Index: projects/routing/sys/net/if_ethersubr.c =================================================================== --- projects/routing/sys/net/if_ethersubr.c (revision 287053) +++ projects/routing/sys/net/if_ethersubr.c (revision 287054) @@ -1,1268 +1,1268 @@ /*- * Copyright (c) 1982, 1989, 1993 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)if_ethersubr.c 8.1 (Berkeley) 6/10/93 * $FreeBSD$ */ #include "opt_inet.h" #include "opt_inet6.h" #include "opt_netgraph.h" #include "opt_mbuf_profiling.h" #include "opt_rss.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #if defined(INET) || defined(INET6) #include #include #include #include #include #endif #ifdef INET6 #include #endif #include #include #ifdef CTASSERT CTASSERT(sizeof (struct ether_header) == ETHER_ADDR_LEN * 2 + 2); CTASSERT(sizeof (struct ether_addr) == ETHER_ADDR_LEN); #endif VNET_DEFINE(struct pfil_head, link_pfil_hook); /* Packet filter hooks */ /* netgraph node hooks for ng_ether(4) */ void (*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp); void (*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m); int (*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp); void (*ng_ether_attach_p)(struct ifnet *ifp); void (*ng_ether_detach_p)(struct ifnet *ifp); void (*vlan_input_p)(struct ifnet *, struct mbuf *); /* if_bridge(4) support */ struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *); int (*bridge_output_p)(struct ifnet *, struct mbuf *, struct sockaddr *, struct rtentry *); void (*bridge_dn_p)(struct mbuf *, struct ifnet *); /* if_lagg(4) support */ struct mbuf *(*lagg_input_p)(struct ifnet *, struct mbuf *); static const u_char etherbroadcastaddr[ETHER_ADDR_LEN] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; static int ether_resolvemulti(struct ifnet *, struct sockaddr **, struct sockaddr *); #ifdef VIMAGE static void ether_reassign(struct ifnet *, struct vnet *, char *); #endif int ether_output_full(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst, struct route *ro); -int ether_output2(struct ifnet *ifp, struct mbuf *m, struct nhop_data *nh, +int ether_output2(struct ifnet *ifp, struct mbuf *m, struct nhop_prepend *nh, int af); static int loopback_frame(struct ifnet *ifp, struct mbuf *m, int family, int hlen); #define ETHER_IS_BROADCAST(addr) \ (bcmp(etherbroadcastaddr, (addr), ETHER_ADDR_LEN) == 0) #define senderr(e) do { error = (e); goto bad;} while (0) static void update_mbuf_csumflags(struct mbuf *src, struct mbuf *dst) { int csum_flags = 0; if (src->m_pkthdr.csum_flags & CSUM_IP) csum_flags |= (CSUM_IP_CHECKED|CSUM_IP_VALID); if (src->m_pkthdr.csum_flags & CSUM_DELAY_DATA) csum_flags |= (CSUM_DATA_VALID|CSUM_PSEUDO_HDR); if (src->m_pkthdr.csum_flags & CSUM_SCTP) csum_flags |= CSUM_SCTP_VALID; dst->m_pkthdr.csum_flags |= csum_flags; if (csum_flags & CSUM_DATA_VALID) dst->m_pkthdr.csum_data = 0xffff; } int ether_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst, struct route *ro) { if (ro != NULL && (ro->ro_flags & RT_NHOP)) - return (ether_output2(ifp, m, (struct nhop_data *)ro->ro_lle, + return (ether_output2(ifp, m, (struct nhop_prepend *)ro->ro_lle, (ro->ro_flags >> 8) & 0xFF)); return (ether_output_full(ifp, m, dst, ro)); } /* * Ethernet output routine. * Encapsulate a packet of type family for the local net. * Use trailer local net encapsulation if enough data in first * packet leaves a multiple of 512 bytes of data in remainder. */ int ether_output_full(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst, struct route *ro) { short type; int error = 0, hdrcmplt = 0; u_char edst[ETHER_ADDR_LEN]; struct llentry *lle = NULL; struct rtentry *rt0 = NULL; struct ether_header *eh; struct pf_mtag *t; int loop_copy = 1; int hlen; /* link layer header length */ int is_gw = 0; uint32_t pflags = 0; if (ro != NULL) { if (!(m->m_flags & (M_BCAST | M_MCAST))) { lle = ro->ro_lle; if (lle != NULL) pflags = lle->la_flags; } rt0 = ro->ro_rt; if (rt0 != NULL && (rt0->rt_flags & RTF_GATEWAY) != 0) is_gw = 1; } #ifdef MAC error = mac_ifnet_check_transmit(ifp, m); if (error) senderr(error); #endif M_PROFILE(m); if (ifp->if_flags & IFF_MONITOR) senderr(ENETDOWN); if (!((ifp->if_flags & IFF_UP) && (ifp->if_drv_flags & IFF_DRV_RUNNING))) senderr(ENETDOWN); hlen = ETHER_HDR_LEN; switch (dst->sa_family) { #ifdef INET case AF_INET: if (lle != NULL && (pflags & LLE_VALID) != 0) memcpy(edst, &lle->ll_addr.mac16, sizeof(edst)); else error = arpresolve(ifp, is_gw, m, dst, edst, &pflags); if (error) return (error == EWOULDBLOCK ? 0 : error); type = htons(ETHERTYPE_IP); break; case AF_ARP: { struct arphdr *ah; ah = mtod(m, struct arphdr *); ah->ar_hrd = htons(ARPHRD_ETHER); loop_copy = 0; /* if this is for us, don't do it */ switch(ntohs(ah->ar_op)) { case ARPOP_REVREQUEST: case ARPOP_REVREPLY: type = htons(ETHERTYPE_REVARP); break; case ARPOP_REQUEST: case ARPOP_REPLY: default: type = htons(ETHERTYPE_ARP); break; } if (m->m_flags & M_BCAST) bcopy(ifp->if_broadcastaddr, edst, ETHER_ADDR_LEN); else bcopy(ar_tha(ah), edst, ETHER_ADDR_LEN); } break; #endif #ifdef INET6 case AF_INET6: if (lle != NULL && (pflags & LLE_VALID)) memcpy(edst, &lle->ll_addr.mac16, sizeof(edst)); else error = nd6_storelladdr(ifp, m, dst, (u_char *)edst, &pflags); if (error) return error; type = htons(ETHERTYPE_IPV6); break; #endif case pseudo_AF_HDRCMPLT: { const struct ether_header *eh; hdrcmplt = 1; /* FALLTHROUGH */ case AF_UNSPEC: loop_copy = 0; /* if this is for us, don't do it */ eh = (const struct ether_header *)dst->sa_data; (void)memcpy(edst, eh->ether_dhost, sizeof (edst)); type = eh->ether_type; break; } default: if_printf(ifp, "can't handle af%d\n", dst->sa_family); senderr(EAFNOSUPPORT); } if ((pflags & LLE_IFADDR) != 0) { update_mbuf_csumflags(m, m); return (if_simloop(ifp, m, dst->sa_family, 0)); } /* * Add local net header. If no space in first mbuf, * allocate another. */ M_PREPEND(m, ETHER_HDR_LEN, M_NOWAIT); if (m == NULL) senderr(ENOBUFS); eh = mtod(m, struct ether_header *); if (hdrcmplt == 0) { memcpy(&eh->ether_type, &type, sizeof(eh->ether_type)); memcpy(eh->ether_dhost, edst, sizeof (edst)); memcpy(eh->ether_shost, IF_LLADDR(ifp),sizeof(eh->ether_shost)); } /* * If a simplex interface, and the packet is being sent to our * Ethernet address or a broadcast address, loopback a copy. * XXX To make a simplex device behave exactly like a duplex * device, we should copy in the case of sending to our own * ethernet address (thus letting the original actually appear * on the wire). However, we don't do that here for security * reasons and compatibility with the original behavior. */ if ((ifp->if_flags & IFF_SIMPLEX) && loop_copy && ((t = pf_find_mtag(m)) == NULL || !t->routed)) { if ((m->m_flags & M_BCAST) || (bcmp(eh->ether_dhost, eh->ether_shost, ETHER_ADDR_LEN) == 0)) { /* Either broadcast or to-us L2 header */ if (loopback_frame(ifp, m, dst->sa_family, hlen) == 1) return (0); } } /* * Bridges require special output handling. */ if (ifp->if_bridge) { BRIDGE_OUTPUT(ifp, m, error); return (error); } #if defined(INET) || defined(INET6) if (ifp->if_carp && (error = (*carp_output_p)(ifp, m, dst))) goto bad; #endif /* Handle ng_ether(4) processing, if any */ if (ifp->if_l2com != NULL) { KASSERT(ng_ether_output_p != NULL, ("ng_ether_output_p is NULL")); if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) { bad: if (m != NULL) m_freem(m); return (error); } if (m == NULL) return (0); } /* Continue with link-layer output */ return ether_output_frame(ifp, m); } /* * We assume this function to be called for * ip[6]_output(), with already pre-compiled L2 header. * * Function assumes all loopback routing is already done on L3, * so the only reason to push packet (copy) to host is M_BCAST flag. */ int -ether_output2(struct ifnet *ifp, struct mbuf *m, struct nhop_data *nh, int af) +ether_output2(struct ifnet *ifp, struct mbuf *m, struct nhop_prepend *nh,int af) { int error; #ifdef MAC error = mac_ifnet_check_transmit(ifp, m); if (error) senderr(error); #endif M_PROFILE(m); if (ifp->if_flags & IFF_MONITOR) senderr(ENETDOWN); if (!((ifp->if_flags & IFF_UP) && (ifp->if_drv_flags & IFF_DRV_RUNNING))) senderr(ENETDOWN); if ((ifp->if_flags & IFF_SIMPLEX) && (m->m_flags & M_BCAST)) { /* We have to copy frame to-us */ if (loopback_frame(NH_LIFP(nh), m, af, nh->nh_count) != 0) return (0); } /* * Bridges require special output handling. */ if (ifp->if_bridge) { BRIDGE_OUTPUT(ifp, m, error); return (error); } #if defined(INET) || defined(INET6) if (ifp->if_carp) { struct sockaddr_in dst; memset(&dst, 0, sizeof(dst)); //dst.sin_addr = error = (*carp_output_p)(ifp, m, (const struct sockaddr *)&dst); if (error != 0) goto bad; } #endif /* Handle ng_ether(4) processing, if any */ if (ifp->if_l2com != NULL) { KASSERT(ng_ether_output_p != NULL, ("ng_ether_output_p is NULL")); if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) { bad: if (m != NULL) m_freem(m); return (error); } if (m == NULL) return (0); } /* Continue with link-layer output */ return (ether_output_frame(ifp, m)); } static int loopback_frame(struct ifnet *ifp, struct mbuf *m, int family, int hlen) { struct ether_header *eh; if (m->m_flags & M_BCAST) { struct mbuf *n; /* * Because if_simloop() modifies the packet, we need a * writable copy through m_dup() instead of a readonly * one as m_copy[m] would give us. The alternative would * be to modify if_simloop() to handle the readonly mbuf, * but performancewise it is mostly equivalent (trading * extra data copying vs. extra locking). * * XXX This is a local workaround. A number of less * often used kernel parts suffer from the same bug. * See PR kern/105943 for a proposed general solution. */ if ((n = m_dup(m, M_NOWAIT)) != NULL) { update_mbuf_csumflags(m, n); if_simloop(ifp, n, family, hlen); } else if_inc_counter(ifp, IFCOUNTER_IQDROPS, 1); } else { eh = mtod(m, struct ether_header *); if (bcmp(eh->ether_dhost, eh->ether_shost, ETHER_ADDR_LEN) == 0) { update_mbuf_csumflags(m, m); if_simloop(ifp, m, family, hlen); return (1); } } return (0); } /* * Ethernet link layer output routine to send a raw frame to the device. * * This assumes that the 14 byte Ethernet header is present and contiguous * in the first mbuf (if BRIDGE'ing). */ int ether_output_frame(struct ifnet *ifp, struct mbuf *m) { int i; if (PFIL_HOOKED(&V_link_pfil_hook)) { i = pfil_run_hooks(&V_link_pfil_hook, &m, ifp, PFIL_OUT, NULL); if (i != 0) return (EACCES); if (m == NULL) return (0); } /* * Queue message on interface, update output statistics if * successful, and start output if interface not yet active. */ return ((ifp->if_transmit)(ifp, m)); } #if defined(INET) || defined(INET6) #endif /* * Process a received Ethernet packet; the packet is in the * mbuf chain m with the ethernet header at the front. */ static void ether_input_internal(struct ifnet *ifp, struct mbuf *m) { struct ether_header *eh; u_short etype; if ((ifp->if_flags & IFF_UP) == 0) { m_freem(m); return; } #ifdef DIAGNOSTIC if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) { if_printf(ifp, "discard frame at !IFF_DRV_RUNNING\n"); m_freem(m); return; } #endif /* * Do consistency checks to verify assumptions * made by code past this point. */ if ((m->m_flags & M_PKTHDR) == 0) { if_printf(ifp, "discard frame w/o packet header\n"); if_inc_counter(ifp, IFCOUNTER_IERRORS, 1); m_freem(m); return; } if (m->m_len < ETHER_HDR_LEN) { /* XXX maybe should pullup? */ if_printf(ifp, "discard frame w/o leading ethernet " "header (len %u pkt len %u)\n", m->m_len, m->m_pkthdr.len); if_inc_counter(ifp, IFCOUNTER_IERRORS, 1); m_freem(m); return; } eh = mtod(m, struct ether_header *); etype = ntohs(eh->ether_type); if (m->m_pkthdr.rcvif == NULL) { if_printf(ifp, "discard frame w/o interface pointer\n"); if_inc_counter(ifp, IFCOUNTER_IERRORS, 1); m_freem(m); return; } #ifdef DIAGNOSTIC if (m->m_pkthdr.rcvif != ifp) { if_printf(ifp, "Warning, frame marked as received on %s\n", m->m_pkthdr.rcvif->if_xname); } #endif random_harvest_queue(m, sizeof(*m), 2, RANDOM_NET_ETHER); CURVNET_SET_QUIET(ifp->if_vnet); if (ETHER_IS_MULTICAST(eh->ether_dhost)) { if (ETHER_IS_BROADCAST(eh->ether_dhost)) m->m_flags |= M_BCAST; else m->m_flags |= M_MCAST; if_inc_counter(ifp, IFCOUNTER_IMCASTS, 1); } #ifdef MAC /* * Tag the mbuf with an appropriate MAC label before any other * consumers can get to it. */ mac_ifnet_create_mbuf(ifp, m); #endif /* * Give bpf a chance at the packet. */ ETHER_BPF_MTAP(ifp, m); /* * If the CRC is still on the packet, trim it off. We do this once * and once only in case we are re-entered. Nothing else on the * Ethernet receive path expects to see the FCS. */ if (m->m_flags & M_HASFCS) { m_adj(m, -ETHER_CRC_LEN); m->m_flags &= ~M_HASFCS; } if (!(ifp->if_capenable & IFCAP_HWSTATS)) if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len); /* Allow monitor mode to claim this frame, after stats are updated. */ if (ifp->if_flags & IFF_MONITOR) { m_freem(m); CURVNET_RESTORE(); return; } /* Handle input from a lagg(4) port */ if (ifp->if_type == IFT_IEEE8023ADLAG) { KASSERT(lagg_input_p != NULL, ("%s: if_lagg not loaded!", __func__)); m = (*lagg_input_p)(ifp, m); if (m != NULL) ifp = m->m_pkthdr.rcvif; else { CURVNET_RESTORE(); return; } } /* * If the hardware did not process an 802.1Q tag, do this now, * to allow 802.1P priority frames to be passed to the main input * path correctly. * TODO: Deal with Q-in-Q frames, but not arbitrary nesting levels. */ if ((m->m_flags & M_VLANTAG) == 0 && etype == ETHERTYPE_VLAN) { struct ether_vlan_header *evl; if (m->m_len < sizeof(*evl) && (m = m_pullup(m, sizeof(*evl))) == NULL) { #ifdef DIAGNOSTIC if_printf(ifp, "cannot pullup VLAN header\n"); #endif if_inc_counter(ifp, IFCOUNTER_IERRORS, 1); m_freem(m); CURVNET_RESTORE(); return; } evl = mtod(m, struct ether_vlan_header *); m->m_pkthdr.ether_vtag = ntohs(evl->evl_tag); m->m_flags |= M_VLANTAG; bcopy((char *)evl, (char *)evl + ETHER_VLAN_ENCAP_LEN, ETHER_HDR_LEN - ETHER_TYPE_LEN); m_adj(m, ETHER_VLAN_ENCAP_LEN); eh = mtod(m, struct ether_header *); } M_SETFIB(m, ifp->if_fib); /* Allow ng_ether(4) to claim this frame. */ if (ifp->if_l2com != NULL) { KASSERT(ng_ether_input_p != NULL, ("%s: ng_ether_input_p is NULL", __func__)); m->m_flags &= ~M_PROMISC; (*ng_ether_input_p)(ifp, &m); if (m == NULL) { CURVNET_RESTORE(); return; } eh = mtod(m, struct ether_header *); } /* * Allow if_bridge(4) to claim this frame. * The BRIDGE_INPUT() macro will update ifp if the bridge changed it * and the frame should be delivered locally. */ if (ifp->if_bridge != NULL) { m->m_flags &= ~M_PROMISC; BRIDGE_INPUT(ifp, m); if (m == NULL) { CURVNET_RESTORE(); return; } eh = mtod(m, struct ether_header *); } #if defined(INET) || defined(INET6) /* * Clear M_PROMISC on frame so that carp(4) will see it when the * mbuf flows up to Layer 3. * FreeBSD's implementation of carp(4) uses the inprotosw * to dispatch IPPROTO_CARP. carp(4) also allocates its own * Ethernet addresses of the form 00:00:5e:00:01:xx, which * is outside the scope of the M_PROMISC test below. * TODO: Maintain a hash table of ethernet addresses other than * ether_dhost which may be active on this ifp. */ if (ifp->if_carp && (*carp_forus_p)(ifp, eh->ether_dhost)) { m->m_flags &= ~M_PROMISC; } else #endif { /* * If the frame received was not for our MAC address, set the * M_PROMISC flag on the mbuf chain. The frame may need to * be seen by the rest of the Ethernet input path in case of * re-entry (e.g. bridge, vlan, netgraph) but should not be * seen by upper protocol layers. */ if (!ETHER_IS_MULTICAST(eh->ether_dhost) && bcmp(IF_LLADDR(ifp), eh->ether_dhost, ETHER_ADDR_LEN) != 0) m->m_flags |= M_PROMISC; } ether_demux(ifp, m); CURVNET_RESTORE(); } /* * Ethernet input dispatch; by default, direct dispatch here regardless of * global configuration. However, if RSS is enabled, hook up RSS affinity * so that when deferred or hybrid dispatch is enabled, we can redistribute * load based on RSS. * * XXXRW: Would be nice if the ifnet passed up a flag indicating whether or * not it had already done work distribution via multi-queue. Then we could * direct dispatch in the event load balancing was already complete and * handle the case of interfaces with different capabilities better. * * XXXRW: Sort of want an M_DISTRIBUTED flag to avoid multiple distributions * at multiple layers? * * XXXRW: For now, enable all this only if RSS is compiled in, although it * works fine without RSS. Need to characterise the performance overhead * of the detour through the netisr code in the event the result is always * direct dispatch. */ static void ether_nh_input(struct mbuf *m) { ether_input_internal(m->m_pkthdr.rcvif, m); } static struct netisr_handler ether_nh = { .nh_name = "ether", .nh_handler = ether_nh_input, .nh_proto = NETISR_ETHER, #ifdef RSS .nh_policy = NETISR_POLICY_CPU, .nh_dispatch = NETISR_DISPATCH_DIRECT, .nh_m2cpuid = rss_m2cpuid, #else .nh_policy = NETISR_POLICY_SOURCE, .nh_dispatch = NETISR_DISPATCH_DIRECT, #endif }; static void ether_init(__unused void *arg) { netisr_register(ðer_nh); } SYSINIT(ether, SI_SUB_INIT_IF, SI_ORDER_ANY, ether_init, NULL); static void vnet_ether_init(__unused void *arg) { int i; /* Initialize packet filter hooks. */ V_link_pfil_hook.ph_type = PFIL_TYPE_AF; V_link_pfil_hook.ph_af = AF_LINK; if ((i = pfil_head_register(&V_link_pfil_hook)) != 0) printf("%s: WARNING: unable to register pfil link hook, " "error %d\n", __func__, i); } VNET_SYSINIT(vnet_ether_init, SI_SUB_PROTO_IF, SI_ORDER_ANY, vnet_ether_init, NULL); static void vnet_ether_destroy(__unused void *arg) { int i; if ((i = pfil_head_unregister(&V_link_pfil_hook)) != 0) printf("%s: WARNING: unable to unregister pfil link hook, " "error %d\n", __func__, i); } VNET_SYSUNINIT(vnet_ether_uninit, SI_SUB_PROTO_IF, SI_ORDER_ANY, vnet_ether_destroy, NULL); static void ether_input(struct ifnet *ifp, struct mbuf *m) { struct mbuf *mn; /* * The drivers are allowed to pass in a chain of packets linked with * m_nextpkt. We split them up into separate packets here and pass * them up. This allows the drivers to amortize the receive lock. */ while (m) { mn = m->m_nextpkt; m->m_nextpkt = NULL; /* * We will rely on rcvif being set properly in the deferred context, * so assert it is correct here. */ KASSERT(m->m_pkthdr.rcvif == ifp, ("%s: ifnet mismatch", __func__)); netisr_dispatch(NETISR_ETHER, m); m = mn; } } /* * Upper layer processing for a received Ethernet packet. */ void ether_demux(struct ifnet *ifp, struct mbuf *m) { struct ether_header *eh; int i, isr; u_short ether_type; KASSERT(ifp != NULL, ("%s: NULL interface pointer", __func__)); /* Do not grab PROMISC frames in case we are re-entered. */ if (PFIL_HOOKED(&V_link_pfil_hook) && !