_tag); 56655505Sshin goto again; 56755505Sshin } 56855505Sshin#endif /* IPFIREWALL_FORWARD */ 56955505Sshin 57055505Sshinpassout: 57155505Sshin /* 127/8 must not appear on wire - RFC1122. */ 57255505Sshin if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 57355505Sshin (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 57455505Sshin if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 57555505Sshin IPSTAT_INC(ips_badaddr); 57655505Sshin error = EADDRNOTAVAIL; 57755505Sshin goto bad; 57855505Sshin } 57955505Sshin } 58055505Sshin 58155505Sshin m->m_pkthdr.csum_flags |= CSUM_IP; 58255505Sshin sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist; 58355505Sshin if (sw_csum & CSUM_DELAY_DATA) { 58455505Sshin in_delayed_cksum(m); 58555505Sshin sw_csum &= ~CSUM_DELAY_DATA; 58655505Sshin } 58755505Sshin#ifdef SCTP 58878064Sume if (sw_csum & CSUM_SCTP) { 58978064Sume sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2)); 59078064Sume sw_csum &= ~CSUM_SCTP; 59178064Sume } 59278064Sume#endif 59378064Sume m->m_pkthdr.csum_flags &= ifp->if_hwassist; 59478064Sume 59578064Sume /* 59678064Sume * If small enough for interface, or the interface will take 59778064Sume * care of the fragmentation for us, we can just send directly. 59878064Sume */ 59978064Sume if (ip->ip_len <= mtu || 60078064Sume (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0 || 60178064Sume ((ip->ip_off & IP_DF) == 0 && (ifp->if_hwassist & CSUM_FRAGMENT))) { 60278064Sume ip->ip_len = htons(ip->ip_len); 60378064Sume ip->ip_off = htons(ip->ip_off); 60478064Sume ip->ip_sum = 0; 60578064Sume if (sw_csum & CSUM_DELAY_IP) 60678064Sume ip->ip_sum = in_cksum(m, hlen); 60778064Sume 60878064Sume /* 60978064Sume * Record statistics for this interface address. 61078064Sume * With CSUM_TSO the byte/packet count will be slightly 61178064Sume * incorrect because we count the IP+TCP headers only 61278064Sume * once instead of for every generated packet. 61378064Sume */ 61478064Sume if (!(flags & IP_FORWARDING) && ia) { 61578064Sume if (m->m_pkthdr.csum_flags & CSUM_TSO) 61678064Sume ia->ia_ifa.if_opackets += 61778064Sume m->m_pkthdr.len / m->m_pkthdr.tso_segsz; 61878064Sume else 61978064Sume ia->ia_ifa.if_opackets++; 62078064Sume ia->ia_ifa.if_obytes += m->m_pkthdr.len; 62178064Sume } 62278064Sume#ifdef MBUF_STRESS_TEST 62378064Sume if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size) 62478064Sume m = m_fragment(m, M_DONTWAIT, mbuf_frag_size); 62578064Sume#endif 62678064Sume /* 62778064Sume * Reset layer specific mbuf flags 62878064Sume * to avoid confusing lower layers. 62978064Sume */ 63078064Sume m->m_flags &= ~(M_PROTOFLAGS); 63178064Sume error = (*ifp->if_output)(ifp, m, 63278064Sume (struct sockaddr *)dst, ro); 63378064Sume goto done; 63478064Sume } 63578064Sume 63678064Sume /* Balk when DF bit is set or the interface didn't support TSO. */ 63778064Sume if ((ip->ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) { 63878064Sume error = EMSGSIZE; 63978064Sume IPSTAT_INC(ips_cantfrag); 64078064Sume goto bad; 64178064Sume } 64278064Sume 64378064Sume /* 64478064Sume * Too large for interface; fragment if possible. If successful, 64578064Sume * on return, m will point to a list of packets to be sent. 64678064Sume */ 64778064Sume error = ip_fragment(ip, &m, mtu, ifp->if_hwassist, sw_csum); 64878064Sume if (error) 649 goto bad; 650 for (; m; m = m0) { 651 m0 = m->m_nextpkt; 652 m->m_nextpkt = 0; 653 if (error == 0) { 654 /* Record statistics for this interface address. */ 655 if (ia != NULL) { 656 ia->ia_ifa.if_opackets++; 657 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 658 } 659 /* 660 * Reset layer specific mbuf flags 661 * to avoid confusing upper layers. 