(m->m_flags & M_PROMISC)) { i = pfil_run_hooks(&V_link_pfil_hook, &m, ifp, PFIL_IN, NULL); if (i != 0 || m == NULL) return; } eh = mtod(m, struct ether_header *); ether_type = ntohs(eh->ether_type); /* * If this frame has a VLAN tag other than 0, call vlan_input() * if its module is loaded. Otherwise, drop. */ if ((m->m_flags & M_VLANTAG) && EVL_VLANOFTAG(m->m_pkthdr.ether_vtag) != 0) { if (ifp->if_vlantrunk == NULL) { if_inc_counter(ifp, IFCOUNTER_NOPROTO, 1); m_freem(m); return; } KASSERT(vlan_input_p != NULL,("%s: VLAN not loaded!", __func__)); /* Clear before possibly re-entering ether_input(). */ m->m_flags &= ~M_PROMISC; (*vlan_input_p)(ifp, m); return; } /* * Pass promiscuously received frames to the upper layer if the user * requested this by setting IFF_PPROMISC. Otherwise, drop them. */ if ((ifp->if_flags & IFF_PPROMISC) == 0 && (m->m_flags & M_PROMISC)) { m_freem(m); return; } /* * Reset layer specific mbuf flags to avoid confusing upper layers. * Strip off Ethernet header. */ m->m_flags &= ~M_VLANTAG; m_clrprotoflags(m); m_adj(m, ETHER_HDR_LEN); /* * Dispatch frame to upper layer. */ switch (ether_type) { #ifdef INET case ETHERTYPE_IP: if ((m = ip_fastforward(m)) == NULL) return; isr = NETISR_IP; break; case ETHERTYPE_ARP: if (ifp->if_flags & IFF_NOARP) { /* Discard packet if ARP is disabled on interface */ m_freem(m); return; } isr = NETISR_ARP; break; #endif #ifdef INET6 case ETHERTYPE_IPV6: isr = NETISR_IPV6; break; #endif default: goto discard; } netisr_dispatch(isr, m); return; discard: /* * Packet is to be discarded. If netgraph is present, * hand the packet to it for last chance processing; * otherwise dispose of it. */ if (ifp->if_l2com != NULL) { KASSERT(ng_ether_input_orphan_p != NULL, ("ng_ether_input_orphan_p is NULL")); /* * Put back the ethernet header so netgraph has a * consistent view of inbound packets. */ M_PREPEND(m, ETHER_HDR_LEN, M_NOWAIT); (*ng_ether_input_orphan_p)(ifp, m); return; } m_freem(m); } /* * Convert Ethernet address to printable (loggable) representation. * This routine is for compatibility; it's better to just use * * printf("%6D", , ":"); * * since there's no static buffer involved. */ char * ether_sprintf(const u_char *ap) { static char etherbuf[18]; snprintf(etherbuf, sizeof (etherbuf), "%6D", ap, ":"); return (etherbuf); } /* * Perform common duties while attaching to interface list */ void ether_ifattach(struct ifnet *ifp, const u_int8_t *lla) { int i; struct ifaddr *ifa; struct sockaddr_dl *sdl; ifp->if_addrlen = ETHER_ADDR_LEN; ifp->if_hdrlen = ETHER_HDR_LEN; if_attach(ifp); ifp->if_mtu = ETHERMTU; ifp->if_output = ether_output; ifp->if_input = ether_input; ifp->if_resolvemulti = ether_resolvemulti; #ifdef VIMAGE ifp->if_reassign = ether_reassign; #endif if (ifp->if_baudrate == 0) ifp->if_baudrate = IF_Mbps(10); /* just a default */ ifp->if_broadcastaddr = etherbroadcastaddr; ifa = ifp->if_addr; KASSERT(ifa != NULL, ("%s: no lladdr!\n", __func__)); sdl = (struct sockaddr_dl *)ifa->ifa_addr; sdl->sdl_type = IFT_ETHER; sdl->sdl_alen = ifp->if_addrlen; bcopy(lla, LLADDR(sdl), ifp->if_addrlen); bpfattach(ifp, DLT_EN10MB, ETHER_HDR_LEN); if (ng_ether_attach_p != NULL) (*ng_ether_attach_p)(ifp); /* Announce Ethernet MAC address if non-zero. */ for (i = 0; i < ifp->if_addrlen; i++) if (lla[i] != 0) break; if (i != ifp->if_addrlen) if_printf(ifp, "Ethernet address: %6D\n", lla, ":"); uuid_ether_add(LLADDR(sdl)); } /* * Perform common duties while detaching an Ethernet interface */ void ether_ifdetach(struct ifnet *ifp) { struct sockaddr_dl *sdl; sdl = (struct sockaddr_dl *)(ifp->if_addr->ifa_addr); uuid_ether_del(LLADDR(sdl)); if (ifp->if_l2com != NULL) { KASSERT(ng_ether_detach_p != NULL, ("ng_ether_detach_p is NULL")); (*ng_ether_detach_p)(ifp); } bpfdetach(ifp); if_detach(ifp); } #ifdef VIMAGE void ether_reassign(struct ifnet *ifp, struct vnet *new_vnet, char *unused __unused) { if (ifp->if_l2com != NULL) { KASSERT(ng_ether_detach_p != NULL, ("ng_ether_detach_p is NULL")); (*ng_ether_detach_p)(ifp); } if (ng_ether_attach_p != NULL) { CURVNET_SET_QUIET(new_vnet); (*ng_ether_attach_p)(ifp); CURVNET_RESTORE(); } } #endif SYSCTL_DECL(_net_link); SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet"); #if 0 /* * This is for reference. We have a table-driven version * of the little-endian crc32 generator, which is faster * than the double-loop. */ uint32_t ether_crc32_le(const uint8_t *buf, size_t len) { size_t i; uint32_t crc; int bit; uint8_t data; crc = 0xffffffff; /* initial value */ for (i = 0; i < len; i++) { for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) { carry = (crc ^ data) & 1; crc >>= 1; if (carry) crc = (crc ^ ETHER_CRC_POLY_LE); } } return (crc); } #else uint32_t ether_crc32_le(const uint8_t *buf, size_t len) { static const uint32_t crctab[] = { 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac, 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c, 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c, 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c }; size_t i; uint32_t crc; crc = 0xffffffff; /* initial value */ for (i = 0; i < len; i++) { crc ^= buf[i]; crc = (crc >> 4) ^ crctab[crc & 0xf]; crc = (crc >> 4) ^ crctab[crc & 0xf]; } return (crc); } #endif uint32_t ether_crc32_be(const uint8_t *buf, size_t len) { size_t i; uint32_t crc, carry; int bit; uint8_t data; crc = 0xffffffff; /* initial value */ for (i = 0; i < len; i++) { for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) { carry = ((crc & 0x80000000) ? 1 : 0) ^ (data & 0x01); crc <<= 1; if (carry) crc = (crc ^ ETHER_CRC_POLY_BE) | carry; } } return (crc); } int ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data) { struct ifaddr *ifa = (struct ifaddr *) data; struct ifreq *ifr = (struct ifreq *) data; int error = 0; switch (command) { case SIOCSIFADDR: ifp->if_flags |= IFF_UP; switch (ifa->ifa_addr->sa_family) { #ifdef INET case AF_INET: ifp->if_init(ifp->if_softc); /* before arpwhohas */ arp_ifinit(ifp, ifa); break; #endif default: ifp->if_init(ifp->if_softc); break; } break; case SIOCGIFADDR: { struct sockaddr *sa; sa = (struct sockaddr *) & ifr->ifr_data; bcopy(IF_LLADDR(ifp), (caddr_t) sa->sa_data, ETHER_ADDR_LEN); } break; case SIOCSIFMTU: /* * Set the interface MTU. */ if (ifr->ifr_mtu > ETHERMTU) { error = EINVAL; } else { ifp->if_mtu = ifr->ifr_mtu; } break; default: error = EINVAL; /* XXX netbsd has ENOTTY??? */ break; } return (error); } static int ether_resolvemulti(struct ifnet *ifp, struct sockaddr **llsa, struct sockaddr *sa) { struct sockaddr_dl *sdl; #ifdef INET struct sockaddr_in *sin; #endif #ifdef INET6 struct sockaddr_in6 *sin6; #endif u_char *e_addr; switch(sa->sa_family) { case AF_LINK: /* * No mapping needed. Just check that it's a valid MC address. */ sdl = (struct sockaddr_dl *)sa; e_addr = LLADDR(sdl); if (!ETHER_IS_MULTICAST(e_addr)) return EADDRNOTAVAIL; *llsa = 0; return 0; #ifdef INET case AF_INET: sin = (struct sockaddr_in *)sa; if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) return EADDRNOTAVAIL; sdl = link_init_sdl(ifp, *llsa, IFT_ETHER); sdl->sdl_alen = ETHER_ADDR_LEN; e_addr = LLADDR(sdl); ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr); *llsa = (struct sockaddr *)sdl; return 0; #endif #ifdef INET6 case AF_INET6: sin6 = (struct sockaddr_in6 *)sa; if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { /* * An IP6 address of 0 means listen to all * of the Ethernet multicast address used for IP6. * (This is used for multicast routers.) */ ifp->if_flags |= IFF_ALLMULTI; *llsa = 0; return 0; } if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) return EADDRNOTAVAIL; sdl = link_init_sdl(ifp, *llsa, IFT_ETHER); sdl->sdl_alen = ETHER_ADDR_LEN; e_addr = LLADDR(sdl); ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr); *llsa = (struct sockaddr *)sdl; return 0; #endif default: /* * Well, the text isn't quite right, but it's the name * that counts... */ return EAFNOSUPPORT; } } static moduledata_t ether_mod = { .name = "ether", }; void ether_vlan_mtap(struct bpf_if *bp, struct mbuf *m, void *data, u_int dlen) { struct ether_vlan_header vlan; struct mbuf mv, mb; KASSERT((m->m_flags & M_VLANTAG) != 0, ("%s: vlan information not present", __func__)); KASSERT(m->m_len >= sizeof(struct ether_header), ("%s: mbuf not large enough for header", __func__)); bcopy(mtod(m, char *), &vlan, sizeof(struct ether_header)); vlan.evl_proto = vlan.evl_encap_proto; vlan.evl_encap_proto = htons(ETHERTYPE_VLAN); vlan.evl_tag = htons(m->m_pkthdr.ether_vtag); m->m_len -= sizeof(struct ether_header); m->m_data += sizeof(struct ether_header); /* * If a data link has been supplied by the caller, then we will need to * re-create a stack allocated mbuf chain with the following structure: * * (1) mbuf #1 will contain the supplied data link * (2) mbuf #2 will contain the vlan header * (3) mbuf #3 will contain the original mbuf's packet data * * Otherwise, submit the packet and vlan header via bpf_mtap2(). */ if (data != NULL) { mv.m_next = m; mv.m_data = (caddr_t)&vlan; mv.m_len = sizeof(vlan); mb.m_next = &mv; mb.m_data = data; mb.m_len = dlen; bpf_mtap(bp, &mb); } else bpf_mtap2(bp, &vlan, sizeof(vlan), m); m->m_len += sizeof(struct ether_header); m->m_data -= sizeof(struct ether_header); } struct mbuf * ether_vlanencap(struct mbuf *m, uint16_t tag) { struct ether_vlan_header *evl; M_PREPEND(m, ETHER_VLAN_ENCAP_LEN, M_NOWAIT); if (m == NULL) return (NULL); /* M_PREPEND takes care of m_len, m_pkthdr.len for us */ if (m->m_len < sizeof(*evl)) { m = m_pullup(m, sizeof(*evl)); if (m == NULL) return (NULL); } /* * Transform the Ethernet header into an Ethernet header * with 802.1Q encapsulation. */ evl = mtod(m, struct ether_vlan_header *); bcopy((char *)evl + ETHER_VLAN_ENCAP_LEN, (char *)evl, ETHER_HDR_LEN - ETHER_TYPE_LEN); evl->evl_encap_proto = htons(ETHERTYPE_VLAN); evl->evl_tag = htons(tag); return (m); } DECLARE_MODULE(ether, ether_mod, SI_SUB_INIT_IF, SI_ORDER_ANY); MODULE_VERSION(ether, 1); Index: projects/routing/sys/net/rt_nhops.c =================================================================== --- projects/routing/sys/net/rt_nhops.c (revision 287053) +++ projects/routing/sys/net/rt_nhops.c (revision 287054) @@ -1,635 +1,666 @@ /*- * Copyright (c) 2014 * Alexander V. Chernikov * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ /* * Temporary file. In future it should be split between net/route.c * and per-AF files like netinet/in_rmx.c | netinet6/in6_rmx.c */ #include "opt_inet.h" #include "opt_inet6.h" #include "opt_route.h" #include "opt_mpath.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef RADIX_MPATH #include #endif #include #include #include #include #include #include #include #include #include struct fwd_info { fib_lookup_t *lookup; void *state; }; #define FWD_FSM_NONE 0 #define FWD_FSM_INIT 1 #define FWD_FSM_FWD 2 struct fwd_control { int fwd_state; /* FSM */ struct fwd_module *fm; }; #if 0 static struct fwd_info *fwd_db[FWD_SIZE]; static struct fwd_control *fwd_ctl[FWD_SIZE]; static TAILQ_HEAD(fwd_module_list, fwd_module) modulehead = TAILQ_HEAD_INITIALIZER(modulehead); static struct fwd_module_list fwd_modules[FWD_SIZE]; static uint8_t fwd_map_af[] = { AF_INET, AF_INET6, }; static struct rwlock fwd_lock; #define FWD_LOCK_INIT() rw_init(&fwd_lock, "fwd_lock") #define FWD_RLOCK() rw_rlock(&fwd_lock) #define FWD_RUNLOCK() rw_runlock(&fwd_lock) #define FWD_WLOCK() rw_wlock(&fwd_lock) #define FWD_WUNLOCK() rw_wunlock(&fwd_lock) int fwd_attach_fib(struct fwd_module *fm, u_int fib); int fwd_destroy_fib(struct fwd_module *fm, u_int fib); #endif +static inline uint16_t fib_rte_to_nh_flags(int rt_flags); #ifdef INET static void fib4_rte_to_nh_extended(struct rtentry *rte, struct in_addr dst, struct nhop4_extended *pnh4); static void fib4_rte_to_nh_basic(struct rtentry *rte, struct in_addr dst, struct nhop4_basic *pnh4); #endif #ifdef INET static void fib6_rte_to_nh_basic(struct rtentry *rte, struct in6_addr dst, struct nhop6_basic *pnh6); #endif MALLOC_DEFINE(M_RTFIB, "rtfib", "routing fwd"); /* * Per-AF fast routines returning minimal needed info. * It is not safe to dereference any pointers since it * may end up with use-after-free case. * Typically it may be used to check if outgoing * interface matches or to calculate proper MTU. * * Note that returned interface pointer is logical one, * e.g. actual transmit ifp may be different. * Difference may be triggered by * 1) loopback routes installed for interface addresses. * e.g. for address 10.0.0.1 with prefix /24 bound to * interface ix0, "logical" interface will be "ix0", * while "trasmit" interface will be "lo0" since this is * loopback route. You should consider using other * functions if you need "transmit" interface or both. * * * Returns 0 on match, error code overwise. */ -#define NHOP_FLAGS_MASK (RTF_REJECT|RTF_BLACKHOLE) //#define NHOP_DIRECT #define RNTORT(p) ((struct rtentry *)(p)) /* * Copies proper nexthop data based on @nh_src nexthop. * * For non-ECMP nexthop function simply copies @nh_src. * For ECMP nexthops flowid is used to select proper * nexthop. * */ static inline void -fib_choose_prepend(uint32_t fibnum, struct nhop_data *nh_src, - uint32_t flowid, struct nhop_data *nh, int af) +fib_choose_prepend(uint32_t fibnum, struct nhop_prepend *nh_src, + uint32_t flowid, struct nhop_prepend *nh, int af) { struct nhop_multi *nh_multi; int idx; - if ((nh_src->nh_flags & NH_FLAGS_RECURSE) != 0) { + if ((nh_src->nh_flags & NHF_RECURSE) != 0) { /* * Recursive nexthop. Choose direct nexthop * based on flowid. */ nh_multi = (struct nhop_multi *)nh_src; idx = nh_multi->nh_nhops[flowid % nh_multi->nh_count]; #if 0 KASSERT((fibnum < rt_numfibs), ("fib4_lookup_prepend§: bad fibnum")); rnh = rt_tables_get_rnh(fibnum, AF_INET); //nh_src = &rnh->nhops[i]; #endif } *nh = *nh_src; /* TODO: Do some light-weight refcounting on egress ifp's */ } static inline void -fib_free_nh(uint32_t fibnum, struct nhop_data *nh, int af) +fib_free_nh_prepend(uint32_t fibnum, struct nhop_prepend *nh, int af) { /* TODO: Do some light-weight refcounting on egress ifp's */ } #ifdef INET void -fib4_free_nh(uint32_t fibnum, struct nhop_data *nh) +fib4_free_nh_prepend(uint32_t fibnum, struct nhop_prepend *nh) { - fib_free_nh(fibnum, nh, AF_INET); + fib_free_nh_prepend(fibnum, nh, AF_INET); } void -fib4_choose_prepend(uint32_t fibnum, struct nhop_data *nh_src, - uint32_t flowid, struct nhop_data *nh, struct nhop4_extended *nh_ext) +fib4_choose_prepend(uint32_t fibnum, struct nhop_prepend *nh_src, + uint32_t flowid, struct nhop_prepend *nh, struct nhop4_extended *nh_ext) { fib_choose_prepend(fibnum, nh_src, flowid, nh, AF_INET); if (nh_ext == NULL) return; nh_ext->nh_ifp = NH_LIFP(nh); nh_ext->nh_mtu = nh->nh_mtu; nh_ext->nh_flags = nh->nh_flags; #if 0 /* TODO: copy source/gw address from extended nexthop data */ nh_ext->nh_addr = ; nh_ext->nh_src= ; #endif } /* * Function performs lookup in IPv4 table fib @fibnum. * * In case of successful lookup @nh header is filled with * appropriate interface info and full L2 header to prepend. * - * If no valid ARP record is present, NH_FLAGS_L2_INCOMPLETE flag + * If no valid ARP record is present, NHF_L2_INCOMPLETE flag * is set and gateway address is stored into nh->d.gw4 * * If @nh_ext is not NULL, additional nexthop data is stored there. * * Returns 0 on success. * */ int fib4_lookup_prepend(uint32_t fibnum, struct in_addr dst, struct mbuf *m, - struct nhop_data *nh, struct nhop4_extended *nh_ext) + struct nhop_prepend *nh, struct nhop4_extended *nh_ext) { struct radix_node_head *rnh; struct radix_node *rn; struct sockaddr_in *gw_sa, sin; struct ifnet *lifp; struct in_addr gw; struct ether_header *eh; int error, flags; //uint32_t flowid; struct rtentry *rte; KASSERT((fibnum < rt_numfibs), ("fib4_lookup_prepend: bad fibnum")); rnh = rt_tables_get_rnh(fibnum, AF_INET); if (rnh == NULL) return (EHOSTUNREACH); /* Prepare lookup key */ memset(&sin, 0, sizeof(sin)); sin.sin_len = sizeof(struct sockaddr_in); sin.sin_addr = dst; RADIX_NODE_HEAD_RLOCK(rnh); rn = rnh->rnh_matchaddr((void *)&sin, rnh); rte = RNTORT(rn); if (rn == NULL || ((rn->rn_flags & RNF_ROOT) != 0) || RT_LINK_IS_UP(rte->rt_ifp) == 0) { RADIX_NODE_HEAD_RUNLOCK(rnh); return (EHOSTUNREACH); } /* * Currently we fill in @nh ourselves. * In near future rte will have nhop index to copy from. */ /* Calculate L3 info */ flags = 0; nh->nh_mtu = min(rte->rt_mtu, rte->rt_ifp->if_mtu); if (rte->rt_flags & RTF_GATEWAY) { gw_sa = (struct sockaddr_in *)rte->rt_gateway; gw = gw_sa->sin_addr; } else gw = dst; /* Set flags */ - flags = rte->rt_flags & NHOP_FLAGS_MASK; + flags = fib_rte_to_nh_flags(rte->rt_flags); gw_sa = (struct sockaddr_in *)rt_key(rte); if (gw_sa->sin_addr.s_addr == 0) - flags |= NHOP_DEFAULT; + flags |= NHF_DEFAULT; /* * TODO: nh L2/L3 resolve. * Currently all we have is rte ifp. * Simply use it. */ lifp = rte->rt_ifp; /* Save both logical and transmit interface indexes */ nh->lifp_idx = lifp->if_index; nh->i.ifp_idx = nh->lifp_idx; if (nh_ext != NULL) { /* Fill in extended info */ fib4_rte_to_nh_extended(rte, dst, nh_ext); } RADIX_NODE_HEAD_RUNLOCK(rnh); nh->nh_flags = flags; /* * Try to lookup L2 info. * Do this using separate LLE locks. * TODO: move this under radix lock. */ if (lifp->if_type == IFT_ETHER) { eh = (struct ether_header *)nh->d.data; /* * Fill in ethernet header. * It should be already presented if we're * sending data via known gateway. */ error = arpresolve_fast(lifp, gw, m ? m->m_flags : 0, eh->ether_dhost); if (error == 0) { memcpy(&eh->ether_shost, IF_LLADDR(lifp), ETHER_ADDR_LEN); eh->ether_type = htons(ETHERTYPE_IP); nh->nh_count = ETHER_HDR_LEN; return (0); } } /* Notify caller that no L2 info is linked */ nh->nh_count = 0; - nh->nh_flags |= NH_FLAGS_L2_INCOMPLETE; + nh->nh_flags |= NHF_L2_INCOMPLETE; /* ..And save gateway address */ nh->d.gw4 = gw; return (0); } int -fib4_sendmbuf(struct ifnet *ifp, struct mbuf *m, struct nhop_data *nh, +fib4_sendmbuf(struct ifnet *ifp, struct mbuf *m, struct nhop_prepend *nh, struct in_addr dst) { int error; - if (nh != NULL && (nh->nh_flags & NH_FLAGS_L2_INCOMPLETE) == 0) { + if (nh != NULL && (nh->nh_flags & NHF_L2_INCOMPLETE) == 0) { /* * Fast path case. Most packets should * be sent from here. * TODO: Make special ifnet * 'if_output_frame' handler for that. */ struct route_compat rc; struct ether_header *eh; rc.ro_flags = AF_INET << 8 | RT_NHOP; rc.ro_nh = nh; M_PREPEND(m, nh->nh_count, M_NOWAIT); if (m == NULL) return (ENOBUFS); eh = mtod(m, struct ether_header *); memcpy(eh, nh->d.data, nh->nh_count); error = (*ifp->if_output)(ifp, m, NULL, (struct route *)&rc); } else { struct sockaddr_in gw_out; memset(&gw_out, 0, sizeof(gw_out)); gw_out.sin_len = sizeof(gw_out); gw_out.sin_family = AF_INET; gw_out.sin_addr = nh ? nh->d.gw4 : dst; error = (*ifp->if_output)(ifp, m, (const struct sockaddr *)&gw_out, NULL); } return (error); } +static inline uint16_t +fib_rte_to_nh_flags(int rt_flags) +{ + uint16_t res; + res = (rt_flags & RTF_REJECT) ? NHF_REJECT : 0; + res |= (rt_flags & RTF_BLACKHOLE) ? NHF_BLACKHOLE : 0; + res |= (rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) ? NHF_REDIRECT : 0; + res |= (rt_flags & RTF_BROADCAST) ? NHF_BROADCAST : 0; + + return (res); +} + + static void fib4_rte_to_nh_extended(struct rtentry *rte, struct in_addr dst, struct nhop4_extended *pnh4) { struct sockaddr_in *gw; struct in_ifaddr *ia; pnh4->nh_ifp = rte->rt_ifp; pnh4->nh_mtu = min(rte->rt_mtu, rte->rt_ifp->if_mtu); if (rte->rt_flags & RTF_GATEWAY) { gw = (struct sockaddr_in *)rte->rt_gateway; pnh4->nh_addr = gw->sin_addr; } else pnh4->nh_addr = dst; ia = ifatoia(rte->rt_ifa); pnh4->nh_src = IA_SIN(ia)->sin_addr; /* Set flags */ - pnh4->nh_flags = rte->rt_flags & NHOP_FLAGS_MASK; - if (rte->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) - pnh4->nh_flags |= NHOP_REDIRECT; + pnh4->nh_flags = fib_rte_to_nh_flags(rte->rt_flags); gw = (struct sockaddr_in *)rt_key(rte); if (gw->sin_addr.s_addr == 0) - pnh4->nh_flags |= NHOP_DEFAULT; + pnh4->nh_flags |= NHF_DEFAULT; } static void fib4_rte_to_nh_basic(struct rtentry *rte, struct in_addr dst, struct nhop4_basic *pnh4) { struct sockaddr_in *gw; pnh4->nh_ifp = rte->rt_ifa->ifa_ifp; pnh4->nh_mtu = min(rte->rt_mtu, rte->rt_ifp->if_mtu); if (rte->rt_flags & RTF_GATEWAY) { gw = (struct sockaddr_in *)rte->rt_gateway; pnh4->nh_addr = gw->sin_addr; } else pnh4->nh_addr = dst; /* Set flags */ - pnh4->nh_flags = rte->rt_flags & NHOP_FLAGS_MASK; - if (rte->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) - pnh4->nh_flags |= NHOP_REDIRECT; + pnh4->nh_flags = fib_rte_to_nh_flags(rte->rt_flags); gw = (struct sockaddr_in *)rt_key(rte); if (gw->sin_addr.s_addr == 0) - pnh4->nh_flags |= NHOP_DEFAULT; + pnh4->nh_flags |= NHF_DEFAULT; } +/* + * Performs IPv4 route table lookup on @dst. Returns 0 on success. + * Stores nexthop info provided @pnh4 structure. + * Note that + * - nh_ifp cannot be safely dereferenced + * - nh_ifp represents ifaddr ifp (e.g. if looking up address on + * interface "ix0" pointer to "ix0" interface will be returned instead + * of "lo0") + * - howewer mtu from "transmit" interface will be returned. + */ int fib4_lookup_nh_basic(uint32_t fibnum, struct in_addr dst, uint32_t flowid, struct nhop4_basic *pnh4) { struct radix_node_head *rnh; struct radix_node *rn; struct sockaddr_in sin; struct rtentry *rte; KASSERT((fibnum < rt_numfibs), ("fib4_lookup_nh_basic: bad fibnum")); rnh = rt_tables_get_rnh(fibnum, AF_INET); if (rnh == NULL) return (ENOENT); /* Prepare lookup key */ memset(&sin, 0, sizeof(sin)); sin.sin_len = sizeof(struct sockaddr_in); sin.sin_addr = dst; RADIX_NODE_HEAD_RLOCK(rnh); rn = rnh->rnh_matchaddr((void *)&sin, rnh); if (rn != NULL && ((rn->rn_flags & RNF_ROOT) == 0)) { rte = RNTORT(rn); /* Ensure route & ifp is UP */ if (RT_LINK_IS_UP(rte->rt_ifp)) { fib4_rte_to_nh_basic(rte, dst, pnh4); RADIX_NODE_HEAD_RUNLOCK(rnh); return (0); } } RADIX_NODE_HEAD_RUNLOCK(rnh); return (ENOENT); } +/* + * Performs IPv4 route table lookup on @dst. Returns 0 on success. + * Stores extende nexthop info provided @pnh4 structure. + * Note that + * - nh_ifp cannot be safely dereferenced unless NHOP_LOOKUP_REF is specified. + * - in that case you need to call fib4_free_nh_ext() + * - nh_ifp represents logical transmit interface (rt_ifp) + * - mtu from logical transmit interface will be returned. + */ int -fib4_lookup_nh_extended(uint32_t fibnum, struct in_addr dst, uint32_t flowid, - struct nhop4_extended *pnh4) +fib4_lookup_nh_ext(uint32_t fibnum, struct in_addr dst, uint32_t flowid, + uint32_t flags, struct nhop4_extended *pnh4) { struct radix_node_head *rnh; struct radix_node *rn; struct sockaddr_in sin; struct rtentry *rte; - KASSERT((fibnum < rt_numfibs), ("fib4_lookup_nh_basic: bad fibnum")); + KASSERT((fibnum < rt_numfibs), ("fib4_lookup_nh_ext: bad fibnum")); rnh = rt_tables_get_rnh(fibnum, AF_INET); if (rnh == NULL) return (ENOENT); /* Prepare lookup key */ memset(&sin, 0, sizeof(sin)); sin.sin_len = sizeof(struct sockaddr_in); sin.sin_addr = dst; RADIX_NODE_HEAD_RLOCK(rnh); rn = rnh->rnh_matchaddr((void *)&sin, rnh); if (rn != NULL && ((rn->rn_flags & RNF_ROOT) == 0)) { rte = RNTORT(rn); /* Ensure route & ifp is UP */ if (RT_LINK_IS_UP(rte->rt_ifp)) { fib4_rte_to_nh_extended(rte, dst, pnh4); + if ((flags & NHOP_LOOKUP_REF) != 0) { + /* TODO: Do lwref on egress ifp's */ + } RADIX_NODE_HEAD_RUNLOCK(rnh); return (0); } } RADIX_NODE_HEAD_RUNLOCK(rnh); return (ENOENT); } void fib4_free_nh_ext(uint32_t fibnum, struct nhop4_extended *pnh4) { } #endif #ifdef INET6 void -fib6_free_nh(uint32_t fibnum, struct nhop_data *nh) +fib6_free_nh_prepend(uint32_t fibnum, struct nhop_prepend *nh) { - fib_free_nh(fibnum, nh, AF_INET6); + fib_free_nh_prepend(fibnum, nh, AF_INET6); } void -fib6_choose_prepend(uint32_t fibnum, struct nhop_data *nh_src, - uint32_t flowid, struct nhop_data *nh, struct nhop6_extended *nh_ext) +fib6_choose_prepend(uint32_t fibnum, struct nhop_prepend *nh_src, + uint32_t flowid, struct nhop_prepend *nh, struct nhop6_extended *nh_ext) { fib_choose_prepend(fibnum, nh_src, flowid, nh, AF_INET6); if (nh_ext == NULL) return; nh_ext->nh_ifp = NH_LIFP(nh); nh_ext->nh_mtu = nh->nh_mtu; nh_ext->nh_flags = nh->nh_flags; /* nh_ext->nh_addr = ; nh_ext->nh_src= ; */ } static void fib6_rte_to_nh_basic(struct rtentry *rte, struct in6_addr dst, struct nhop6_basic *pnh6) { struct sockaddr_in6 *gw; pnh6->nh_ifp = rte->rt_ifa->ifa_ifp; pnh6->nh_mtu = min(rte->rt_mtu, rte->rt_ifp->if_mtu); if (rte->rt_flags & RTF_GATEWAY) { gw = (struct sockaddr_in6 *)rte->rt_gateway; pnh6->nh_addr = gw->sin6_addr; } else pnh6->nh_addr = dst; /* Set flags */ - pnh6->nh_flags = rte->rt_flags & NHOP_FLAGS_MASK; + pnh6->nh_flags = fib_rte_to_nh_flags(rte->rt_flags); gw = (struct sockaddr_in6 *)rt_key(rte); if (IN6_IS_ADDR_UNSPECIFIED(&gw->sin6_addr)) - pnh6->nh_flags |= NHOP_DEFAULT; + pnh6->nh_flags |= NHF_DEFAULT; } int fib6_lookup_nh_basic(uint32_t fibnum, struct in6_addr dst, uint32_t flowid, struct nhop6_basic *pnh6) { struct radix_node_head *rnh; struct radix_node *rn; struct sockaddr_in6 sin6; struct rtentry *rte; KASSERT((fibnum < rt_numfibs), ("fib6_lookup_nh_basic: bad fibnum")); rnh = rt_tables_get_rnh(fibnum, AF_INET); if (rnh == NULL) return (ENOENT); /* Prepare lookup key */ memset(&sin6, 0, sizeof(sin6)); sin6.sin6_addr = dst; RADIX_NODE_HEAD_RLOCK(rnh); rn = rnh->rnh_matchaddr((void *)&sin6, rnh); if (rn != NULL && ((rn->rn_flags & RNF_ROOT) == 0)) { rte = RNTORT(rn); /* Ensure route & ifp is UP */ if (RT_LINK_IS_UP(rte->rt_ifp)) { fib6_rte_to_nh_basic(rte, dst, pnh6); RADIX_NODE_HEAD_RUNLOCK(rnh); return (0); } } RADIX_NODE_HEAD_RUNLOCK(rnh); return (ENOENT); } #endif #if 0 typedef void nhop_change_cb_t(void *state); struct nhop_tracker { TAILQ_ENTRY(nhop_tracker) next; nhop_change_cb_t *f; void *state; uint32_t fibnum; struct sockaddr_storage ss; }; struct nhop_tracker * nhop_alloc_tracked(uint32_t fibnum, struct sockaddr *sa, nhop_change_cb_t *f, void *state) { struct nhop_tracker *nt; nt = malloc(sizeof(struct nhop_tracker), M_RTFIB, M_WAITOK | M_ZERO); nt->f = f; nt-state = state; nt->fibnum = fibnum; memcpy(&nt->ss, sa, sa->sa_len); return (nt); } int nhop_bind(struct nhop_tracker *nt) { NHOP_LOCK(nnh); NHOP_UNLOCK(nnh); return (0); } #endif Index: projects/routing/sys/net/rt_nhops.h =================================================================== --- projects/routing/sys/net/rt_nhops.h (revision 287053) +++ projects/routing/sys/net/rt_nhops.h (revision 287054) @@ -1,268 +1,274 @@ /*- * Copyright (c) 2014 * Alexander V. Chernikov * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #ifndef _NET_RT_NHOPS_H_ #define _NET_RT_NHOPS_H_ #define NH_TYPE_DIRECT 1 /* Directly reachable, no data */ #define NH_TYPE_BLACKHOLE 2 /* Blackhole route */ #define NH_TYPE_REJECT 3 /* Send reject */ #define NH_TYPE_L2 4 /* Provides full prepend header */ #define NH_TYPE_MUTATOR 5 /* NH+callback function */ #define NH_TYPE_MULTIPATH 6 /* Multipath route */ struct nhop_ctl_info { uint64_t refcnt; /* Use references */ uint64_t flags; /* Options */ }; /* Multipath nhop info */ struct nhop_mpath_info { uint16_t nhop; /* Netxthop id */ }; /* mutator info */ struct nhop_mutator_info; -struct nhop_data; +struct nhop_prepend; -typedef int nhop_mutate_t(struct mbuf **, struct nhop_data *nd, void *storage); +typedef int nhop_mutate_t(struct mbuf **, struct nhop_prepend *nd, void *storage); struct nhop_mutator_info { nhop_mutate_t *func; char data[]; }; /* Structures used for forwarding purposes */ -#define MAX_PREPEND_LEN 56 /* Max data that can be prepended */ +#define MAX_PREPEND_LEN 48 /* Max data that can be prepended */ /* Non-recursive nexthop */ -struct nhop_data { - uint8_t nh_flags; /* NH flags */ - uint8_t nh_count; /* Number of nexthops or data length */ +struct nhop_prepend { + uint16_t nh_flags; /* NH flags */ + uint8_t nh_count; /* Number of nexthops or data length */ + uint8_t spare0; uint16_t nh_mtu; /* given nhop MTU */ uint16_t lifp_idx; /* Logical interface index */ union { uint16_t ifp_idx; /* Transmit interface index */ uint16_t nhop_idx; /* L2 multipath nhop index */ } i; + uint16_t spare1[3]; union { char data[MAX_PREPEND_LEN]; /* data to prepend */ #ifdef INET struct in_addr gw4; /* IPv4 gw address */ #endif #ifdef INET6 struct in6_addr gw6; /* IPv4 gw address */ #endif } d; }; + /* Internal flags */ -#define NH_FLAGS_RECURSE 0x01 /* Nexthop structure is recursive */ -#define NH_FLAGS_L2_NHOP 0x02 /* L2 interface has to be selected */ -#define NH_FLAGS_L2_ME 0x04 /* dst L2 address is our address */ -#define NH_FLAGS_L2_INCOMPLETE 0x08 /* L2 header not prepended */ +#define NHF_RECURSE 0x0001 /* Nexthop structure is recursive */ +#define NHF_L2_NHOP 0x0002 /* L2 interface has to be selected */ +#define NHF_L2_ME 0x0004 /* dst L2 address is our address */ +#define NHF_L2_INCOMPLETE 0x0008 /* L2 header not prepended */ +/* External flags */ +#define NHF_REJECT 0x0010 /* RTF_REJECT */ +#define NHF_BLACKHOLE 0x0020 /* RTF_BLACKHOLE */ +#define NHF_REDIRECT 0x0040 /* RTF_DYNAMIC|RTF_MODIFIED */ +#define NHF_DEFAULT 0x0080 /* Default route */ +#define NHF_BROADCAST 0x0100 /* RTF_BROADCAST */ + #define NH_LIFP(nh) ifnet_byindex_locked((nh)->lifp_idx) #define NH_TIFP(nh) ifnet_byindex_locked((nh)->i.ifp_idx) /* L2/L3 recursive nexthop */ struct nhop_multi { uint8_t nh_flags; /* NH flags */ uint8_t nh_count; /* Number of nexthops or data length */ uint8_t spare[2]; uint16_t nh_nhops[30]; /* Nexthop indexes */ }; /* Control plane nexthop data */ struct nhop_info { }; /* Per-AF per-fib nhop table */ struct nhops_descr { uint32_t nhop_size; /* Nehthop data size */ uint32_t nhops_max; /* Max number of nhops */ void *nhops_data; /* Pointer to nhop data table */ void *nhops_info; /* Pointer to nhop info table */ }; #if 0 -typedef int nhop_resolve_t(struct sockaddr *dst, u_int fib, struct nhop_data *nd, struct nhop_info *nf); +typedef int nhop_resolve_t(struct sockaddr *dst, u_int fib, struct nhop_prepend *nd, struct nhop_info *nf); int lla_create_notify(struct sockaddr *dst, u_int fib, lla_notify_t *func, void *state, int flags); #endif /* Basic nexthop info used for uRPF/mtu checks */ struct nhop4_basic { struct ifnet *nh_ifp; /* Logical egress interface */ uint16_t nh_mtu; /* nexthop mtu */ uint16_t nh_flags; /* nhop flags */ struct in_addr nh_addr; /* GW/DST IPv4 address */ }; struct nhop6_basic { struct ifnet *nh_ifp; /* Logical egress interface */ uint16_t nh_mtu; /* nexthop mtu */ uint16_t nh_flags; /* nhop flags */ uint8_t spare[4]; struct in6_addr nh_addr; /* GW/DST IPv4 address */ }; struct nhop64_basic { union { struct nhop4_basic nh4; struct nhop6_basic nh6; } u; }; /* Extended nexthop info used for control protocols */ struct nhop4_extended { struct ifnet *nh_ifp; /* Logical egress interface */ uint16_t nh_mtu; /* nexthop mtu */ uint16_t nh_flags; /* nhop flags */ uint8_t spare[4]; struct in_addr nh_addr; /* GW/DST IPv4 address */ struct in_addr nh_src; /* default source IPv4 address */ uint64_t spare2[2]; }; struct nhop6_extended { struct ifnet *nh_ifp; /* Logical egress interface */ uint16_t nh_mtu; /* nexthop mtu */ uint16_t nh_flags; /* nhop flags */ uint8_t spare[4]; struct in6_addr nh_addr; /* GW/DST IPv6 address */ struct in6_addr nh_src; /* default source IPv6 address */ uint64_t spare2[2]; }; struct nhop64_extended { union { struct nhop4_extended nh4; struct nhop6_extended nh6; } u; }; struct route_info { - struct nhop_data *ri_nh; /* Desired nexthop to use */ + struct nhop_prepend *ri_nh; /* Desired nexthop to use */ struct nhop64_basic *ri_nh_info; /* Get selected route info */ uint16_t ri_mtu; uint16_t spare[3]; }; struct route_compat { - struct nhop_data *ro_nh; + struct nhop_prepend *ro_nh; void *spare0; void *spare1; int ro_flags; }; int fib4_lookup_nh_basic(uint32_t fibnum, struct in_addr dst, uint32_t flowid, struct nhop4_basic *pnh4); int fib6_lookup_nh_basic(uint32_t fibnum, struct in6_addr dst, uint32_t flowid, struct nhop6_basic *pnh6); -int fib4_lookup_nh_extended(uint32_t fibnum, struct in_addr dst, - uint32_t flowid, struct nhop4_extended *pnh4); +int fib4_lookup_nh_ext(uint32_t fibnum, struct in_addr dst, + uint32_t flowid, uint32_t flags, struct nhop4_extended *pnh4); void fib4_free_nh_ext(uint32_t fibnum, struct nhop4_extended *pnh4); +#define NHOP_LOOKUP_REF 0x01 -void fib4_free_nh(uint32_t fibnum, struct nhop_data *nh); -void fib4_choose_prepend(uint32_t fibnum, struct nhop_data *nh_src, - uint32_t flowid, struct nhop_data *nh, struct nhop4_extended *nh_ext); +void fib4_free_nh_prepend(uint32_t fibnum, struct nhop_prepend *nh); +void fib4_choose_prepend(uint32_t fibnum, struct nhop_prepend *nh_src, + uint32_t flowid, struct nhop_prepend *nh, struct nhop4_extended *nh_ext); int fib4_lookup_prepend(uint32_t fibnum, struct in_addr dst, struct mbuf *m, - struct nhop_data *nh, struct nhop4_extended *nh_ext); + struct nhop_prepend *nh, struct nhop4_extended *nh_ext); -int fib4_sendmbuf(struct ifnet *ifp, struct mbuf *m, struct nhop_data *nh, +int fib4_sendmbuf(struct ifnet *ifp, struct mbuf *m, struct nhop_prepend *nh, struct in_addr dst); -void fib6_free_nh(uint32_t fibnum, struct nhop_data *nh); -void fib6_choose_prepend(uint32_t fibnum, struct nhop_data *nh_src, - uint32_t flowid, struct nhop_data *nh, struct nhop6_extended *nh_ext); - -#define NHOP_REJECT 0x08 /* RTF_REJECT */ -#define NHOP_BLACKHOLE 0x1000 /* RTF_BLACKHOLE */ -#define NHOP_REDIRECT 0x10 /* RTF_DYNAMIC|RTF_MODIFIED */ -#define NHOP_DEFAULT 0x80 /* Default route */ +void fib6_free_nh_prepend(uint32_t fibnum, struct nhop_prepend *nh); +void fib6_choose_prepend(uint32_t fibnum, struct nhop_prepend *nh_src, + uint32_t flowid, struct nhop_prepend *nh, struct nhop6_extended *nh_ext); #define FWD_INET 0 #define FWD_INET6 1 #define FWD_SIZE 2 #define FWD_NAME_MAX 15 #define FWD_MULTIPATH 0x0001 /* has multipath support */ #define FWD_OLDMASKS 0x0002 /* has support for non-contig masks */ #define FWD_DEFAULT 0x0004 /* installs as default fib mechanism */ #define FWD_MANAGELOCK 0x0004 /* manage its own locking */ typedef void *fib_init_t(u_int fibnum); typedef void fib_destroy_t(void *state); typedef int fib_dump_t(void *state, struct radix_node_head *rnh); typedef int fib_change_t(void *state, int req, struct rtentry *rte, struct rt_addrinfo *info); typedef int fib_lookup_t(void *state, void *key, uint64_t *attr, u_int flowid, void *nhop); /* Structure used by external module */ struct fwd_module_info { uint8_t fwd_family; /* family we're registering to */ char name[FWD_NAME_MAX]; /* fwd module name */ uint32_t capabilities; fib_init_t *fib_init; fib_destroy_t *fib_destroy; fib_dump_t *fib_dump; fib_change_t *fib_change; fib_lookup_t *fib_lookup; }; /* Internal version of previous structure */ struct fwd_module { TAILQ_ENTRY(fwd_module) list; uint8_t fwd_family; char name[FWD_NAME_MAX]; uint32_t capabilities; fib_init_t *fib_init; fib_destroy_t *fib_destroy; fib_dump_t *fib_dump; fib_change_t *fib_change; fib_lookup_t *fib_lookup; }; int fwd_attach_module(struct fwd_module_info *m, void **); int fwd_destroy_module(void *state); int fwd_change_fib(struct radix_node_head *rnh, int req, struct rtentry *rte, struct rt_addrinfo *info); #endif Index: projects/routing/sys/netinet/if_ether.c =================================================================== --- projects/routing/sys/netinet/if_ether.c (revision 287053) +++ projects/routing/sys/netinet/if_ether.c (revision 287054) @@ -1,1037 +1,1037 @@ /*- * Copyright (c) 1982, 1986, 1988, 1993 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)if_ether.c 8.1 (Berkeley) 6/10/93 */ /* * Ethernet address resolution protocol. * TODO: * add "inuse/lock" bit (or ref. count) along with valid bit */ #include __FBSDID("$FreeBSD$"); #include "opt_inet.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef INET #include #endif #include #include #include #include #define SIN(s) ((const struct sockaddr_in *)(s)) #define SDL(s) ((struct sockaddr_dl *)s) SYSCTL_DECL(_net_link_ether); static SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, ""); static SYSCTL_NODE(_net_link_ether, PF_ARP, arp, CTLFLAG_RW, 0, ""); /* timer values */ static VNET_DEFINE(int, arpt_keep) = (20*60); /* once resolved, good for 20 * minutes */ static VNET_DEFINE(int, arp_maxtries) = 5; static VNET_DEFINE(int, arp_proxyall) = 0; static VNET_DEFINE(int, arpt_down) = 20; /* keep incomplete entries for * 20 seconds */ VNET_PCPUSTAT_DEFINE(struct arpstat, arpstat); /* ARP statistics, see if_arp.h */ VNET_PCPUSTAT_SYSINIT(arpstat); #ifdef VIMAGE VNET_PCPUSTAT_SYSUNINIT(arpstat); #endif /* VIMAGE */ static VNET_DEFINE(int, arp_maxhold) = 1; #define V_arpt_keep VNET(arpt_keep) #define V_arpt_down VNET(arpt_down) #define V_arp_maxtries VNET(arp_maxtries) #define V_arp_proxyall VNET(arp_proxyall) #define V_arp_maxhold VNET(arp_maxhold) SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_age, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(arpt_keep), 0, "ARP entry lifetime in seconds"); SYSCTL_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(arp_maxtries), 0, "ARP resolution attempts before returning error"); SYSCTL_INT(_net_link_ether_inet, OID_AUTO, proxyall, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(arp_proxyall), 0, "Enable proxy ARP for all suitable requests"); SYSCTL_INT(_net_link_ether_inet, OID_AUTO, wait, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(arpt_down), 0, "Incomplete ARP entry lifetime in seconds"); SYSCTL_VNET_PCPUSTAT(_net_link_ether_arp, OID_AUTO, stats, struct arpstat, arpstat, "ARP statistics (struct arpstat, net/if_arp.h)"); SYSCTL_INT(_net_link_ether_inet, OID_AUTO, maxhold, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(arp_maxhold), 0, "Number of packets to hold per ARP entry"); static void arp_init(void); static void arpintr(struct mbuf *); static void arptimer(void *); #ifdef INET static void in_arpinput(struct mbuf *); #endif static const struct netisr_handler arp_nh = { .nh_name = "arp", .nh_handler = arpintr, .nh_proto = NETISR_ARP, .nh_policy = NETISR_POLICY_SOURCE, }; #ifdef AF_INET /* * called by in_scrubprefix() to remove entry from the table when * the interface goes away */ void arp_ifscrub(struct ifnet *ifp, uint32_t addr) { struct sockaddr_in addr4; bzero((void *)&addr4, sizeof(addr4)); addr4.sin_len = sizeof(addr4); addr4.sin_family = AF_INET; addr4.sin_addr.s_addr = addr; IF_AFDATA_WLOCK(ifp); lla_lookup(LLTABLE(ifp), (LLE_DELETE | LLE_IFADDR), (struct sockaddr *)&addr4); IF_AFDATA_WUNLOCK(ifp); } #endif /* * Timeout routine. Age arp_tab entries periodically. */ static void arptimer(void *arg) { struct llentry *lle = (struct llentry *)arg; struct ifnet *ifp; if (lle->la_flags & LLE_STATIC) { return; } LLE_WLOCK(lle); if (callout_pending(&lle->la_timer)) { /* * Here we are a bit odd here in the treatment of * active/pending. If the pending bit is set, it got * rescheduled before I ran. The active * bit we ignore, since if it was stopped * in ll_tablefree() and was currently running * it would have return 0 so the code would * not have deleted it since the callout could * not be stopped so we want to go through * with the delete here now. If the callout * was restarted, the pending bit will be back on and * we just want to bail since the callout_reset would * return 1 and our reference would have been removed * by arpresolve() below. */ LLE_WUNLOCK(lle); return; } ifp = lle->lle_tbl->llt_ifp; CURVNET_SET(ifp->if_vnet); if ((lle->la_flags & LLE_DELETED) == 0) { int evt; if (lle->la_flags & LLE_VALID) evt = LLENTRY_EXPIRED; else evt = LLENTRY_TIMEDOUT; EVENTHANDLER_INVOKE(lle_event, lle, evt); } callout_stop(&lle->la_timer); /* XXX: LOR avoidance. We still have ref on lle. */ LLE_WUNLOCK(lle); IF_AFDATA_LOCK(ifp); LLE_WLOCK(lle); /* Guard against race with other llentry_free(). */ if (lle->la_flags & LLE_LINKED) { size_t pkts_dropped; LLE_REMREF(lle); pkts_dropped = llentry_free(lle); ARPSTAT_ADD(dropped, pkts_dropped); } else LLE_FREE_LOCKED(lle); IF_AFDATA_UNLOCK(ifp); ARPSTAT_INC(timeouts); CURVNET_RESTORE(); } /* * Broadcast an ARP request. Caller specifies: * - arp header source ip address * - arp header target ip address * - arp header source ethernet address */ void arprequest(struct ifnet *ifp, const struct in_addr *sip, const struct in_addr *tip, u_char *enaddr) { struct mbuf *m; struct arphdr *ah; struct sockaddr sa; u_char *carpaddr = NULL; if (sip == NULL) { /* * The caller did not supply a source address, try to find * a compatible one among those assigned to this interface. */ struct ifaddr *ifa; IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { if (ifa->ifa_addr->sa_family != AF_INET) continue; if (ifa->ifa_carp) { if ((*carp_iamatch_p)(ifa, &carpaddr) == 0) continue; sip = &IA_SIN(ifa)->sin_addr; } else { carpaddr = NULL; sip = &IA_SIN(ifa)->sin_addr; } if (0 == ((sip->s_addr ^ tip->s_addr) & IA_MASKSIN(ifa)->sin_addr.s_addr)) break; /* found it. */ } IF_ADDR_RUNLOCK(ifp); if (sip == NULL) { printf("%s: cannot find matching address\n", __func__); return; } } if (enaddr == NULL) enaddr = carpaddr ? carpaddr : (u_char *)IF_LLADDR(ifp); if ((m = m_gethdr(M_NOWAIT, MT_DATA)) == NULL) return; m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) + 2 * ifp->if_addrlen; m->m_pkthdr.len = m->m_len; M_ALIGN(m, m->m_len); ah = mtod(m, struct arphdr *); bzero((caddr_t)ah, m->m_len); #ifdef MAC mac_netinet_arp_send(ifp, m); #endif ah->ar_pro = htons(ETHERTYPE_IP); ah->ar_hln = ifp->if_addrlen; /* hardware address length */ ah->ar_pln = sizeof(struct in_addr); /* protocol address length */ ah->ar_op = htons(ARPOP_REQUEST); bcopy(enaddr, ar_sha(ah), ah->ar_hln); bcopy(sip, ar_spa(ah), ah->ar_pln); bcopy(tip, ar_tpa(ah), ah->ar_pln); sa.sa_family = AF_ARP; sa.sa_len = 2; m->m_flags |= M_BCAST; m_clrprotoflags(m); /* Avoid confusing lower layers. */ (*ifp->if_output)(ifp, m, &sa, NULL); ARPSTAT_INC(txrequests); } /* * * Saves lle address for @dst in @dst_addr. * Returns 0 if address was found&valid. */ int arpresolve_fast(struct ifnet *ifp, struct in_addr dst, u_int mflags, u_char *dst_addr) { int do_arp, error; struct llentry *la; struct sockaddr_in sin; if (mflags & M_BCAST) { memcpy(dst_addr, ifp->if_broadcastaddr, ifp->if_addrlen); return (0); } if (mflags & M_MCAST) { ETHER_MAP_IP_MULTICAST(&dst, dst_addr); return (0); } do_arp = 0; error = EAGAIN; memset(&sin, 0, sizeof(sin)); sin.sin_addr = dst; sin.sin_family = AF_INET; sin.sin_len = sizeof(sin); IF_AFDATA_RLOCK(ifp); la = lla_lookup(LLTABLE(ifp), 0, (const struct sockaddr *)&sin); /* * XXX: We need to convert all these checks to single one */ if (la != NULL && (la->la_flags & LLE_VALID) && ((la->la_flags & LLE_STATIC) || la->la_expire > time_uptime)) { bcopy(&la->ll_addr, dst_addr, ifp->if_addrlen); /* * If entry has an expiry time and it is approaching, * see if we need to send an ARP request within this * arpt_down interval. */ if (!(la->la_flags & LLE_STATIC) && time_uptime + la->la_preempt > la->la_expire) { do_arp = 1; la->la_preempt--; } error = 0; } if (la != NULL) LLE_RUNLOCK(la); IF_AFDATA_RUNLOCK(ifp); /* * XXX: For compat reasons only. * We should delay the job to slowpath queue. */ if (do_arp != 0) arprequest(ifp, NULL, &dst, NULL); return (error); } /* * Resolve an IP address into an ethernet address. * On input: * ifp is the interface we use * is_gw != if @dst represents gateway to some destination * m is the mbuf. May be NULL if we don't have a packet. * dst is the next hop, * desten is where we want the address. * flags returns lle entry flags. * * On success, desten and flags are filled in and the function returns 0; * If the packet must be held pending resolution, we return EWOULDBLOCK * On other errors, we return the corresponding error code. * Note that m_freem() handles NULL. */ int arpresolve(struct ifnet *ifp, int is_gw, struct mbuf *m, const struct sockaddr *dst, u_char *desten, uint32_t *pflags) { struct llentry *la = 0; u_int flags = 0; struct mbuf *curr = NULL; struct mbuf *next = NULL; int error, renew; if (pflags != NULL) *pflags = 0; if (m != NULL) { if (m->m_flags & M_BCAST) { /* broadcast */ (void)memcpy(desten, ifp->if_broadcastaddr, ifp->if_addrlen); return (0); } if (m->m_flags & M_MCAST) { /* multicast */ ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten); return (0); } } retry: IF_AFDATA_RLOCK(ifp); la = lla_lookup(LLTABLE(ifp), flags, dst); IF_AFDATA_RUNLOCK(ifp); if ((la == NULL) && ((flags & LLE_EXCLUSIVE) == 0) && ((ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) == 0)) { flags |= (LLE_CREATE | LLE_EXCLUSIVE); IF_AFDATA_WLOCK(ifp); la = lla_lookup(LLTABLE(ifp), flags, dst); IF_AFDATA_WUNLOCK(ifp); } if (la == NULL) { if (flags & LLE_CREATE) log(LOG_DEBUG, "arpresolve: can't allocate llinfo for %s on %s\n", inet_ntoa(SIN(dst)->sin_addr), ifp->if_xname); m_freem(m); return (EINVAL); } if ((la->la_flags & LLE_VALID) && ((la->la_flags & LLE_STATIC) || la->la_expire > time_uptime)) { bcopy(&la->ll_addr, desten, ifp->if_addrlen); renew = 0; /* * If entry has an expiry time and it is approaching, * see if we need to send an ARP request within this * arpt_down interval. */ if (!(la->la_flags & LLE_STATIC) && time_uptime + la->la_preempt > la->la_expire) { renew = 1; la->la_preempt--; } if (pflags != NULL) *pflags = la->la_flags; if (flags & LLE_EXCLUSIVE) LLE_WUNLOCK(la); else LLE_RUNLOCK(la); if (renew == 1) arprequest(ifp, NULL, &SIN(dst)->sin_addr, NULL); return (0); } if (la->la_flags & LLE_STATIC) { /* should not happen! */ log(LOG_DEBUG, "arpresolve: ouch, empty static llinfo for %s\n", inet_ntoa(SIN(dst)->sin_addr)); m_freem(m); error = EINVAL; goto done; } renew = (la->la_asked == 0 || la->la_expire != time_uptime); if ((renew || m != NULL) && (flags & LLE_EXCLUSIVE) == 0) { flags |= LLE_EXCLUSIVE; LLE_RUNLOCK(la); goto retry; } /* * There is an arptab entry, but no ethernet address * response yet. Add the mbuf to the list, dropping * the oldest packet if we have exceeded the system * setting. */ if (m != NULL) { if (la->la_numheld >= V_arp_maxhold) { if (la->la_hold != NULL) { next = la->la_hold->m_nextpkt; m_freem(la->la_hold); la->la_hold = next; la->la_numheld--; ARPSTAT_INC(dropped); } } if (la->la_hold != NULL) { curr = la->la_hold; while (curr->m_nextpkt != NULL) curr = curr->m_nextpkt; curr->m_nextpkt = m; } else la->la_hold = m; la->la_numheld++; if (renew == 0 && (flags & LLE_EXCLUSIVE)) { flags &= ~LLE_EXCLUSIVE; LLE_DOWNGRADE(la); } } /* * Return EWOULDBLOCK if we have tried less than arp_maxtries. It * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH * if we have already sent arp_maxtries ARP requests. Retransmit the * ARP request, but not faster than one request per second. */ if (la->la_asked < V_arp_maxtries) error = EWOULDBLOCK; /* First request. */ else error = is_gw != 0 ? EHOSTUNREACH : EHOSTDOWN; if (renew) { int canceled; LLE_ADDREF(la); la->la_expire = time_uptime; canceled = callout_reset(&la->la_timer, hz * V_arpt_down, arptimer, la); if (canceled) LLE_REMREF(la); la->la_asked++; LLE_WUNLOCK(la); arprequest(ifp, NULL, &SIN(dst)->sin_addr, NULL); return (error); } done: if (flags & LLE_EXCLUSIVE) LLE_WUNLOCK(la); else LLE_RUNLOCK(la); return (error); } /* * Common length and type checks are done here, * then the protocol-specific routine is called. */ static void arpintr(struct mbuf *m) { struct arphdr *ar; if (m->m_len < sizeof(struct arphdr) && ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) { log(LOG_NOTICE, "arp: runt packet -- m_pullup failed\n"); return; } ar = mtod(m, struct arphdr *); if (ntohs(ar->ar_hrd) != ARPHRD_ETHER && ntohs(ar->ar_hrd) != ARPHRD_IEEE802 && ntohs(ar->ar_hrd) != ARPHRD_ARCNET && ntohs(ar->ar_hrd) != ARPHRD_IEEE1394 && ntohs(ar->ar_hrd) != ARPHRD_INFINIBAND) { log(LOG_NOTICE, "arp: unknown hardware address format (0x%2D)" " (from %*D to %*D)\n", (unsigned char *)&ar->ar_hrd, "", ETHER_ADDR_LEN, (u_char *)ar_sha(ar), ":", ETHER_ADDR_LEN, (u_char *)ar_tha(ar), ":"); m_freem(m); return; } if (m->m_len < arphdr_len(ar)) { if ((m = m_pullup(m, arphdr_len(ar))) == NULL) { log(LOG_NOTICE, "arp: runt packet\n"); m_freem(m); return; } ar = mtod(m, struct arphdr *); } ARPSTAT_INC(received); switch (ntohs(ar->ar_pro)) { #ifdef INET case ETHERTYPE_IP: in_arpinput(m); return; #endif } m_freem(m); } #ifdef INET /* * ARP for Internet protocols on 10 Mb/s Ethernet. * Algorithm is that given in RFC 826. * In addition, a sanity check is performed on the sender * protocol address, to catch impersonators. * We no longer handle negotiations for use of trailer protocol: * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent * along with IP replies if we wanted trailers sent to us, * and also sent them in response to IP replies. * This allowed either end to announce the desire to receive * trailer packets. * We no longer reply to requests for ETHERTYPE_TRAIL protocol either, * but formerly didn't normally send requests. */ static int log_arp_wrong_iface = 1; static int log_arp_movements = 1; static int log_arp_permanent_modify = 1; static int allow_multicast = 0; static struct timeval arp_lastlog; static int arp_curpps; static int arp_maxpps = 1; SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW, &log_arp_wrong_iface, 0, "log arp packets arriving on the wrong interface"); SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW, &log_arp_movements, 0, "log arp replies from MACs different than the one in the cache"); SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_permanent_modify, CTLFLAG_RW, &log_arp_permanent_modify, 0, "log arp replies from MACs different than the one in the permanent arp entry"); SYSCTL_INT(_net_link_ether_inet, OID_AUTO, allow_multicast, CTLFLAG_RW, &allow_multicast, 0, "accept multicast addresses"); SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_log_per_second, CTLFLAG_RW, &arp_maxpps, 0, "Maximum number of remotely triggered ARP messages that can be " "logged per second"); #define ARP_LOG(pri, ...) do { \ if (ppsratecheck(&arp_lastlog, &arp_curpps, arp_maxpps)) \ log((pri), "arp: " __VA_ARGS__); \ } while (0) static void in_arpinput(struct mbuf *m) { struct rm_priotracker in_ifa_tracker; struct arphdr *ah; struct ifnet *ifp = m->m_pkthdr.rcvif; struct llentry *la = NULL; struct ifaddr *ifa; struct in_ifaddr *ia; struct sockaddr sa; struct in_addr isaddr, itaddr, myaddr; u_int8_t *enaddr = NULL; int op, flags; int req_len; int bridged = 0, is_bridge = 0; int carped; struct nhop4_extended nh_ext; struct sockaddr_in sin; sin.sin_len = sizeof(struct sockaddr_in); sin.sin_family = AF_INET; sin.sin_addr.s_addr = 0; if (ifp->if_bridge) bridged = 1; if (ifp->if_type == IFT_BRIDGE) is_bridge = 1; req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr)); if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) { ARP_LOG(LOG_NOTICE, "runt packet -- m_pullup failed\n"); return; } ah = mtod(m, struct arphdr *); /* * ARP is only for IPv4 so we can reject packets with * a protocol length not equal to an IPv4 address. */ if (ah->ar_pln != sizeof(struct in_addr)) { ARP_LOG(LOG_NOTICE, "requested protocol length != %zu\n", sizeof(struct in_addr)); goto drop; } if (allow_multicast == 0 && ETHER_IS_MULTICAST(ar_sha(ah))) { ARP_LOG(LOG_NOTICE, "%*D is multicast\n", ifp->if_addrlen, (u_char *)ar_sha(ah), ":"); goto drop; } op = ntohs(ah->ar_op); (void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr)); (void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr)); if (op == ARPOP_REPLY) ARPSTAT_INC(rxreplies); /* * For a bridge, we want to check the address irrespective * of the receive interface. (This will change slightly * when we have clusters of interfaces). */ IN_IFADDR_RLOCK(&in_ifa_tracker); LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) { if (((bridged && ia->ia_ifp->if_bridge == ifp->if_bridge) || ia->ia_ifp == ifp) && itaddr.s_addr == ia->ia_addr.sin_addr.s_addr && (ia->ia_ifa.ifa_carp == NULL || (*carp_iamatch_p)(&ia->ia_ifa, &enaddr))) { ifa_ref(&ia->ia_ifa); IN_IFADDR_RUNLOCK(&in_ifa_tracker); goto match; } } LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash) if (((bridged && ia->ia_ifp->if_bridge == ifp->if_bridge) || ia->ia_ifp == ifp) && isaddr.s_addr == ia->ia_addr.sin_addr.s_addr) { ifa_ref(&ia->ia_ifa); IN_IFADDR_RUNLOCK(&in_ifa_tracker); goto match; } #define BDG_MEMBER_MATCHES_ARP(addr, ifp, ia) \ (ia->ia_ifp->if_bridge == ifp->if_softc && \ !bcmp(IF_LLADDR(ia->ia_ifp), IF_LLADDR(ifp), ifp->if_addrlen) && \ addr == ia->ia_addr.sin_addr.s_addr) /* * Check the case when bridge shares its MAC address with * some of its children, so packets are claimed by bridge * itself (bridge_input() does it first), but they are really * meant to be destined to the bridge member. */ if (is_bridge) { LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) { if (BDG_MEMBER_MATCHES_ARP(itaddr.s_addr, ifp, ia)) { ifa_ref(&ia->ia_ifa); ifp = ia->ia_ifp; IN_IFADDR_RUNLOCK(&in_ifa_tracker); goto match; } } } #undef BDG_MEMBER_MATCHES_ARP IN_IFADDR_RUNLOCK(&in_ifa_tracker); /* * No match, use the first inet address on the receive interface * as a dummy address for the rest of the function. */ IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) if (ifa->ifa_addr->sa_family == AF_INET && (ifa->ifa_carp == NULL || (*carp_iamatch_p)(ifa, &enaddr))) { ia = ifatoia(ifa); ifa_ref(ifa); IF_ADDR_RUNLOCK(ifp); goto match; } IF_ADDR_RUNLOCK(ifp); /* * If bridging, fall back to using any inet address. */ IN_IFADDR_RLOCK(&in_ifa_tracker); if (!bridged || (ia = TAILQ_FIRST(&V_in_ifaddrhead)) == NULL) { IN_IFADDR_RUNLOCK(&in_ifa_tracker); goto drop; } ifa_ref(&ia->ia_ifa); IN_IFADDR_RUNLOCK(&in_ifa_tracker); match: if (!enaddr) enaddr = (u_int8_t *)IF_LLADDR(ifp); carped = (ia->ia_ifa.ifa_carp != NULL); myaddr = ia->ia_addr.sin_addr; ifa_free(&ia->ia_ifa); if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen)) goto drop; /* it's from me, ignore it. */ if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) { ARP_LOG(LOG_NOTICE, "link address is broadcast for IP address " "%s!