662 */ 663 m->m_flags &= ~(M_PROTOFLAGS); 664 665 error = (*ifp->if_output)(ifp, m, 666 (struct sockaddr *)dst, ro); 667 } else 668 m_freem(m); 669 } 670 671 if (error == 0) 672 IPSTAT_INC(ips_fragmented); 673 674done: 675 if (ro == &iproute && ro->ro_rt && !nortfree) { 676 RTFREE(ro->ro_rt); 677 } 678 if (ia != NULL) 679 ifa_free(&ia->ia_ifa); 680 return (error); 681bad: 682 m_freem(m); 683 goto done; 684} 685 686/* 687 * Create a chain of fragments which fit the given mtu. m_frag points to the 688 * mbuf to be fragmented; on return it points to the chain with the fragments. 689 * Return 0 if no error. If error, m_frag may contain a partially built 690 * chain of fragments that should be freed by the caller. 691 * 692 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist) 693 * sw_csum contains the delayed checksums flags (e.g., CSUM_DELAY_IP). 694 */ 695int 696ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu, 697 u_long if_hwassist_flags, int sw_csum) 698{ 699 int error = 0; 700 int hlen = ip->ip_hl << 2; 701 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */ 702 int off; 703 struct mbuf *m0 = *m_frag; /* the original packet */ 704 int firstlen; 705 struct mbuf **mnext; 706 int nfrags; 707 708 if (ip->ip_off & IP_DF) { /* Fragmentation not allowed */ 709 IPSTAT_INC(ips_cantfrag); 710 return EMSGSIZE; 711 } 712 713 /* 714 * Must be able to put at least 8 bytes per fragment. 715 */ 716 if (len < 8) 717 return EMSGSIZE; 718 719 /* 720 * If the interface will not calculate checksums on 721 * fragmented packets, then do it here. 722 */ 723 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA && 724 (if_hwassist_flags & CSUM_IP_FRAGS) == 0) { 725 in_delayed_cksum(m0); 726 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 727 } 728#ifdef SCTP 729 if (m0->m_pkthdr.csum_flags & CSUM_SCTP && 730 (if_hwassist_flags & CSUM_IP_FRAGS) == 0) { 731 sctp_delayed_cksum(m0, hlen); 732 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP; 733 } 734#endif 735 if (len > PAGE_SIZE) { 736 /* 737 * Fragment large datagrams such that each segment 738 * contains a multiple of PAGE_SIZE amount of data, 739 * plus headers. This enables a receiver to perform 740 * page-flipping zero-copy optimizations. 741 * 742 * XXX When does this help given that sender and receiver 743 * could have different page sizes, and also mtu could 744 * be less than the receiver's page size ? 745 */ 746 int newlen; 747 struct mbuf *m; 748 749 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next) 750 off += m->m_len; 751 752 /* 753 * firstlen (off - hlen) must be aligned on an 754 * 8-byte boundary 755 */ 756 if (off < hlen) 757 goto smart_frag_failure; 758 off = ((off - hlen) & ~7) + hlen; 759 newlen = (~PAGE_MASK) & mtu; 760 if ((newlen + sizeof (struct ip)) > mtu) { 761 /* we failed, go back the default */ 762smart_frag_failure: 763 newlen = len; 764 off = hlen + len; 765 } 766 len = newlen; 767 768 } else { 769 off = hlen + len; 770 } 771 772 firstlen = off - hlen; 773 mnext = &m0->m_nextpkt; /* pointer to next packet */ 774 775 /* 776 * Loop through length of segment after first fragment, 777 * make new header and copy data of each part and link onto chain. 