\n", inet_ntoa(isaddr)); goto drop; } /* * Warn if another host is using the same IP address, but only if the * IP address isn't 0.0.0.0, which is used for DHCP only, in which * case we suppress the warning to avoid false positive complaints of * potential misconfiguration. */ if (!bridged && !carped && isaddr.s_addr == myaddr.s_addr && myaddr.s_addr != 0) { ARP_LOG(LOG_ERR, "%*D is using my IP address %s on %s!\n", ifp->if_addrlen, (u_char *)ar_sha(ah), ":", inet_ntoa(isaddr), ifp->if_xname); itaddr = myaddr; ARPSTAT_INC(dupips); goto reply; } if (ifp->if_flags & IFF_STATICARP) goto reply; bzero(&sin, sizeof(sin)); sin.sin_len = sizeof(struct sockaddr_in); sin.sin_family = AF_INET; sin.sin_addr = isaddr; flags = (itaddr.s_addr == myaddr.s_addr) ? LLE_CREATE : 0; flags |= LLE_EXCLUSIVE; IF_AFDATA_LOCK(ifp); la = lla_lookup(LLTABLE(ifp), flags, (struct sockaddr *)&sin); IF_AFDATA_UNLOCK(ifp); if (la != NULL) { /* the following is not an error when doing bridging */ if (!bridged && la->lle_tbl->llt_ifp != ifp) { if (log_arp_wrong_iface) ARP_LOG(LOG_WARNING, "%s is on %s " "but got reply from %*D on %s\n", inet_ntoa(isaddr), la->lle_tbl->llt_ifp->if_xname, ifp->if_addrlen, (u_char *)ar_sha(ah), ":", ifp->if_xname); LLE_WUNLOCK(la); goto reply; } if ((la->la_flags & LLE_VALID) && bcmp(ar_sha(ah), &la->ll_addr, ifp->if_addrlen)) { if (la->la_flags & LLE_STATIC) { LLE_WUNLOCK(la); if (log_arp_permanent_modify) ARP_LOG(LOG_ERR, "%*D attempts to modify " "permanent entry for %s on %s\n", ifp->if_addrlen, (u_char *)ar_sha(ah), ":", inet_ntoa(isaddr), ifp->if_xname); goto reply; } if (log_arp_movements) { ARP_LOG(LOG_INFO, "%s moved from %*D " "to %*D on %s\n", inet_ntoa(isaddr), ifp->if_addrlen, (u_char *)&la->ll_addr, ":", ifp->if_addrlen, (u_char *)ar_sha(ah), ":", ifp->if_xname); } } if (ifp->if_addrlen != ah->ar_hln) { LLE_WUNLOCK(la); ARP_LOG(LOG_WARNING, "from %*D: addr len: new %d, " "i/f %d (ignored)\n", ifp->if_addrlen, (u_char *) ar_sha(ah), ":", ah->ar_hln, ifp->if_addrlen); goto drop; } (void)memcpy(&la->ll_addr, ar_sha(ah), ifp->if_addrlen); la->la_flags |= LLE_VALID; EVENTHANDLER_INVOKE(lle_event, la, LLENTRY_RESOLVED); if (!(la->la_flags & LLE_STATIC)) { int canceled; LLE_ADDREF(la); la->la_expire = time_uptime + V_arpt_keep; canceled = callout_reset(&la->la_timer, hz * V_arpt_keep, arptimer, la); if (canceled) LLE_REMREF(la); } la->la_asked = 0; la->la_preempt = V_arp_maxtries; /* * The packets are all freed within the call to the output * routine. * * NB: The lock MUST be released before the call to the * output routine. */ if (la->la_hold != NULL) { struct mbuf *m_hold, *m_hold_next; m_hold = la->la_hold; la->la_hold = NULL; la->la_numheld = 0; memcpy(&sa, L3_ADDR(la), sizeof(sa)); LLE_WUNLOCK(la); for (; m_hold != NULL; m_hold = m_hold_next) { m_hold_next = m_hold->m_nextpkt; m_hold->m_nextpkt = NULL; /* Avoid confusing lower layers. */ m_clrprotoflags(m_hold); (*ifp->if_output)(ifp, m_hold, &sa, NULL); } } else LLE_WUNLOCK(la); } reply: if (op != ARPOP_REQUEST) goto drop; ARPSTAT_INC(rxrequests); if (itaddr.s_addr == myaddr.s_addr) { /* Shortcut.. the receiving interface is the target. */ (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); (void)memcpy(ar_sha(ah), enaddr, ah->ar_hln); } else { struct llentry *lle = NULL; sin.sin_addr = itaddr; IF_AFDATA_RLOCK(ifp); lle = lla_lookup(LLTABLE(ifp), 0, (struct sockaddr *)&sin); IF_AFDATA_RUNLOCK(ifp); if ((lle != NULL) && (lle->la_flags & LLE_PUB)) { (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); (void)memcpy(ar_sha(ah), &lle->ll_addr, ah->ar_hln); LLE_RUNLOCK(lle); } else { if (lle != NULL) LLE_RUNLOCK(lle); if (!V_arp_proxyall) goto drop; sin.sin_addr = itaddr; /* XXX MRT use table 0 for arp reply */ - if (fib4_lookup_nh_extended(0, itaddr, 0, &nh_ext) != 0) + if (fib4_lookup_nh_ext(0, itaddr, 0, 0, &nh_ext) != 0) goto drop; /* * Don't send proxies for nodes on the same interface * as this one came out of, or we'll get into a fight * over who claims what Ether address. */ if (nh_ext.nh_ifp == ifp) goto drop; (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); (void)memcpy(ar_sha(ah), enaddr, ah->ar_hln); /* * Also check that the node which sent the ARP packet * is on the interface we expect it to be on. This * avoids ARP chaos if an interface is connected to the * wrong network. */ sin.sin_addr = isaddr; /* XXX MRT use table 0 for arp checks */ - if (fib4_lookup_nh_extended(0, isaddr, 0, &nh_ext) != 0) + if (fib4_lookup_nh_ext(0, isaddr, 0, 0, &nh_ext) != 0) goto drop; if (nh_ext.nh_ifp != ifp) { ARP_LOG(LOG_INFO, "proxy: ignoring request" " from %s via wrong interface %s\n", inet_ntoa(isaddr), ifp->if_xname); goto drop; } #ifdef DEBUG_PROXY printf("arp: proxying for %s\n", inet_ntoa(itaddr)); #endif } } if (itaddr.s_addr == myaddr.s_addr && IN_LINKLOCAL(ntohl(itaddr.s_addr))) { /* RFC 3927 link-local IPv4; always reply by broadcast. */ #ifdef DEBUG_LINKLOCAL printf("arp: sending reply for link-local addr %s\n", inet_ntoa(itaddr)); #endif m->m_flags |= M_BCAST; m->m_flags &= ~M_MCAST; } else { /* default behaviour; never reply by broadcast. */ m->m_flags &= ~(M_BCAST|M_MCAST); } (void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln); (void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln); ah->ar_op = htons(ARPOP_REPLY); ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */ m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln); m->m_pkthdr.len = m->m_len; m->m_pkthdr.rcvif = NULL; sa.sa_family = AF_ARP; sa.sa_len = 2; m_clrprotoflags(m); /* Avoid confusing lower layers. */ (*ifp->if_output)(ifp, m, &sa, NULL); ARPSTAT_INC(txreplies); return; drop: m_freem(m); } #endif void arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa) { struct llentry *lle; if (ifa->ifa_carp != NULL) return; if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY) { arprequest(ifp, &IA_SIN(ifa)->sin_addr, &IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp)); /* * interface address is considered static entry * because the output of the arp utility shows * that L2 entry as permanent */ IF_AFDATA_LOCK(ifp); lle = lla_lookup(LLTABLE(ifp), (LLE_CREATE | LLE_IFADDR | LLE_STATIC), (struct sockaddr *)IA_SIN(ifa)); IF_AFDATA_UNLOCK(ifp); if (lle == NULL) log(LOG_INFO, "arp_ifinit: cannot create arp " "entry for interface address\n"); else LLE_RUNLOCK(lle); } ifa->ifa_rtrequest = NULL; } void arp_ifinit2(struct ifnet *ifp, struct ifaddr *ifa, u_char *enaddr) { if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY) arprequest(ifp, &IA_SIN(ifa)->sin_addr, &IA_SIN(ifa)->sin_addr, enaddr); ifa->ifa_rtrequest = NULL; } static void arp_init(void) { netisr_register(&arp_nh); } SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0); Index: projects/routing/sys/netinet/in_pcb.c =================================================================== --- projects/routing/sys/netinet/in_pcb.c (revision 287053) +++ projects/routing/sys/netinet/in_pcb.c (revision 287054) @@ -1,2628 +1,2629 @@ /*- * Copyright (c) 1982, 1986, 1991, 1993, 1995 * The Regents of the University of California. * Copyright (c) 2007-2009 Robert N. M. Watson * Copyright (c) 2010-2011 Juniper Networks, Inc. * All rights reserved. * * Portions of this software were developed by Robert N. M. Watson under * contract to Juniper Networks, Inc. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)in_pcb.c 8.4 (Berkeley) 5/24/95 */ #include __FBSDID("$FreeBSD$"); #include "opt_ddb.h" #include "opt_ipsec.h" #include "opt_inet.h" #include "opt_inet6.h" #include "opt_pcbgroup.h" #include "opt_rss.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef DDB #include #endif #include #include #include #include #include #include #include #if defined(INET) || defined(INET6) #include #include #include #include #include #include #endif #ifdef INET #include #endif #ifdef INET6 #include #include #include #include #endif /* INET6 */ #include #ifdef IPSEC #include #include #endif /* IPSEC */ #include static struct callout ipport_tick_callout; /* * These configure the range of local port addresses assigned to * "unspecified" outgoing connections/packets/whatever. */ VNET_DEFINE(int, ipport_lowfirstauto) = IPPORT_RESERVED - 1; /* 1023 */ VNET_DEFINE(int, ipport_lowlastauto) = IPPORT_RESERVEDSTART; /* 600 */ VNET_DEFINE(int, ipport_firstauto) = IPPORT_EPHEMERALFIRST; /* 10000 */ VNET_DEFINE(int, ipport_lastauto) = IPPORT_EPHEMERALLAST; /* 65535 */ VNET_DEFINE(int, ipport_hifirstauto) = IPPORT_HIFIRSTAUTO; /* 49152 */ VNET_DEFINE(int, ipport_hilastauto) = IPPORT_HILASTAUTO; /* 65535 */ /* * Reserved ports accessible only to root. There are significant * security considerations that must be accounted for when changing these, * but the security benefits can be great. Please be careful. */ VNET_DEFINE(int, ipport_reservedhigh) = IPPORT_RESERVED - 1; /* 1023 */ VNET_DEFINE(int, ipport_reservedlow); /* Variables dealing with random ephemeral port allocation. */ VNET_DEFINE(int, ipport_randomized) = 1; /* user controlled via sysctl */ VNET_DEFINE(int, ipport_randomcps) = 10; /* user controlled via sysctl */ VNET_DEFINE(int, ipport_randomtime) = 45; /* user controlled via sysctl */ VNET_DEFINE(int, ipport_stoprandom); /* toggled by ipport_tick */ VNET_DEFINE(int, ipport_tcpallocs); static VNET_DEFINE(int, ipport_tcplastcount); #define V_ipport_tcplastcount VNET(ipport_tcplastcount) static void in_pcbremlists(struct inpcb *inp); #ifdef INET static struct inpcb *in_pcblookup_hash_locked(struct inpcbinfo *pcbinfo, struct in_addr faddr, u_int fport_arg, struct in_addr laddr, u_int lport_arg, int lookupflags, struct ifnet *ifp); #define RANGECHK(var, min, max) \ if ((var) < (min)) { (var) = (min); } \ else if ((var) > (max)) { (var) = (max); } static int sysctl_net_ipport_check(SYSCTL_HANDLER_ARGS) { int error; error = sysctl_handle_int(oidp, arg1, arg2, req); if (error == 0) { RANGECHK(V_ipport_lowfirstauto, 1, IPPORT_RESERVED - 1); RANGECHK(V_ipport_lowlastauto, 1, IPPORT_RESERVED - 1); RANGECHK(V_ipport_firstauto, IPPORT_RESERVED, IPPORT_MAX); RANGECHK(V_ipport_lastauto, IPPORT_RESERVED, IPPORT_MAX); RANGECHK(V_ipport_hifirstauto, IPPORT_RESERVED, IPPORT_MAX); RANGECHK(V_ipport_hilastauto, IPPORT_RESERVED, IPPORT_MAX); } return (error); } #undef RANGECHK static SYSCTL_NODE(_net_inet_ip, IPPROTO_IP, portrange, CTLFLAG_RW, 0, "IP Ports"); SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowfirst, CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW, &VNET_NAME(ipport_lowfirstauto), 0, &sysctl_net_ipport_check, "I", ""); SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowlast, CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW, &VNET_NAME(ipport_lowlastauto), 0, &sysctl_net_ipport_check, "I", ""); SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, first, CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW, &VNET_NAME(ipport_firstauto), 0, &sysctl_net_ipport_check, "I", ""); SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, last, CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW, &VNET_NAME(ipport_lastauto), 0, &sysctl_net_ipport_check, "I", ""); SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hifirst, CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW, &VNET_NAME(ipport_hifirstauto), 0, &sysctl_net_ipport_check, "I", ""); SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hilast, CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW, &VNET_NAME(ipport_hilastauto), 0, &sysctl_net_ipport_check, "I", ""); SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, reservedhigh, CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE, &VNET_NAME(ipport_reservedhigh), 0, ""); SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, reservedlow, CTLFLAG_RW|CTLFLAG_SECURE, &VNET_NAME(ipport_reservedlow), 0, ""); SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, randomized, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipport_randomized), 0, "Enable random port allocation"); SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, randomcps, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipport_randomcps), 0, "Maximum number of random port " "allocations before switching to a sequental one"); SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, randomtime, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipport_randomtime), 0, "Minimum time to keep sequental port " "allocation before switching to a random one"); #endif /* INET */ /* * in_pcb.c: manage the Protocol Control Blocks. * * NOTE: It is assumed that most of these functions will be called with * the pcbinfo lock held, and often, the inpcb lock held, as these utility * functions often modify hash chains or addresses in pcbs. */ /* * Initialize an inpcbinfo -- we should be able to reduce the number of * arguments in time. */ void in_pcbinfo_init(struct inpcbinfo *pcbinfo, const char *name, struct inpcbhead *listhead, int hash_nelements, int porthash_nelements, char *inpcbzone_name, uma_init inpcbzone_init, uma_fini inpcbzone_fini, uint32_t inpcbzone_flags, u_int hashfields) { INP_INFO_LOCK_INIT(pcbinfo, name); INP_HASH_LOCK_INIT(pcbinfo, "pcbinfohash"); /* XXXRW: argument? */ #ifdef VIMAGE pcbinfo->ipi_vnet = curvnet; #endif pcbinfo->ipi_listhead = listhead; LIST_INIT(pcbinfo->ipi_listhead); pcbinfo->ipi_count = 0; pcbinfo->ipi_hashbase = hashinit(hash_nelements, M_PCB, &pcbinfo->ipi_hashmask); pcbinfo->ipi_porthashbase = hashinit(porthash_nelements, M_PCB, &pcbinfo->ipi_porthashmask); #ifdef PCBGROUP in_pcbgroup_init(pcbinfo, hashfields, hash_nelements); #endif pcbinfo->ipi_zone = uma_zcreate(inpcbzone_name, sizeof(struct inpcb), NULL, NULL, inpcbzone_init, inpcbzone_fini, UMA_ALIGN_PTR, inpcbzone_flags); uma_zone_set_max(pcbinfo->ipi_zone, maxsockets); uma_zone_set_warning(pcbinfo->ipi_zone, "kern.ipc.maxsockets limit reached"); } /* * Destroy an inpcbinfo. */ void in_pcbinfo_destroy(struct inpcbinfo *pcbinfo) { KASSERT(pcbinfo->ipi_count == 0, ("%s: ipi_count = %u", __func__, pcbinfo->ipi_count)); hashdestroy(pcbinfo->ipi_hashbase, M_PCB, pcbinfo->ipi_hashmask); hashdestroy(pcbinfo->ipi_porthashbase, M_PCB, pcbinfo->ipi_porthashmask); #ifdef PCBGROUP in_pcbgroup_destroy(pcbinfo); #endif uma_zdestroy(pcbinfo->ipi_zone); INP_HASH_LOCK_DESTROY(pcbinfo); INP_INFO_LOCK_DESTROY(pcbinfo); } /* * Allocate a PCB and associate it with the socket. * On success return with the PCB locked. */ int in_pcballoc(struct socket *so, struct inpcbinfo *pcbinfo) { struct inpcb *inp; int error; INP_INFO_WLOCK_ASSERT(pcbinfo); error = 0; inp = uma_zalloc(pcbinfo->ipi_zone, M_NOWAIT); if (inp == NULL) return (ENOBUFS); bzero(inp, inp_zero_size); inp->inp_pcbinfo = pcbinfo; inp->inp_socket = so; inp->inp_cred = crhold(so->so_cred); inp->inp_inc.inc_fibnum = so->so_fibnum; #ifdef MAC error = mac_inpcb_init(inp, M_NOWAIT); if (error != 0) goto out; mac_inpcb_create(so, inp); #endif #ifdef IPSEC error = ipsec_init_policy(so, &inp->inp_sp); if (error != 0) { #ifdef MAC mac_inpcb_destroy(inp); #endif goto out; } #endif /*IPSEC*/ #ifdef INET6 if (INP_SOCKAF(so) == AF_INET6) { inp->inp_vflag |= INP_IPV6PROTO; if (V_ip6_v6only) inp->inp_flags |= IN6P_IPV6_V6ONLY; } #endif LIST_INSERT_HEAD(pcbinfo->ipi_listhead, inp, inp_list); pcbinfo->ipi_count++; so->so_pcb = (caddr_t)inp; #ifdef INET6 if (V_ip6_auto_flowlabel) inp->inp_flags |= IN6P_AUTOFLOWLABEL; #endif INP_WLOCK(inp); inp->inp_gencnt = ++pcbinfo->ipi_gencnt; refcount_init(&inp->inp_refcount, 1); /* Reference from inpcbinfo */ #if defined(IPSEC) || defined(MAC) out: if (error != 0) { crfree(inp->inp_cred); uma_zfree(pcbinfo->ipi_zone, inp); } #endif return (error); } #ifdef INET int in_pcbbind(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred) { int anonport, error; INP_WLOCK_ASSERT(inp); INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo); if (inp->inp_lport != 0 || inp->inp_laddr.s_addr != INADDR_ANY) return (EINVAL); anonport = nam == NULL || ((struct sockaddr_in *)nam)->sin_port == 0; error = in_pcbbind_setup(inp, nam, &inp->inp_laddr.s_addr, &inp->inp_lport, cred); if (error) return (error); if (in_pcbinshash(inp) != 0) { inp->inp_laddr.s_addr = INADDR_ANY; inp->inp_lport = 0; return (EAGAIN); } if (anonport) inp->inp_flags |= INP_ANONPORT; return (0); } #endif /* * Select a local port (number) to use. */ #if defined(INET) || defined(INET6) int in_pcb_lport(struct inpcb *inp, struct in_addr *laddrp, u_short *lportp, struct ucred *cred, int lookupflags) { struct inpcbinfo *pcbinfo; struct inpcb *tmpinp; unsigned short *lastport; int count, dorandom, error; u_short aux, first, last, lport; #ifdef INET struct in_addr laddr; #endif pcbinfo = inp->inp_pcbinfo; /* * Because no actual state changes occur here, a global write lock on * the pcbinfo isn't required. */ INP_LOCK_ASSERT(inp); INP_HASH_LOCK_ASSERT(pcbinfo); if (inp->inp_flags & INP_HIGHPORT) { first = V_ipport_hifirstauto; /* sysctl */ last = V_ipport_hilastauto; lastport = &pcbinfo->ipi_lasthi; } else if (inp->inp_flags & INP_LOWPORT) { error = priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT, 0); if (error) return (error); first = V_ipport_lowfirstauto; /* 1023 */ last = V_ipport_lowlastauto; /* 600 */ lastport = &pcbinfo->ipi_lastlow; } else { first = V_ipport_firstauto; /* sysctl */ last = V_ipport_lastauto; lastport = &pcbinfo->ipi_lastport; } /* * For UDP(-Lite), use random port allocation as long as the user * allows it. For TCP (and as of yet unknown) connections, * use random port allocation only if the user allows it AND * ipport_tick() allows it. */ if (V_ipport_randomized && (!V_ipport_stoprandom || pcbinfo == &V_udbinfo || pcbinfo == &V_ulitecbinfo)) dorandom = 1; else dorandom = 0; /* * It makes no sense to do random port allocation if * we have the only port available. */ if (first == last) dorandom = 0; /* Make sure to not include UDP(-Lite) packets in the count. */ if (pcbinfo != &V_udbinfo || pcbinfo != &V_ulitecbinfo) V_ipport_tcpallocs++; /* * Instead of having two loops further down counting up or down * make sure that first is always <= last and go with only one * code path implementing all logic. */ if (first > last) { aux = first; first = last; last = aux; } #ifdef INET /* Make the compiler happy. */ laddr.s_addr = 0; if ((inp->inp_vflag & (INP_IPV4|INP_IPV6)) == INP_IPV4) { KASSERT(laddrp != NULL, ("%s: laddrp NULL for v4 inp %p", __func__, inp)); laddr = *laddrp; } #endif tmpinp = NULL; /* Make compiler happy. */ lport = *lportp; if (dorandom) *lastport = first + (arc4random() % (last - first)); count = last - first; do { if (count-- < 0) /* completely used? */ return (EADDRNOTAVAIL); ++*lastport; if (*lastport < first || *lastport > last) *lastport = first; lport = htons(*lastport); #ifdef INET6 if ((inp->inp_vflag & INP_IPV6) != 0) tmpinp = in6_pcblookup_local(pcbinfo, &inp->in6p_laddr, lport, lookupflags, cred); #endif #if defined(INET) && defined(INET6) else #endif #ifdef INET tmpinp = in_pcblookup_local(pcbinfo, laddr, lport, lookupflags, cred); #endif } while (tmpinp != NULL); #ifdef INET if ((inp->inp_vflag & (INP_IPV4|INP_IPV6)) == INP_IPV4) laddrp->s_addr = laddr.s_addr; #endif *lportp = lport; return (0); } /* * Return cached socket options. */ short inp_so_options(const struct inpcb *inp) { short so_options; so_options = 0; if ((inp->inp_flags2 & INP_REUSEPORT) != 0) so_options |= SO_REUSEPORT; if ((inp->inp_flags2 & INP_REUSEADDR) != 0) so_options |= SO_REUSEADDR; return (so_options); } #endif /* INET || INET6 */ /* * Check if a new BINDMULTI socket is allowed to be created. * * ni points to the new inp. * oi points to the exisitng inp. * * This checks whether the existing inp also has BINDMULTI and * whether the credentials match. */ int in_pcbbind_check_bindmulti(const struct inpcb *ni, const struct inpcb *oi) { /* Check permissions match */ if ((ni->inp_flags2 & INP_BINDMULTI) && (ni->inp_cred->cr_uid != oi->inp_cred->cr_uid)) return (0); /* Check the existing inp has BINDMULTI set */ if ((ni->inp_flags2 & INP_BINDMULTI) && ((oi->inp_flags2 & INP_BINDMULTI) == 0)) return (0); /* * We're okay - either INP_BINDMULTI isn't set on ni, or * it is and it matches the checks. */ return (1); } #ifdef INET /* * Set up a bind operation on a PCB, performing port allocation * as required, but do not actually modify the PCB. Callers can * either complete the bind by setting inp_laddr/inp_lport and * calling in_pcbinshash(), or they can just use the resulting * port and address to authorise the sending of a once-off packet. * * On error, the values of *laddrp and *lportp are not changed. */ int in_pcbbind_setup(struct inpcb *inp, struct sockaddr *nam, in_addr_t *laddrp, u_short *lportp, struct ucred *cred) { struct socket *so = inp->inp_socket; struct sockaddr_in *sin; struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; struct in_addr laddr; u_short lport = 0; int lookupflags = 0, reuseport = (so->so_options & SO_REUSEPORT); int error; /* * No state changes, so read locks are sufficient here. */ INP_LOCK_ASSERT(inp); INP_HASH_LOCK_ASSERT(pcbinfo); if (TAILQ_EMPTY(&V_in_ifaddrhead)) /* XXX broken! */ return (EADDRNOTAVAIL); laddr.s_addr = *laddrp; if (nam != NULL && laddr.s_addr != INADDR_ANY) return (EINVAL); if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0) lookupflags = INPLOOKUP_WILDCARD; if (nam == NULL) { if ((error = prison_local_ip4(cred, &laddr)) != 0) return (error); } else { sin = (struct sockaddr_in *)nam; if (nam->sa_len != sizeof (*sin)) return (EINVAL); #ifdef notdef /* * We should check the family, but old programs * incorrectly fail to initialize it. */ if (sin->sin_family != AF_INET) return (EAFNOSUPPORT); #endif error = prison_local_ip4(cred, &sin->sin_addr); if (error) return (error); if (sin->sin_port != *lportp) { /* Don't allow the port to change. */ if (*lportp != 0) return (EINVAL); lport = sin->sin_port; } /* NB: lport is left as 0 if the port isn't being changed. */ if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) { /* * Treat SO_REUSEADDR as SO_REUSEPORT for multicast; * allow complete duplication of binding if * SO_REUSEPORT is set, or if SO_REUSEADDR is set * and a multicast address is bound on both * new and duplicated sockets. */ if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) != 0) reuseport = SO_REUSEADDR|SO_REUSEPORT; } else if (sin->sin_addr.s_addr != INADDR_ANY) { sin->sin_port = 0; /* yech... */ bzero(&sin->sin_zero, sizeof(sin->sin_zero)); /* * Is the address a local IP address? * If INP_BINDANY is set, then the socket may be bound * to any endpoint address, local or not. */ if ((inp->inp_flags & INP_BINDANY) == 0 && ifa_ifwithaddr_check((struct sockaddr *)sin) == 0) return (EADDRNOTAVAIL); } laddr = sin->sin_addr; if (lport) { struct inpcb *t; struct tcptw *tw; /* GROSS */ if (ntohs(lport) <= V_ipport_reservedhigh && ntohs(lport) >= V_ipport_reservedlow && priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT, 0)) return (EACCES); if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)) && priv_check_cred(inp->inp_cred, PRIV_NETINET_REUSEPORT, 0) != 0) { t = in_pcblookup_local(pcbinfo, sin->sin_addr, lport, INPLOOKUP_WILDCARD, cred); /* * XXX * This entire block sorely needs a rewrite. */ if (t && ((inp->inp_flags2 & INP_BINDMULTI) == 0) && ((t->inp_flags & INP_TIMEWAIT) == 0) && (so->so_type != SOCK_STREAM || ntohl(t->inp_faddr.s_addr) == INADDR_ANY) && (ntohl(sin->sin_addr.s_addr) != INADDR_ANY || ntohl(t->inp_laddr.s_addr) != INADDR_ANY || (t->inp_flags2 & INP_REUSEPORT) == 0) && (inp->inp_cred->cr_uid != t->inp_cred->cr_uid)) return (EADDRINUSE); /* * If the socket is a BINDMULTI socket, then * the credentials need to match and the * original socket also has to have been bound * with BINDMULTI. */ if (t && (! in_pcbbind_check_bindmulti(inp, t))) return (EADDRINUSE); } t = in_pcblookup_local(pcbinfo, sin->sin_addr, lport, lookupflags, cred); if (t && (t->inp_flags & INP_TIMEWAIT)) { /* * XXXRW: If an incpb has had its timewait * state recycled, we treat the address as * being in use (for now). This is better * than a panic, but not desirable. */ tw = intotw(t); if (tw == NULL || (reuseport & tw->tw_so_options) == 0) return (EADDRINUSE); } else if (t && ((inp->inp_flags2 & INP_BINDMULTI) == 0) && (reuseport & inp_so_options(t)) == 0) { #ifdef INET6 if (ntohl(sin->sin_addr.s_addr) != INADDR_ANY || ntohl(t->inp_laddr.s_addr) != INADDR_ANY || (inp->inp_vflag & INP_IPV6PROTO) == 0 || (t->inp_vflag & INP_IPV6PROTO) == 0) #endif return (EADDRINUSE); if (t && (! in_pcbbind_check_bindmulti(inp, t))) return (EADDRINUSE); } } } if (*lportp != 0) lport = *lportp; if (lport == 0) { error = in_pcb_lport(inp, &laddr, &lport, cred, lookupflags); if (error != 0) return (error); } *laddrp = laddr.s_addr; *lportp = lport; return (0); } /* * Connect from a socket to a specified address. * Both address and port must be specified in argument sin. * If don't have a local address for this socket yet, * then pick one. */ int in_pcbconnect_mbuf(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred, struct mbuf *m) { u_short lport, fport; in_addr_t laddr, faddr; int anonport, error; INP_WLOCK_ASSERT(inp); INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo); lport = inp->inp_lport; laddr = inp->inp_laddr.s_addr; anonport = (lport == 0); error = in_pcbconnect_setup(inp, nam, &laddr, &lport, &faddr, &fport, NULL, cred); if (error) return (error); /* Do the initial binding of the local address if required. */ if (inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0) { inp->inp_lport = lport; inp->inp_laddr.s_addr = laddr; if (in_pcbinshash(inp) != 0) { inp->inp_laddr.s_addr = INADDR_ANY; inp->inp_lport = 0; return (EAGAIN); } } /* Commit the remaining changes. */ inp->inp_lport = lport; inp->inp_laddr.s_addr = laddr; inp->inp_faddr.s_addr = faddr; inp->inp_fport = fport; in_pcbrehash_mbuf(inp, m); if (anonport) inp->inp_flags |= INP_ANONPORT; return (0); } int in_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred) { return (in_pcbconnect_mbuf(inp, nam, cred, NULL)); } /* * Do proper source address selection on an unbound socket in case * of connect. Take jails into account as well. */ int in_pcbladdr(struct inpcb *inp, struct in_addr *faddr, struct in_addr *laddr, struct ucred *cred) { struct ifaddr *ifa; struct sockaddr *sa; struct sockaddr_in *sin, sin_storage; struct nhop4_extended nh_ext, *pnh4; u_int fibnum; int error; KASSERT(laddr != NULL, ("%s: laddr NULL", __func__)); /* * Bypass source address selection and use the primary jail IP * if requested. */ if (cred != NULL && !prison_saddrsel_ip4(cred, laddr)) return (0); error = 0; sin = &sin_storage; sin->sin_family = AF_INET; sin->sin_len = sizeof(struct sockaddr_in); sin->sin_addr.s_addr = faddr->s_addr; /* * If route is known our src addr is taken from the i/f, * else punt. * * Find out route to destination. */ fibnum = inp->inp_inc.inc_fibnum; pnh4 = &nh_ext; memset(&nh_ext, 0, sizeof(nh_ext)); if ((inp->inp_socket->so_options & SO_DONTROUTE) == 0) - error = fib4_lookup_nh_extended(fibnum, *faddr, 0, &nh_ext); + error = fib4_lookup_nh_ext(fibnum, *faddr, 0, NHOP_LOOKUP_REF, + &nh_ext); if (error != 0) { pnh4 = NULL; error = 0; } /* * If we found a route, use the address corresponding to * the outgoing interface. * * Otherwise assume faddr is reachable on a directly connected * network and try to find a corresponding interface to take * the source address from. */ if (pnh4 == NULL) { struct in_ifaddr *ia; struct ifnet *ifp; ia = ifatoia(ifa_ifwithdstaddr((struct sockaddr *)sin, inp->inp_socket->so_fibnum)); if (ia == NULL) ia = ifatoia(ifa_ifwithnet((struct sockaddr *)sin, 0, inp->inp_socket->so_fibnum)); if (ia == NULL) { error = ENETUNREACH; goto done; } if (cred == NULL || !prison_flag(cred, PR_IP4)) { laddr->s_addr = ia->ia_addr.sin_addr.s_addr; ifa_free(&ia->ia_ifa); goto done; } ifp = ia->ia_ifp; ifa_free(&ia->ia_ifa); ia = NULL; IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { sa = ifa->ifa_addr; if (sa->sa_family != AF_INET) continue; sin = (struct sockaddr_in *)sa; if (prison_check_ip4(cred, &sin->sin_addr) == 0) { ia = (struct in_ifaddr *)ifa; break; } } if (ia != NULL) { laddr->s_addr = ia->ia_addr.sin_addr.s_addr; IF_ADDR_RUNLOCK(ifp); goto done; } IF_ADDR_RUNLOCK(ifp); /* 3. As a last resort return the 'default' jail address. */ error = prison_get_ip4(cred, laddr); goto done; } /* * If the outgoing interface on the route found is not * a loopback interface, use the address from that interface. * In case of jails do those three steps: * 1. check if the interface address belongs to the jail. If so use it. * 2. check if we have any address on the outgoing interface * belonging to this jail. If so use it. * 3. as a last resort return the 'default' jail address. */ if ((nh_ext.nh_ifp->if_flags & IFF_LOOPBACK) == 0) { struct in_ifaddr *ia; struct ifnet *ifp; struct in_addr addr; /* If not jailed, use the default returned. */ if (cred == NULL || !prison_flag(cred, PR_IP4)) { laddr->s_addr = nh_ext.nh_src.s_addr; goto done; } /* Jailed. */ /* 1. Check if the iface address belongs to the jail. */ addr = nh_ext.nh_src; if (prison_check_ip4(cred, &addr) == 0) { laddr->s_addr = nh_ext.nh_src.s_addr; goto done; } /* * 2. Check if we have any address on the outgoing interface * belonging to this jail. */ ia = NULL; ifp = nh_ext.nh_ifp; IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { sa = ifa->ifa_addr; if (sa->sa_family != AF_INET) continue; sin = (struct sockaddr_in *)sa; if (prison_check_ip4(cred, &sin->sin_addr) == 0) { ia = (struct in_ifaddr *)ifa; break; } } if (ia != NULL) { laddr->s_addr = ia->ia_addr.sin_addr.s_addr; IF_ADDR_RUNLOCK(ifp); goto done; } IF_ADDR_RUNLOCK(ifp); /* 3. As a last resort return the 'default' jail address. */ error = prison_get_ip4(cred, laddr); goto done; } /* * The outgoing interface is marked with 'loopback net', so a route * to ourselves is here. * Try to find the interface of the destination address and then * take the address from there. That interface is not necessarily * a loopback interface. * In case of jails, check that it is an address of the jail * and if we cannot find, fall back to the 'default' jail address. */ if ((nh_ext.nh_ifp->if_flags & IFF_LOOPBACK) != 0) { struct sockaddr_in sain; struct in_ifaddr *ia; bzero(&sain, sizeof(struct sockaddr_in)); sain.sin_family = AF_INET; sain.sin_len = sizeof(struct sockaddr_in); sain.sin_addr.s_addr = faddr->s_addr; ia = ifatoia(ifa_ifwithdstaddr(sintosa(&sain), inp->inp_socket->so_fibnum)); if (ia == NULL) ia = ifatoia(ifa_ifwithnet(sintosa(&sain), 0, inp->inp_socket->so_fibnum)); if (ia == NULL) ia = ifatoia(ifa_ifwithaddr(sintosa(&sain))); if (cred == NULL || !prison_flag(cred, PR_IP4)) { if (ia == NULL) { error = ENETUNREACH; goto done; } laddr->s_addr = ia->ia_addr.sin_addr.s_addr; ifa_free(&ia->ia_ifa); goto done; } /* Jailed. */ if (ia != NULL) { struct ifnet *ifp; ifp = ia->ia_ifp; ifa_free(&ia->ia_ifa); ia = NULL; IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { sa = ifa->ifa_addr; if (sa->sa_family != AF_INET) continue; sin = (struct sockaddr_in *)sa; if (prison_check_ip4(cred, &sin->sin_addr) == 0) { ia = (struct in_ifaddr *)ifa; break; } } if (ia != NULL) { laddr->s_addr = ia->ia_addr.sin_addr.s_addr; IF_ADDR_RUNLOCK(ifp); goto done; } IF_ADDR_RUNLOCK(ifp); } /* 3. As a last resort return the 'default' jail address. */ error = prison_get_ip4(cred, laddr); goto done; } done: if (pnh4 != NULL) fib4_free_nh_ext(fibnum, pnh4); return (error); } /* * Set up for a connect from a socket to the specified address. * On entry, *laddrp and *lportp should contain the current local * address and port for the PCB; these are updated to the values * that should be placed in inp_laddr and inp_lport to complete * the connect. * * On success, *faddrp and *fportp will be set to the remote address * and port. These are not updated in the error case. * * If the operation fails because the connection already exists, * *oinpp will be set to the PCB of that connection so that the * caller can decide to override it. In all other cases, *oinpp * is set to NULL. */ int in_pcbconnect_setup(struct inpcb *inp, struct sockaddr *nam, in_addr_t *laddrp, u_short *lportp, in_addr_t *faddrp, u_short *fportp, struct inpcb **oinpp, struct ucred *cred) { struct rm_priotracker in_ifa_tracker; struct sockaddr_in *sin = (struct sockaddr_in *)nam; struct in_ifaddr *ia; struct inpcb *oinp; struct in_addr laddr, faddr; u_short lport, fport; int error; /* * Because a global state change doesn't actually occur here, a read * lock is sufficient. */ INP_LOCK_ASSERT(inp); INP_HASH_LOCK_ASSERT(inp->inp_pcbinfo); if (oinpp != NULL) *oinpp = NULL; if (nam->sa_len != sizeof (*sin)) return (EINVAL); if (sin->sin_family != AF_INET) return (EAFNOSUPPORT); if (sin->sin_port == 0) return (EADDRNOTAVAIL); laddr.s_addr = *laddrp; lport = *lportp; faddr = sin->sin_addr; fport = sin->sin_port; if (!TAILQ_EMPTY(&V_in_ifaddrhead)) { /* * If the destination address is INADDR_ANY, * use the primary local address. * If the supplied address is INADDR_BROADCAST, * and the primary interface supports broadcast, * choose the broadcast address for that interface. */ if (faddr.s_addr == INADDR_ANY) { IN_IFADDR_RLOCK(&in_ifa_tracker); faddr = IA_SIN(TAILQ_FIRST(&V_in_ifaddrhead))->sin_addr; IN_IFADDR_RUNLOCK(&in_ifa_tracker); if (cred != NULL && (error = prison_get_ip4(cred, &faddr)) != 0) return (error); } else if (faddr.s_addr == (u_long)INADDR_BROADCAST) { IN_IFADDR_RLOCK(&in_ifa_tracker); if (TAILQ_FIRST(&V_in_ifaddrhead)->ia_ifp->if_flags & IFF_BROADCAST) faddr = satosin(&TAILQ_FIRST( &V_in_ifaddrhead)->ia_broadaddr)->sin_addr; IN_IFADDR_RUNLOCK(&in_ifa_tracker); } } if (laddr.s_addr == INADDR_ANY) { error = in_pcbladdr(inp, &faddr, &laddr, cred); /* * If the destination address is multicast and an outgoing * interface has been set as a multicast option, prefer the * address of that interface as our source address. */ if (IN_MULTICAST(ntohl(faddr.s_addr)) && inp->inp_moptions != NULL) { struct ip_moptions *imo; struct ifnet *ifp; imo = inp->inp_moptions; if (imo->imo_multicast_ifp != NULL) { ifp = imo->imo_multicast_ifp; IN_IFADDR_RLOCK(&in_ifa_tracker); TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { if ((ia->ia_ifp == ifp) && (cred == NULL || prison_check_ip4(cred, &ia->ia_addr.sin_addr) == 0)) break; } if (ia == NULL) error = EADDRNOTAVAIL; else { laddr = ia->ia_addr.sin_addr; error = 0; } IN_IFADDR_RUNLOCK(&in_ifa_tracker); } } if (error) return (error); } oinp = in_pcblookup_hash_locked(inp->inp_pcbinfo, faddr, fport, laddr, lport, 0, NULL); if (oinp != NULL) { if (oinpp != NULL) *oinpp = oinp; return (EADDRINUSE); } if (lport == 0) { error = in_pcbbind_setup(inp, NULL, &laddr.s_addr, &lport, cred); if (error) return (error); } *laddrp = laddr.s_addr; *lportp = lport; *faddrp = faddr.s_addr; *fportp = fport; return (0); } void in_pcbdisconnect(struct inpcb *inp) { INP_WLOCK_ASSERT(inp); INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo); inp->inp_faddr.s_addr = INADDR_ANY; inp->inp_fport = 0; in_pcbrehash(inp); } #endif /* INET */ /* * in_pcbdetach() is responsibe for disassociating a socket from an inpcb. * For most protocols, this will be invoked immediately prior to calling * in_pcbfree(). However, with TCP the inpcb may significantly outlive the * socket, in which case in_pcbfree() is deferred. */ void in_pcbdetach(struct inpcb *inp) { KASSERT(inp->inp_socket != NULL, ("%s: inp_socket == NULL", __func__)); inp->inp_socket->so_pcb = NULL; inp->inp_socket = NULL; } /* * in_pcbref() bumps the reference count on an inpcb in order to maintain * stability of an inpcb pointer despite the inpcb lock being released. This * is used in TCP when the inpcbinfo lock needs to be acquired or upgraded, * but where the inpcb lock may already held, or when acquiring a reference * via a pcbgroup. * * in_pcbref() should be used only to provide brief memory stability, and * must always be followed by a call to INP_WLOCK() and in_pcbrele() to * garbage collect the inpcb if it has been in_pcbfree()'d from another * context. Until