778 * Here, m0 is the original packet, m is the fragment being created. 779 * The fragments are linked off the m_nextpkt of the original 780 * packet, which after processing serves as the first fragment. 781 */ 782 for (nfrags = 1; off < ip->ip_len; off += len, nfrags++) { 783 struct ip *mhip; /* ip header on the fragment */ 784 struct mbuf *m; 785 int mhlen = sizeof (struct ip); 786 787 MGETHDR(m, M_DONTWAIT, MT_DATA); 788 if (m == NULL) { 789 error = ENOBUFS; 790 IPSTAT_INC(ips_odropped); 791 goto done; 792 } 793 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG; 794 /* 795 * In the first mbuf, leave room for the link header, then 796 * copy the original IP header including options. The payload 797 * goes into an additional mbuf chain returned by m_copym(). 798 */ 799 m->m_data += max_linkhdr; 800 mhip = mtod(m, struct ip *); 801 *mhip = *ip; 802 if (hlen > sizeof (struct ip)) { 803 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 804 mhip->ip_v = IPVERSION; 805 mhip->ip_hl = mhlen >> 2; 806 } 807 m->m_len = mhlen; 808 /* XXX do we need to add ip->ip_off below ? */ 809 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off; 810 if (off + len >= ip->ip_len) { /* last fragment */ 811 len = ip->ip_len - off; 812 m->m_flags |= M_LASTFRAG; 813 } else 814 mhip->ip_off |= IP_MF; 815 mhip->ip_len = htons((u_short)(len + mhlen)); 816 m->m_next = m_copym(m0, off, len, M_DONTWAIT); 817 if (m->m_next == NULL) { /* copy failed */ 818 m_free(m); 819 error = ENOBUFS; /* ??? */ 820 IPSTAT_INC(ips_odropped); 821 goto done; 822 } 823 m->m_pkthdr.len = mhlen + len; 824 m->m_pkthdr.rcvif = NULL; 825#ifdef MAC 826 mac_netinet_fragment(m0, m); 827#endif 828 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags; 829 mhip->ip_off = htons(mhip->ip_off); 830 mhip->ip_sum = 0; 831 if (sw_csum & CSUM_DELAY_IP) 832 mhip->ip_sum = in_cksum(m, mhlen); 833 *mnext = m; 834 mnext = &m->m_nextpkt; 835 } 836 IPSTAT_ADD(ips_ofragments, nfrags); 837 838 /* set first marker for fragment chain */ 839 m0->m_flags |= M_FIRSTFRAG | M_FRAG; 840 m0->m_pkthdr.csum_data = nfrags; 841 842 /* 843 * Update first fragment by trimming what's been copied out 844 * and updating header. 845 */ 846 m_adj(m0, hlen + firstlen - ip->ip_len); 847 m0->m_pkthdr.len = hlen + firstlen; 848 ip->ip_len = htons((u_short)m0->m_pkthdr.len); 849 ip->ip_off |= IP_MF; 850 ip->ip_off = htons(ip->ip_off); 851 ip->ip_sum = 0; 852 if (sw_csum & CSUM_DELAY_IP) 853 ip->ip_sum = in_cksum(m0, hlen); 854 855done: 856 *m_frag = m0; 857 return error; 858} 859 860void 861in_delayed_cksum(struct mbuf *m) 862{ 863 struct ip *ip; 864 u_short csum, offset; 865 866 ip = mtod(m, struct ip *); 867 offset = ip->ip_hl << 2 ; 868 csum = in_cksum_skip(m, ip->ip_len, offset); 869 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0) 870 csum = 0xffff; 871 offset += m->m_pkthdr.csum_data; /* checksum offset */ 872 873 if (offset + sizeof(u_short) > m->m_len) { 874 printf("delayed m_pullup, m->len: %d off: %d p: %d\n", 875 m->m_len, offset, ip->ip_p); 876 /* 877 * XXX 878 * this shouldn't happen, but if it does, the 879 * correct behavior may be to insert the checksum 880 * in the appropriate next mbuf in the chain. 