in_pcbrele() has returned that the inpcb is still valid, * lock and rele are the *only* safe operations that may be performed on the * inpcb. * * While the inpcb will not be freed, releasing the inpcb lock means that the * connection's state may change, so the caller should be careful to * revalidate any cached state on reacquiring the lock. Drop the reference * using in_pcbrele(). */ void in_pcbref(struct inpcb *inp) { KASSERT(inp->inp_refcount > 0, ("%s: refcount 0", __func__)); refcount_acquire(&inp->inp_refcount); } /* * Drop a refcount on an inpcb elevated using in_pcbref(); because a call to * in_pcbfree() may have been made between in_pcbref() and in_pcbrele(), we * return a flag indicating whether or not the inpcb remains valid. If it is * valid, we return with the inpcb lock held. * * Notice that, unlike in_pcbref(), the inpcb lock must be held to drop a * reference on an inpcb. Historically more work was done here (actually, in * in_pcbfree_internal()) but has been moved to in_pcbfree() to avoid the * need for the pcbinfo lock in in_pcbrele(). Deferring the free is entirely * about memory stability (and continued use of the write lock). */ int in_pcbrele_rlocked(struct inpcb *inp) { struct inpcbinfo *pcbinfo; KASSERT(inp->inp_refcount > 0, ("%s: refcount 0", __func__)); INP_RLOCK_ASSERT(inp); if (refcount_release(&inp->inp_refcount) == 0) { /* * If the inpcb has been freed, let the caller know, even if * this isn't the last reference. */ if (inp->inp_flags2 & INP_FREED) { INP_RUNLOCK(inp); return (1); } return (0); } KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL", __func__)); INP_RUNLOCK(inp); pcbinfo = inp->inp_pcbinfo; uma_zfree(pcbinfo->ipi_zone, inp); return (1); } int in_pcbrele_wlocked(struct inpcb *inp) { struct inpcbinfo *pcbinfo; KASSERT(inp->inp_refcount > 0, ("%s: refcount 0", __func__)); INP_WLOCK_ASSERT(inp); if (refcount_release(&inp->inp_refcount) == 0) return (0); KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL", __func__)); INP_WUNLOCK(inp); pcbinfo = inp->inp_pcbinfo; uma_zfree(pcbinfo->ipi_zone, inp); return (1); } /* * Temporary wrapper. */ int in_pcbrele(struct inpcb *inp) { return (in_pcbrele_wlocked(inp)); } /* * Unconditionally schedule an inpcb to be freed by decrementing its * reference count, which should occur only after the inpcb has been detached * from its socket. If another thread holds a temporary reference (acquired * using in_pcbref()) then the free is deferred until that reference is * released using in_pcbrele(), but the inpcb is still unlocked. Almost all * work, including removal from global lists, is done in this context, where * the pcbinfo lock is held. */ void in_pcbfree(struct inpcb *inp) { struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL", __func__)); INP_INFO_WLOCK_ASSERT(pcbinfo); INP_WLOCK_ASSERT(inp); /* XXXRW: Do as much as possible here. */ #ifdef IPSEC if (inp->inp_sp != NULL) ipsec_delete_pcbpolicy(inp); #endif inp->inp_gencnt = ++pcbinfo->ipi_gencnt; in_pcbremlists(inp); #ifdef INET6 if (inp->inp_vflag & INP_IPV6PROTO) { ip6_freepcbopts(inp->in6p_outputopts); if (inp->in6p_moptions != NULL) ip6_freemoptions(inp->in6p_moptions); } #endif if (inp->inp_options) (void)m_free(inp->inp_options); #ifdef INET if (inp->inp_moptions != NULL) inp_freemoptions(inp->inp_moptions); #endif inp->inp_vflag = 0; inp->inp_flags2 |= INP_FREED; crfree(inp->inp_cred); #ifdef MAC mac_inpcb_destroy(inp); #endif if (!in_pcbrele_wlocked(inp)) INP_WUNLOCK(inp); } /* * in_pcbdrop() removes an inpcb from hashed lists, releasing its address and * port reservation, and preventing it from being returned by inpcb lookups. * * It is used by TCP to mark an inpcb as unused and avoid future packet * delivery or event notification when a socket remains open but TCP has * closed. This might occur as a result of a shutdown()-initiated TCP close * or a RST on the wire, and allows the port binding to be reused while still * maintaining the invariant that so_pcb always points to a valid inpcb until * in_pcbdetach(). * * XXXRW: Possibly in_pcbdrop() should also prevent future notifications by * in_pcbnotifyall() and in_pcbpurgeif0()? */ void in_pcbdrop(struct inpcb *inp) { INP_WLOCK_ASSERT(inp); /* * XXXRW: Possibly we should protect the setting of INP_DROPPED with * the hash lock...? */ inp->inp_flags |= INP_DROPPED; if (inp->inp_flags & INP_INHASHLIST) { struct inpcbport *phd = inp->inp_phd; INP_HASH_WLOCK(inp->inp_pcbinfo); LIST_REMOVE(inp, inp_hash); LIST_REMOVE(inp, inp_portlist); if (LIST_FIRST(&phd->phd_pcblist) == NULL) { LIST_REMOVE(phd, phd_hash); free(phd, M_PCB); } INP_HASH_WUNLOCK(inp->inp_pcbinfo); inp->inp_flags &= ~INP_INHASHLIST; #ifdef PCBGROUP in_pcbgroup_remove(inp); #endif } } #ifdef INET /* * Common routines to return the socket addresses associated with inpcbs. */ struct sockaddr * in_sockaddr(in_port_t port, struct in_addr *addr_p) { struct sockaddr_in *sin; sin = malloc(sizeof *sin, M_SONAME, M_WAITOK | M_ZERO); sin->sin_family = AF_INET; sin->sin_len = sizeof(*sin); sin->sin_addr = *addr_p; sin->sin_port = port; return (struct sockaddr *)sin; } int in_getsockaddr(struct socket *so, struct sockaddr **nam) { struct inpcb *inp; struct in_addr addr; in_port_t port; inp = sotoinpcb(so); KASSERT(inp != NULL, ("in_getsockaddr: inp == NULL")); INP_RLOCK(inp); port = inp->inp_lport; addr = inp->inp_laddr; INP_RUNLOCK(inp); *nam = in_sockaddr(port, &addr); return 0; } int in_getpeeraddr(struct socket *so, struct sockaddr **nam) { struct inpcb *inp; struct in_addr addr; in_port_t port; inp = sotoinpcb(so); KASSERT(inp != NULL, ("in_getpeeraddr: inp == NULL")); INP_RLOCK(inp); port = inp->inp_fport; addr = inp->inp_faddr; INP_RUNLOCK(inp); *nam = in_sockaddr(port, &addr); return 0; } void in_pcbnotifyall(struct inpcbinfo *pcbinfo, struct in_addr faddr, int errno, struct inpcb *(*notify)(struct inpcb *, int)) { struct inpcb *inp, *inp_temp; INP_INFO_WLOCK(pcbinfo); LIST_FOREACH_SAFE(inp, pcbinfo->ipi_listhead, inp_list, inp_temp) { INP_WLOCK(inp); #ifdef INET6 if ((inp->inp_vflag & INP_IPV4) == 0) { INP_WUNLOCK(inp); continue; } #endif if (inp->inp_faddr.s_addr != faddr.s_addr || inp->inp_socket == NULL) { INP_WUNLOCK(inp); continue; } if ((*notify)(inp, errno)) INP_WUNLOCK(inp); } INP_INFO_WUNLOCK(pcbinfo); } void in_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp) { struct inpcb *inp; struct ip_moptions *imo; int i, gap; INP_INFO_RLOCK(pcbinfo); LIST_FOREACH(inp, pcbinfo->ipi_listhead, inp_list) { INP_WLOCK(inp); imo = inp->inp_moptions; if ((inp->inp_vflag & INP_IPV4) && imo != NULL) { /* * Unselect the outgoing interface if it is being * detached. */ if (imo->imo_multicast_ifp == ifp) imo->imo_multicast_ifp = NULL; /* * Drop multicast group membership if we joined * through the interface being detached. */ for (i = 0, gap = 0; i < imo->imo_num_memberships; i++) { if (imo->imo_membership[i]->inm_ifp == ifp) { in_delmulti(imo->imo_membership[i]); gap++; } else if (gap != 0) imo->imo_membership[i - gap] = imo->imo_membership[i]; } imo->imo_num_memberships -= gap; } INP_WUNLOCK(inp); } INP_INFO_RUNLOCK(pcbinfo); } /* * Lookup a PCB based on the local address and port. Caller must hold the * hash lock. No inpcb locks or references are acquired. */ #define INP_LOOKUP_MAPPED_PCB_COST 3 struct inpcb * in_pcblookup_local(struct inpcbinfo *pcbinfo, struct in_addr laddr, u_short lport, int lookupflags, struct ucred *cred) { struct inpcb *inp; #ifdef INET6 int matchwild = 3 + INP_LOOKUP_MAPPED_PCB_COST; #else int matchwild = 3; #endif int wildcard; KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0, ("%s: invalid lookup flags %d", __func__, lookupflags)); INP_HASH_LOCK_ASSERT(pcbinfo); if ((lookupflags & INPLOOKUP_WILDCARD) == 0) { struct inpcbhead *head; /* * Look for an unconnected (wildcard foreign addr) PCB that * matches the local address and port we're looking for. */ head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport, 0, pcbinfo->ipi_hashmask)]; LIST_FOREACH(inp, head, inp_hash) { #ifdef INET6 /* XXX inp locking */ if ((inp->inp_vflag & INP_IPV4) == 0) continue; #endif if (inp->inp_faddr.s_addr == INADDR_ANY && inp->inp_laddr.s_addr == laddr.s_addr && inp->inp_lport == lport) { /* * Found? */ if (cred == NULL || prison_equal_ip4(cred->cr_prison, inp->inp_cred->cr_prison)) return (inp); } } /* * Not found. */ return (NULL); } else { struct inpcbporthead *porthash; struct inpcbport *phd; struct inpcb *match = NULL; /* * Best fit PCB lookup. * * First see if this local port is in use by looking on the * port hash list. */ porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport, pcbinfo->ipi_porthashmask)]; LIST_FOREACH(phd, porthash, phd_hash) { if (phd->phd_port == lport) break; } if (phd != NULL) { /* * Port is in use by one or more PCBs. Look for best * fit. */ LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) { wildcard = 0; if (cred != NULL && !prison_equal_ip4(inp->inp_cred->cr_prison, cred->cr_prison)) continue; #ifdef INET6 /* XXX inp locking */ if ((inp->inp_vflag & INP_IPV4) == 0) continue; /* * We never select the PCB that has * INP_IPV6 flag and is bound to :: if * we have another PCB which is bound * to 0.0.0.0. If a PCB has the * INP_IPV6 flag, then we set its cost * higher than IPv4 only PCBs. * * Note that the case only happens * when a socket is bound to ::, under * the condition that the use of the * mapped address is allowed. */ if ((inp->inp_vflag & INP_IPV6) != 0) wildcard += INP_LOOKUP_MAPPED_PCB_COST; #endif if (inp->inp_faddr.s_addr != INADDR_ANY) wildcard++; if (inp->inp_laddr.s_addr != INADDR_ANY) { if (laddr.s_addr == INADDR_ANY) wildcard++; else if (inp->inp_laddr.s_addr != laddr.s_addr) continue; } else { if (laddr.s_addr != INADDR_ANY) wildcard++; } if (wildcard < matchwild) { match = inp; matchwild = wildcard; if (matchwild == 0) break; } } } return (match); } } #undef INP_LOOKUP_MAPPED_PCB_COST #ifdef PCBGROUP /* * Lookup PCB in hash list, using pcbgroup tables. */ static struct inpcb * in_pcblookup_group(struct inpcbinfo *pcbinfo, struct inpcbgroup *pcbgroup, struct in_addr faddr, u_int fport_arg, struct in_addr laddr, u_int lport_arg, int lookupflags, struct ifnet *ifp) { struct inpcbhead *head; struct inpcb *inp, *tmpinp; u_short fport = fport_arg, lport = lport_arg; /* * First look for an exact match. */ tmpinp = NULL; INP_GROUP_LOCK(pcbgroup); head = &pcbgroup->ipg_hashbase[INP_PCBHASH(faddr.s_addr, lport, fport, pcbgroup->ipg_hashmask)]; LIST_FOREACH(inp, head, inp_pcbgrouphash) { #ifdef INET6 /* XXX inp locking */ if ((inp->inp_vflag & INP_IPV4) == 0) continue; #endif if (inp->inp_faddr.s_addr == faddr.s_addr && inp->inp_laddr.s_addr == laddr.s_addr && inp->inp_fport == fport && inp->inp_lport == lport) { /* * XXX We should be able to directly return * the inp here, without any checks. * Well unless both bound with SO_REUSEPORT? */ if (prison_flag(inp->inp_cred, PR_IP4)) goto found; if (tmpinp == NULL) tmpinp = inp; } } if (tmpinp != NULL) { inp = tmpinp; goto found; } #ifdef RSS /* * For incoming connections, we may wish to do a wildcard * match for an RSS-local socket. */ if ((lookupflags & INPLOOKUP_WILDCARD) != 0) { struct inpcb *local_wild = NULL, *local_exact = NULL; #ifdef INET6 struct inpcb *local_wild_mapped = NULL; #endif struct inpcb *jail_wild = NULL; struct inpcbhead *head; int injail; /* * Order of socket selection - we always prefer jails. * 1. jailed, non-wild. * 2. jailed, wild. * 3. non-jailed, non-wild. * 4. non-jailed, wild. */ head = &pcbgroup->ipg_hashbase[INP_PCBHASH(INADDR_ANY, lport, 0, pcbgroup->ipg_hashmask)]; LIST_FOREACH(inp, head, inp_pcbgrouphash) { #ifdef INET6 /* XXX inp locking */ if ((inp->inp_vflag & INP_IPV4) == 0) continue; #endif if (inp->inp_faddr.s_addr != INADDR_ANY || inp->inp_lport != lport) continue; injail = prison_flag(inp->inp_cred, PR_IP4); if (injail) { if (prison_check_ip4(inp->inp_cred, &laddr) != 0) continue; } else { if (local_exact != NULL) continue; } if (inp->inp_laddr.s_addr == laddr.s_addr) { if (injail) goto found; else local_exact = inp; } else if (inp->inp_laddr.s_addr == INADDR_ANY) { #ifdef INET6 /* XXX inp locking, NULL check */ if (inp->inp_vflag & INP_IPV6PROTO) local_wild_mapped = inp; else #endif if (injail) jail_wild = inp; else local_wild = inp; } } /* LIST_FOREACH */ inp = jail_wild; if (inp == NULL) inp = local_exact; if (inp == NULL) inp = local_wild; #ifdef INET6 if (inp == NULL) inp = local_wild_mapped; #endif if (inp != NULL) goto found; } #endif /* * Then look for a wildcard match, if requested. */ if ((lookupflags & INPLOOKUP_WILDCARD) != 0) { struct inpcb *local_wild = NULL, *local_exact = NULL; #ifdef INET6 struct inpcb *local_wild_mapped = NULL; #endif struct inpcb *jail_wild = NULL; struct inpcbhead *head; int injail; /* * Order of socket selection - we always prefer jails. * 1. jailed, non-wild. * 2. jailed, wild. * 3. non-jailed, non-wild. * 4. non-jailed, wild. */ head = &pcbinfo->ipi_wildbase[INP_PCBHASH(INADDR_ANY, lport, 0, pcbinfo->ipi_wildmask)]; LIST_FOREACH(inp, head, inp_pcbgroup_wild) { #ifdef INET6 /* XXX inp locking */ if ((inp->inp_vflag & INP_IPV4) == 0) continue; #endif if (inp->inp_faddr.s_addr != INADDR_ANY || inp->inp_lport != lport) continue; injail = prison_flag(inp->inp_cred, PR_IP4); if (injail) { if (prison_check_ip4(inp->inp_cred, &laddr) != 0) continue; } else { if (local_exact != NULL) continue; } if (inp->inp_laddr.s_addr == laddr.s_addr) { if (injail) goto found; else local_exact = inp; } else if (inp->inp_laddr.s_addr == INADDR_ANY) { #ifdef INET6 /* XXX inp locking, NULL check */ if (inp->inp_vflag & INP_IPV6PROTO) local_wild_mapped = inp; else #endif if (injail) jail_wild = inp; else local_wild = inp; } } /* LIST_FOREACH */ inp = jail_wild; if (inp == NULL) inp = local_exact; if (inp == NULL) inp = local_wild; #ifdef INET6 if (inp == NULL) inp = local_wild_mapped; #endif if (inp != NULL) goto found; } /* if (lookupflags & INPLOOKUP_WILDCARD) */ INP_GROUP_UNLOCK(pcbgroup); return (NULL); found: in_pcbref(inp); INP_GROUP_UNLOCK(pcbgroup); if (lookupflags & INPLOOKUP_WLOCKPCB) { INP_WLOCK(inp); if (in_pcbrele_wlocked(inp)) return (NULL); } else if (lookupflags & INPLOOKUP_RLOCKPCB) { INP_RLOCK(inp); if (in_pcbrele_rlocked(inp)) return (NULL); } else panic("%s: locking bug", __func__); return (inp); } #endif /* PCBGROUP */ /* * Lookup PCB in hash list, using pcbinfo tables. This variation assumes * that the caller has locked the hash list, and will not perform any further * locking or reference operations on either the hash list or the connection. */ static struct inpcb * in_pcblookup_hash_locked(struct inpcbinfo *pcbinfo, struct in_addr faddr, u_int fport_arg, struct in_addr laddr, u_int lport_arg, int lookupflags, struct ifnet *ifp) { struct inpcbhead *head; struct inpcb *inp, *tmpinp; u_short fport = fport_arg, lport = lport_arg; KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0, ("%s: invalid lookup flags %d", __func__, lookupflags)); INP_HASH_LOCK_ASSERT(pcbinfo); /* * First look for an exact match. */ tmpinp = NULL; head = &pcbinfo->ipi_hashbase[INP_PCBHASH(faddr.s_addr, lport, fport, pcbinfo->ipi_hashmask)]; LIST_FOREACH(inp, head, inp_hash) { #ifdef INET6 /* XXX inp locking */ if ((inp->inp_vflag & INP_IPV4) == 0) continue; #endif if (inp->inp_faddr.s_addr == faddr.s_addr && inp->inp_laddr.s_addr == laddr.s_addr && inp->inp_fport == fport && inp->inp_lport == lport) { /* * XXX We should be able to directly return * the inp here, without any checks. * Well unless both bound with SO_REUSEPORT? */ if (prison_flag(inp->inp_cred, PR_IP4)) return (inp); if (tmpinp == NULL) tmpinp = inp; } } if (tmpinp != NULL) return (tmpinp); /* * Then look for a wildcard match, if requested. */ if ((lookupflags & INPLOOKUP_WILDCARD) != 0) { struct inpcb *local_wild = NULL, *local_exact = NULL; #ifdef INET6 struct inpcb *local_wild_mapped = NULL; #endif struct inpcb *jail_wild = NULL; int injail; /* * Order of socket selection - we always prefer jails. * 1. jailed, non-wild. * 2. jailed, wild. * 3. non-jailed, non-wild. * 4. non-jailed, wild. */ head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport, 0, pcbinfo->ipi_hashmask)]; LIST_FOREACH(inp, head, inp_hash) { #ifdef INET6 /* XXX inp locking */ if ((inp->inp_vflag & INP_IPV4) == 0) continue; #endif if (inp->inp_faddr.s_addr != INADDR_ANY || inp->inp_lport != lport) continue; injail = prison_flag(inp->inp_cred, PR_IP4); if (injail) { if (prison_check_ip4(inp->inp_cred, &laddr) != 0) continue; } else { if (local_exact != NULL) continue; } if (inp->inp_laddr.s_addr == laddr.s_addr) { if (injail) return (inp); else local_exact = inp; } else if (inp->inp_laddr.s_addr == INADDR_ANY) { #ifdef INET6 /* XXX inp locking, NULL check */ if (inp->inp_vflag & INP_IPV6PROTO) local_wild_mapped = inp; else #endif if (injail) jail_wild = inp; else local_wild = inp; } } /* LIST_FOREACH */ if (jail_wild != NULL) return (jail_wild); if (local_exact != NULL) return (local_exact); if (local_wild != NULL) return (local_wild); #ifdef INET6 if (local_wild_mapped != NULL) return (local_wild_mapped); #endif } /* if ((lookupflags & INPLOOKUP_WILDCARD) != 0) */ return (NULL); } /* * Lookup PCB in hash list, using pcbinfo tables. This variation locks the * hash list lock, and will return the inpcb locked (i.e., requires * INPLOOKUP_LOCKPCB). */ static struct inpcb * in_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in_addr faddr, u_int fport, struct in_addr laddr, u_int lport, int lookupflags, struct ifnet *ifp) { struct inpcb *inp; INP_HASH_RLOCK(pcbinfo); inp = in_pcblookup_hash_locked(pcbinfo, faddr, fport, laddr, lport, (lookupflags & ~(INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)), ifp); if (inp != NULL) { in_pcbref(inp); INP_HASH_RUNLOCK(pcbinfo); if (lookupflags & INPLOOKUP_WLOCKPCB) { INP_WLOCK(inp); if (in_pcbrele_wlocked(inp)) return (NULL); } else if (lookupflags & INPLOOKUP_RLOCKPCB) { INP_RLOCK(inp); if (in_pcbrele_rlocked(inp)) return (NULL); } else panic("%s: locking bug", __func__); } else INP_HASH_RUNLOCK(pcbinfo); return (inp); } /* * Public inpcb lookup routines, accepting a 4-tuple, and optionally, an mbuf * from which a pre-calculated hash value may be extracted. * * Possibly more of this logic should be in in_pcbgroup.c. */ struct inpcb * in_pcblookup(struct inpcbinfo *pcbinfo, struct in_addr faddr, u_int fport, struct in_addr laddr, u_int lport, int lookupflags, struct ifnet *ifp) { #if defined(PCBGROUP) && !defined(RSS) struct inpcbgroup *pcbgroup; #endif KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0, ("%s: invalid lookup flags %d", __func__, lookupflags)); KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0, ("%s: LOCKPCB not set", __func__)); /* * When not using RSS, use connection groups in preference to the * reservation table when looking up 4-tuples. When using RSS, just * use the reservation table, due to the cost of the Toeplitz hash * in software. * * XXXRW: This policy belongs in the pcbgroup code, as in principle * we could be doing RSS with a non-Toeplitz hash that is affordable * in software. */ #if defined(PCBGROUP) && !defined(RSS) if (in_pcbgroup_enabled(pcbinfo)) { pcbgroup = in_pcbgroup_bytuple(pcbinfo, laddr, lport, faddr, fport); return (in_pcblookup_group(pcbinfo, pcbgroup, faddr, fport, laddr, lport, lookupflags, ifp)); } #endif return (in_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport, lookupflags, ifp)); } struct inpcb * in_pcblookup_mbuf(struct inpcbinfo *pcbinfo, struct in_addr faddr, u_int fport, struct in_addr laddr, u_int lport, int lookupflags, struct ifnet *ifp, struct mbuf *m) { #ifdef PCBGROUP struct inpcbgroup *pcbgroup; #endif KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0, ("%s: invalid lookup flags %d", __func__, lookupflags)); KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0, ("%s: LOCKPCB not set", __func__)); #ifdef PCBGROUP /* * If we can use a hardware-generated hash to look up the connection * group, use that connection group to find the inpcb. Otherwise * fall back on a software hash -- or the reservation table if we're * using RSS. * * XXXRW: As above, that policy belongs in the pcbgroup code. */ if (in_pcbgroup_enabled(pcbinfo) && !