881 */ 882 return; 883 } 884 *(u_short *)(m->m_data + offset) = csum; 885} 886 887/* 888 * IP socket option processing. 889 */ 890int 891ip_ctloutput(struct socket *so, struct sockopt *sopt) 892{ 893 struct inpcb *inp = sotoinpcb(so); 894 int error, optval; 895 896 error = optval = 0; 897 if (sopt->sopt_level != IPPROTO_IP) { 898 error = EINVAL; 899 900 if (sopt->sopt_level == SOL_SOCKET && 901 sopt->sopt_dir == SOPT_SET) { 902 switch (sopt->sopt_name) { 903 case SO_REUSEADDR: 904 INP_WLOCK(inp); 905 if (IN_MULTICAST(ntohl(inp->inp_laddr.s_addr))) { 906 if ((so->so_options & 907 (SO_REUSEADDR | SO_REUSEPORT)) != 0) 908 inp->inp_flags2 |= INP_REUSEPORT; 909 else 910 inp->inp_flags2 &= ~INP_REUSEPORT; 911 } 912 INP_WUNLOCK(inp); 913 error = 0; 914 break; 915 case SO_REUSEPORT: 916 INP_WLOCK(inp); 917 if ((so->so_options & SO_REUSEPORT) != 0) 918 inp->inp_flags2 |= INP_REUSEPORT; 919 else 920 inp->inp_flags2 &= ~INP_REUSEPORT; 921 INP_WUNLOCK(inp); 922 error = 0; 923 break; 924 case SO_SETFIB: 925 INP_WLOCK(inp); 926 inp->inp_inc.inc_fibnum = so->so_fibnum; 927 INP_WUNLOCK(inp); 928 error = 0; 929 break; 930 default: 931 break; 932 } 933 } 934 return (error); 935 } 936 937 switch (sopt->sopt_dir) { 938 case SOPT_SET: 939 switch (sopt->sopt_name) { 940 case IP_OPTIONS: 941#ifdef notyet 942 case IP_RETOPTS: 943#endif 944 { 945 struct mbuf *m; 946 if (sopt->sopt_valsize > MLEN) { 947 error = EMSGSIZE; 948 break; 949 } 950 MGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT, MT_DATA); 951 if (m == NULL) { 952 error = ENOBUFS; 953 break; 954 } 955 m->m_len = sopt->sopt_valsize; 956 error = sooptcopyin(sopt, mtod(m, char *), m->m_len, 957 m->m_len); 958 if (error) { 959 m_free(m); 960 break; 961 } 962 INP_WLOCK(inp); 963 error = ip_pcbopts(inp, sopt->sopt_name, m); 964 INP_WUNLOCK(inp); 965 return (error); 966 } 967 968 case IP_BINDANY: 969 if (sopt->sopt_td != NULL) { 970 error = priv_check(sopt->sopt_td, 971 PRIV_NETINET_BINDANY); 972 if (error) 973 break; 974 } 975 /* FALLTHROUGH */ 976 case IP_TOS: 977 case IP_TTL: 978 case IP_MINTTL: 979 case IP_RECVOPTS: 980 case IP_RECVRETOPTS: 981 case IP_RECVDSTADDR: 982 case IP_RECVTTL: 983 case IP_RECVIF: 984 case IP_FAITH: 985 case IP_ONESBCAST: 986 case IP_DONTFRAG: 987 case IP_RECVTOS: 988 error = sooptcopyin(sopt, &optval, sizeof optval, 989 sizeof optval); 990 if (error) 991 break; 992 993 switch (sopt->sopt_name) { 994 case IP_TOS: 995 inp->inp_ip_tos = optval; 996 break; 997 998 case IP_TTL: 999 inp->inp_ip_ttl = optval; 1000 break; 1001 1002 case IP_MINTTL: 1003 if (optval >= 0 && optval <= MAXTTL) 1004 inp->inp_ip_minttl = optval; 1005 else 1006 error = EINVAL; 1007 break; 1008 1009#define OPTSET(bit) do { \ 1010 INP_WLOCK(inp); \ 1011 if (optval) \ 1012 inp->inp_flags |= bit; \ 1013 else \ 1014 inp->inp_flags &= ~bit; \ 1015 INP_WUNLOCK(inp); \ 1016} while (0) 1017 1018 case IP_RECVOPTS: 1019 OPTSET(INP_RECVOPTS); 1020 break; 1021 1022 case IP_RECVRETOPTS: 1023 OPTSET(INP_RECVRETOPTS); 1024 break; 1025 1026 case IP_RECVDSTADDR: 1027 OPTSET(INP_RECVDSTADDR); 1028 break; 1029 1030 case IP_RECVTTL: 1031 OPTSET(INP_RECVTTL); 1032 break; 1033 1034 case IP_RECVIF: 1035 OPTSET(INP_RECVIF); 1036 break; 1037 1038 case IP_FAITH: 1039 OPTSET(INP_FAITH); 1040 break; 1041 1042 case IP_ONESBCAST: 1043 OPTSET(INP_ONESBCAST); 1044 break; 1045 case IP_DONTFRAG: 1046 OPTSET(INP_DONTFRAG); 1047 break; 1048 case IP_BINDANY: 1049 OPTSET(INP_BINDANY); 1050 break; 1051 case IP_RECVTOS: 1052 OPTSET(INP_RECVTOS); 1053 break; 1054 } 1055 break; 1056#undef OPTSET 1057 1058 /* 1059 * Multicast socket options are processed by the in_mcast 1060 * module. 1061 */ 1062 case IP_MULTICAST_IF: 1063 case IP_MULTICAST_VIF: 1064 case IP_MULTICAST_TTL: 1065 case IP_MULTICAST_LOOP: 1066 case IP_ADD_MEMBERSHIP: 1067 case IP_DROP_MEMBERSHIP: 1068 case IP_ADD_SOURCE_MEMBERSHIP: 1069 case IP_DROP_SOURCE_MEMBERSHIP: 1070 case IP_BLOCK_SOURCE: 1071 case IP_UNBLOCK_SOURCE: 1072 case IP_MSFILTER: 1073 case MCAST_JOIN_GROUP: 1074 case MCAST_LEAVE_GROUP: 1075 case MCAST_JOIN_SOURCE_GROUP: 1076 case MCAST_LEAVE_SOURCE_GROUP: 1077 case MCAST_BLOCK_SOURCE: 1078 case MCAST_UNBLOCK_SOURCE: 1079 error = inp_setmoptions(inp, sopt); 1080 break; 1081 1082 case IP_PORTRANGE: 1083 error = sooptcopyin(sopt, &optval, sizeof optval, 1084 sizeof optval); 1085 if (error) 1086 break; 1087 1088 INP_WLOCK(inp); 1089 switch (optval) { 1090 case IP_PORTRANGE_DEFAULT: 1091 inp->inp_flags &= ~(INP_LOWPORT); 1092 inp->inp_flags &= ~(INP_HIGHPORT); 1093 break; 1094 1095 case IP_PORTRANGE_HIGH: 1096 inp->inp_flags &= ~(INP_LOWPORT); 1097 inp->inp_flags |= INP_HIGHPORT; 1098 break; 1099 1100 case IP_PORTRANGE_LOW: 1101 inp->inp_flags &= ~(INP_HIGHPORT); 1102 inp->inp_flags |= INP_LOWPORT; 1103 break; 1104 1105 default: 1106 error = EINVAL; 1107 break; 1108 } 1109 INP_WUNLOCK(inp); 1110 break; 1111 1112#ifdef IPSEC 1113 case IP_IPSEC_POLICY: 1114 { 1115 caddr_t req; 1116 struct mbuf *m; 1117 1118 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */ 1119 break; 1120 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */ 1121 break; 1122 req = mtod(m, caddr_t); 1123 error = ipsec_set_policy(inp, sopt->sopt_name, req, 1124 m->m_len, (sopt->sopt_td != NULL) ? 1125 sopt->sopt_td->td_ucred : NULL); 1126 m_freem(m); 1127 break; 1128 } 1129#endif /* IPSEC */ 1130 1131 default: 1132 error = ENOPROTOOPT; 1133 break; 1134 } 1135 break; 1136 1137 case SOPT_GET: 1138 switch (sopt->sopt_name) { 1139 case IP_OPTIONS: 1140 case IP_RETOPTS: 1141 if (inp->inp_options) 1142 error = sooptcopyout(sopt, 1143 mtod(inp->inp_options, 1144 char *), 1145 inp->inp_options->m_len); 1146 else 1147 sopt->sopt_valsize = 0; 1148 break; 1149 1150 case IP_TOS: 1151 case IP_TTL: 1152 case IP_MINTTL: 1153 case IP_RECVOPTS: 1154 case IP_RECVRETOPTS: 1155 case IP_RECVDSTADDR: 1156 case IP_RECVTTL: 1157 case IP_RECVIF: 1158 case IP_PORTRANGE: 1159 case IP_FAITH: 1160 case IP_ONESBCAST: 1161 case IP_DONTFRAG: 1162 case IP_BINDANY: 1163 case IP_RECVTOS: 1164 switch (sopt->sopt_name) { 1165 1166 case IP_TOS: 1167 optval = inp->inp_ip_tos; 1168 break; 1169 1170 case IP_TTL: 1171 optval = inp->inp_ip_ttl; 1172 break; 1173 1174 case IP_MINTTL: 1175 optval = inp->inp_ip_minttl; 1176 break; 1177 1178#define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1179 1180 case IP_RECVOPTS: 1181 optval = OPTBIT(INP_RECVOPTS); 1182 break; 1183 1184 case IP_RECVRETOPTS: 1185 optval = OPTBIT(INP_RECVRETOPTS); 1186 break; 1187 1188 case IP_RECVDSTADDR: 1189 optval = OPTBIT(INP_RECVDSTADDR); 1190 break; 1191 1192 case IP_RECVTTL: 1193 optval = OPTBIT(INP_RECVTTL); 1194 break; 1195 1196 case IP_RECVIF: 1197 optval = OPTBIT(INP_RECVIF); 1198 break; 1199 1200 case IP_PORTRANGE: 1201 if (inp->inp_flags & INP_HIGHPORT) 1202 optval = IP_PORTRANGE_HIGH; 1203 else if (inp->inp_flags & INP_LOWPORT) 1204 optval = IP_PORTRANGE_LOW; 1205 else 1206 optval = 0; 1207 break; 1208 1209 case IP_FAITH: 1210 optval = OPTBIT(INP_FAITH); 1211 break; 1212 1213 case IP_ONESBCAST: 1214 optval = OPTBIT(INP_ONESBCAST); 1215 break; 1216 case IP_DONTFRAG: 1217 optval = OPTBIT(INP_DONTFRAG); 1218 break; 1219 case IP_BINDANY: 1220 optval = OPTBIT(INP_BINDANY); 1221 break; 1222 case IP_RECVTOS: 1223 optval = OPTBIT(INP_RECVTOS); 1224 break; 1225 } 1226 error = sooptcopyout(sopt, &optval, sizeof optval); 1227 break; 1228 1229 /* 1230 * Multicast socket options are processed by the in_mcast 1231 * module. 1232 */ 1233 case IP_MULTICAST_IF: 1234 case IP_MULTICAST_VIF: 1235 case IP_MULTICAST_TTL: 1236 case IP_MULTICAST_LOOP: 1237 case IP_MSFILTER: 1238 error = inp_getmoptions(inp, sopt); 1239 break; 1240 1241#ifdef IPSEC 1242 case IP_IPSEC_POLICY: 1243 { 1244 struct mbuf *m = NULL; 1245 caddr_t req = NULL; 1246 size_t len = 0; 1247 1248 if (m != 0) { 1249 req = mtod(m, caddr_t); 1250 len = m->m_len; 1251 } 1252 error = ipsec_get_policy(sotoinpcb(so), req, len, &m); 1253 if (error == 0) 1254 error = soopt_mcopyout(sopt, m); /* XXX */ 1255 if (error == 0) 1256 m_freem(m); 1257 break; 1258 } 1259#endif /* IPSEC */ 1260 1261 default: 1262 error = ENOPROTOOPT; 1263 break; 1264 } 1265 break; 1266 } 1267 return (error); 1268} 1269 1270/* 1271 * Routine called from ip_output() to loop back a copy of an IP multicast 1272 * packet to the input queue of a specified interface. Note that this 1273 * calls the output routine of the loopback "driver", but with an interface 1274 * pointer that might NOT be a loopback interface -- evil, but easier than 1275 * replicating that code here. 1276 */ 1277static void 1278ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst, 1279 int hlen) 1280{ 1281 register struct ip *ip; 1282 struct mbuf *copym; 1283 1284 /* 1285 * Make a deep copy of the packet because we're going to 1286 * modify the pack in order to generate checksums. 1287 */ 1288 copym = m_dup(m, M_DONTWAIT); 1289 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen)) 1290 copym = m_pullup(copym, hlen); 1291 if (copym != NULL) { 1292 /* If needed, compute the checksum and mark it as valid. */ 1293 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 1294 in_delayed_cksum(copym); 1295 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 1296 copym->m_pkthdr.csum_flags |= 1297 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 1298 copym->m_pkthdr.csum_data = 0xffff; 1299 } 1300 /* 1301 * We don't bother to fragment if the IP length is greater 1302 * than the interface's MTU. Can this possibly matter? 1303 */ 1304 ip = mtod(copym, struct ip *); 1305 ip->ip_len = htons(ip->ip_len); 1306 ip->ip_off = htons(ip->ip_off); 1307 ip->ip_sum = 0; 1308 ip->ip_sum = in_cksum(copym, hlen); 1309#if 1 /* XXX */ 1310 if (dst->sin_family != AF_INET) { 1311 printf("ip_mloopback: bad address family %d\n", 1312 dst->sin_family); 1313 dst->sin_family = AF_INET; 1314 } 1315#endif 1316 if_simloop(ifp, copym, dst->sin_family, 0); 1317 } 1318} 1319