(M_HASHTYPE_TEST(m, M_HASHTYPE_NONE))) { pcbgroup = in_pcbgroup_byhash(pcbinfo, M_HASHTYPE_GET(m), m->m_pkthdr.flowid); if (pcbgroup != NULL) return (in_pcblookup_group(pcbinfo, pcbgroup, faddr, fport, laddr, lport, lookupflags, ifp)); #ifndef RSS pcbgroup = in_pcbgroup_bytuple(pcbinfo, laddr, lport, faddr, fport); return (in_pcblookup_group(pcbinfo, pcbgroup, faddr, fport, laddr, lport, lookupflags, ifp)); #endif } #endif return (in_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport, lookupflags, ifp)); } #endif /* INET */ /* * Insert PCB onto various hash lists. */ static int in_pcbinshash_internal(struct inpcb *inp, int do_pcbgroup_update) { struct inpcbhead *pcbhash; struct inpcbporthead *pcbporthash; struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; struct inpcbport *phd; u_int32_t hashkey_faddr; INP_WLOCK_ASSERT(inp); INP_HASH_WLOCK_ASSERT(pcbinfo); KASSERT((inp->inp_flags & INP_INHASHLIST) == 0, ("in_pcbinshash: INP_INHASHLIST")); #ifdef INET6 if (inp->inp_vflag & INP_IPV6) hashkey_faddr = INP6_PCBHASHKEY(&inp->in6p_faddr); else #endif hashkey_faddr = inp->inp_faddr.s_addr; pcbhash = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr, inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)]; pcbporthash = &pcbinfo->ipi_porthashbase[ INP_PCBPORTHASH(inp->inp_lport, pcbinfo->ipi_porthashmask)]; /* * Go through port list and look for a head for this lport. */ LIST_FOREACH(phd, pcbporthash, phd_hash) { if (phd->phd_port == inp->inp_lport) break; } /* * If none exists, malloc one and tack it on. */ if (phd == NULL) { phd = malloc(sizeof(struct inpcbport), M_PCB, M_NOWAIT); if (phd == NULL) { return (ENOBUFS); /* XXX */ } phd->phd_port = inp->inp_lport; LIST_INIT(&phd->phd_pcblist); LIST_INSERT_HEAD(pcbporthash, phd, phd_hash); } inp->inp_phd = phd; LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist); LIST_INSERT_HEAD(pcbhash, inp, inp_hash); inp->inp_flags |= INP_INHASHLIST; #ifdef PCBGROUP if (do_pcbgroup_update) in_pcbgroup_update(inp); #endif return (0); } /* * For now, there are two public interfaces to insert an inpcb into the hash * lists -- one that does update pcbgroups, and one that doesn't. The latter * is used only in the TCP syncache, where in_pcbinshash is called before the * full 4-tuple is set for the inpcb, and we don't want to install in the * pcbgroup until later. * * XXXRW: This seems like a misfeature. in_pcbinshash should always update * connection groups, and partially initialised inpcbs should not be exposed * to either reservation hash tables or pcbgroups. */ int in_pcbinshash(struct inpcb *inp) { return (in_pcbinshash_internal(inp, 1)); } int in_pcbinshash_nopcbgroup(struct inpcb *inp) { return (in_pcbinshash_internal(inp, 0)); } /* * Move PCB to the proper hash bucket when { faddr, fport } have been * changed. NOTE: This does not handle the case of the lport changing (the * hashed port list would have to be updated as well), so the lport must * not change after in_pcbinshash() has been called. */ void in_pcbrehash_mbuf(struct inpcb *inp, struct mbuf *m) { struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; struct inpcbhead *head; u_int32_t hashkey_faddr; INP_WLOCK_ASSERT(inp); INP_HASH_WLOCK_ASSERT(pcbinfo); KASSERT(inp->inp_flags & INP_INHASHLIST, ("in_pcbrehash: !INP_INHASHLIST")); #ifdef INET6 if (inp->inp_vflag & INP_IPV6) hashkey_faddr = INP6_PCBHASHKEY(&inp->in6p_faddr); else #endif hashkey_faddr = inp->inp_faddr.s_addr; head = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr, inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)]; LIST_REMOVE(inp, inp_hash); LIST_INSERT_HEAD(head, inp, inp_hash); #ifdef PCBGROUP if (m != NULL) in_pcbgroup_update_mbuf(inp, m); else in_pcbgroup_update(inp); #endif } void in_pcbrehash(struct inpcb *inp) { in_pcbrehash_mbuf(inp, NULL); } /* * Remove PCB from various lists. */ static void in_pcbremlists(struct inpcb *inp) { struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; INP_INFO_WLOCK_ASSERT(pcbinfo); INP_WLOCK_ASSERT(inp); inp->inp_gencnt = ++pcbinfo->ipi_gencnt; if (inp->inp_flags & INP_INHASHLIST) { struct inpcbport *phd = inp->inp_phd; INP_HASH_WLOCK(pcbinfo); LIST_REMOVE(inp, inp_hash); LIST_REMOVE(inp, inp_portlist); if (LIST_FIRST(&phd->phd_pcblist) == NULL) { LIST_REMOVE(phd, phd_hash); free(phd, M_PCB); } INP_HASH_WUNLOCK(pcbinfo); inp->inp_flags &= ~INP_INHASHLIST; } LIST_REMOVE(inp, inp_list); pcbinfo->ipi_count--; #ifdef PCBGROUP in_pcbgroup_remove(inp); #endif } /* * A set label operation has occurred at the socket layer, propagate the * label change into the in_pcb for the socket. */ void in_pcbsosetlabel(struct socket *so) { #ifdef MAC struct inpcb *inp; inp = sotoinpcb(so); KASSERT(inp != NULL, ("in_pcbsosetlabel: so->so_pcb == NULL")); INP_WLOCK(inp); SOCK_LOCK(so); mac_inpcb_sosetlabel(so, inp); SOCK_UNLOCK(so); INP_WUNLOCK(inp); #endif } /* * ipport_tick runs once per second, determining if random port allocation * should be continued. If more than ipport_randomcps ports have been * allocated in the last second, then we return to sequential port * allocation. We return to random allocation only once we drop below * ipport_randomcps for at least ipport_randomtime seconds. */ static void ipport_tick(void *xtp) { VNET_ITERATOR_DECL(vnet_iter); VNET_LIST_RLOCK_NOSLEEP(); VNET_FOREACH(vnet_iter) { CURVNET_SET(vnet_iter); /* XXX appease INVARIANTS here */ if (V_ipport_tcpallocs <= V_ipport_tcplastcount + V_ipport_randomcps) { if (V_ipport_stoprandom > 0) V_ipport_stoprandom--; } else V_ipport_stoprandom = V_ipport_randomtime; V_ipport_tcplastcount = V_ipport_tcpallocs; CURVNET_RESTORE(); } VNET_LIST_RUNLOCK_NOSLEEP(); callout_reset(&ipport_tick_callout, hz, ipport_tick, NULL); } static void ip_fini(void *xtp) { callout_stop(&ipport_tick_callout); } /* * The ipport_callout should start running at about the time we attach the * inet or inet6 domains. */ static void ipport_tick_init(const void *unused __unused) { /* Start ipport_tick. */ callout_init(&ipport_tick_callout, 1); callout_reset(&ipport_tick_callout, 1, ipport_tick, NULL); EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL, SHUTDOWN_PRI_DEFAULT); } SYSINIT(ipport_tick_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ipport_tick_init, NULL); void inp_wlock(struct inpcb *inp) { INP_WLOCK(inp); } void inp_wunlock(struct inpcb *inp) { INP_WUNLOCK(inp); } void inp_rlock(struct inpcb *inp) { INP_RLOCK(inp); } void inp_runlock(struct inpcb *inp) { INP_RUNLOCK(inp); } #ifdef INVARIANTS void inp_lock_assert(struct inpcb *inp) { INP_WLOCK_ASSERT(inp); } void inp_unlock_assert(struct inpcb *inp) { INP_UNLOCK_ASSERT(inp); } #endif void inp_apply_all(void (*func)(struct inpcb *, void *), void *arg) { struct inpcb *inp; INP_INFO_RLOCK(&V_tcbinfo); LIST_FOREACH(inp, V_tcbinfo.ipi_listhead, inp_list) { INP_WLOCK(inp); func(inp, arg); INP_WUNLOCK(inp); } INP_INFO_RUNLOCK(&V_tcbinfo); } struct socket * inp_inpcbtosocket(struct inpcb *inp) { INP_WLOCK_ASSERT(inp); return (inp->inp_socket); } struct tcpcb * inp_inpcbtotcpcb(struct inpcb *inp) { INP_WLOCK_ASSERT(inp); return ((struct tcpcb *)inp->inp_ppcb); } int inp_ip_tos_get(const struct inpcb *inp) { return (inp->inp_ip_tos); } void inp_ip_tos_set(struct inpcb *inp, int val) { inp->inp_ip_tos = val; } void inp_4tuple_get(struct inpcb *inp, uint32_t *laddr, uint16_t *lp, uint32_t *faddr, uint16_t *fp) { INP_LOCK_ASSERT(inp); *laddr = inp->inp_laddr.s_addr; *faddr = inp->inp_faddr.s_addr; *lp = inp->inp_lport; *fp = inp->inp_fport; } struct inpcb * so_sotoinpcb(struct socket *so) { return (sotoinpcb(so)); } struct tcpcb * so_sototcpcb(struct socket *so) { return (sototcpcb(so)); } #ifdef DDB static void db_print_indent(int indent) { int i; for (i = 0; i < indent; i++) db_printf(" "); } static void db_print_inconninfo(struct in_conninfo *inc, const char *name, int indent) { char faddr_str[48], laddr_str[48]; db_print_indent(indent); db_printf("%s at %p\n", name, inc); indent += 2; #ifdef INET6 if (inc->inc_flags & INC_ISIPV6) { /* IPv6. */ ip6_sprintf(laddr_str, &inc->inc6_laddr); ip6_sprintf(faddr_str, &inc->inc6_faddr); } else #endif { /* IPv4. */ inet_ntoa_r(inc->inc_laddr, laddr_str); inet_ntoa_r(inc->inc_faddr, faddr_str); } db_print_indent(indent); db_printf("inc_laddr %s inc_lport %u\n", laddr_str, ntohs(inc->inc_lport)); db_print_indent(indent); db_printf("inc_faddr %s inc_fport %u\n", faddr_str, ntohs(inc->inc_fport)); } static void db_print_inpflags(int inp_flags) { int comma; comma = 0; if (inp_flags & INP_RECVOPTS) { db_printf("%sINP_RECVOPTS", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_RECVRETOPTS) { db_printf("%sINP_RECVRETOPTS", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_RECVDSTADDR) { db_printf("%sINP_RECVDSTADDR", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_HDRINCL) { db_printf("%sINP_HDRINCL", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_HIGHPORT) { db_printf("%sINP_HIGHPORT", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_LOWPORT) { db_printf("%sINP_LOWPORT", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_ANONPORT) { db_printf("%sINP_ANONPORT", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_RECVIF) { db_printf("%sINP_RECVIF", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_MTUDISC) { db_printf("%sINP_MTUDISC", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_RECVTTL) { db_printf("%sINP_RECVTTL", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_DONTFRAG) { db_printf("%sINP_DONTFRAG", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_RECVTOS) { db_printf("%sINP_RECVTOS", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_IPV6_V6ONLY) { db_printf("%sIN6P_IPV6_V6ONLY", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_PKTINFO) { db_printf("%sIN6P_PKTINFO", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_HOPLIMIT) { db_printf("%sIN6P_HOPLIMIT", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_HOPOPTS) { db_printf("%sIN6P_HOPOPTS", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_DSTOPTS) { db_printf("%sIN6P_DSTOPTS", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_RTHDR) { db_printf("%sIN6P_RTHDR", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_RTHDRDSTOPTS) { db_printf("%sIN6P_RTHDRDSTOPTS", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_TCLASS) { db_printf("%sIN6P_TCLASS", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_AUTOFLOWLABEL) { db_printf("%sIN6P_AUTOFLOWLABEL", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_TIMEWAIT) { db_printf("%sINP_TIMEWAIT", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_ONESBCAST) { db_printf("%sINP_ONESBCAST", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_DROPPED) { db_printf("%sINP_DROPPED", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_SOCKREF) { db_printf("%sINP_SOCKREF", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_RFC2292) { db_printf("%sIN6P_RFC2292", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_MTU) { db_printf("IN6P_MTU%s", comma ? ", " : ""); comma = 1; } } static void db_print_inpvflag(u_char inp_vflag) { int comma; comma = 0; if (inp_vflag & INP_IPV4) { db_printf("%sINP_IPV4", comma ? ", " : ""); comma = 1; } if (inp_vflag & INP_IPV6) { db_printf("%sINP_IPV6", comma ? ", " : ""); comma = 1; } if (inp_vflag & INP_IPV6PROTO) { db_printf("%sINP_IPV6PROTO", comma ? ", " : ""); comma = 1; } } static void db_print_inpcb(struct inpcb *inp, const char *name, int indent) { db_print_indent(indent); db_printf("%s at %p\n", name, inp); indent += 2; db_print_indent(indent); db_printf("inp_flow: 0x%x\n", inp->inp_flow); db_print_inconninfo(&inp->inp_inc, "inp_conninfo", indent); db_print_indent(indent); db_printf("inp_ppcb: %p inp_pcbinfo: %p inp_socket: %p\n", inp->inp_ppcb, inp->inp_pcbinfo, inp->inp_socket); db_print_indent(indent); db_printf("inp_label: %p inp_flags: 0x%x (", inp->inp_label, inp->inp_flags); db_print_inpflags(inp->inp_flags); db_printf(")\n"); db_print_indent(indent); db_printf("inp_sp: %p inp_vflag: 0x%x (", inp->inp_sp, inp->inp_vflag); db_print_inpvflag(inp->inp_vflag); db_printf(")\n"); db_print_indent(indent); db_printf("inp_ip_ttl: %d inp_ip_p: %d inp_ip_minttl: %d\n", inp->inp_ip_ttl, inp->inp_ip_p, inp->inp_ip_minttl); db_print_indent(indent); #ifdef INET6 if (inp->inp_vflag & INP_IPV6) { db_printf("in6p_options: %p in6p_outputopts: %p " "in6p_moptions: %p\n", inp->in6p_options, inp->in6p_outputopts, inp->in6p_moptions); db_printf("in6p_icmp6filt: %p in6p_cksum %d " "in6p_hops %u\n", inp->in6p_icmp6filt, inp->in6p_cksum, inp->in6p_hops); } else #endif { db_printf("inp_ip_tos: %d inp_ip_options: %p " "inp_ip_moptions: %p\n", inp->inp_ip_tos, inp->inp_options, inp->inp_moptions); } db_print_indent(indent); db_printf("inp_phd: %p inp_gencnt: %ju\n", inp->inp_phd, (uintmax_t)inp->inp_gencnt); } DB_SHOW_COMMAND(inpcb, db_show_inpcb) { struct inpcb *inp; if (!have_addr) { db_printf("usage: show inpcb \n"); return; } inp = (struct inpcb *)addr; db_print_inpcb(inp, "inpcb", 0); } #endif /* DDB */ Index: projects/routing/sys/netinet/ip_fastfwd.c =================================================================== --- projects/routing/sys/netinet/ip_fastfwd.c (revision 287053) +++ projects/routing/sys/netinet/ip_fastfwd.c (revision 287054) @@ -1,567 +1,567 @@ /*- * Copyright (c) 2003 Andre Oppermann, Internet Business Solutions AG * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. The name of the author may not be used to endorse or promote * products derived from this software without specific prior written * permission. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ /* * ip_fastforward gets its speed from processing the forwarded packet to * completion (if_output on the other side) without any queues or netisr's. * The receiving interface DMAs the packet into memory, the upper half of * driver calls ip_fastforward, we do our routing table lookup and directly * send it off to the outgoing interface, which DMAs the packet to the * network card. The only part of the packet we touch with the CPU is the * IP header (unless there are complex firewall rules touching other parts * of the packet, but that is up to you). We are essentially limited by bus * bandwidth and how fast the network card/driver can set up receives and * transmits. * * We handle basic errors, IP header errors, checksum errors, * destination unreachable, fragmentation and fragmentation needed and * report them via ICMP to the sender. * * Else if something is not pure IPv4 unicast forwarding we fall back to * the normal ip_input processing path. We should only be called from * interfaces connected to the outside world. * * Firewalling is fully supported including divert, ipfw fwd and ipfilter * ipnat and address rewrite. * * IPSEC is not supported if this host is a tunnel broker. IPSEC is * supported for connections to/from local host. * * We try to do the least expensive (in CPU ops) checks and operations * first to catch junk with as little overhead as possible. * * We take full advantage of hardware support for IP checksum and * fragmentation offloading. * * We don't do ICMP redirect in the fast forwarding path. I have had my own * cases where two core routers with Zebra routing suite would send millions * ICMP redirects to connected hosts if the destination router was not the * default gateway. In one case it was filling the routing table of a host * with approximately 300.000 cloned redirect entries until it ran out of * kernel memory. However the networking code proved very robust and it didn't * crash or fail in other ways. */ /* * Many thanks to Matt Thomas of NetBSD for basic structure of ip_flow.c which * is being followed here. */ #include __FBSDID("$FreeBSD$"); #include "opt_ipfw.h" #include "opt_ipstealth.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include static VNET_DEFINE(int, ipfastforward_active); #define V_ipfastforward_active VNET(ipfastforward_active) SYSCTL_INT(_net_inet_ip, OID_AUTO, fastforwarding, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipfastforward_active), 0, "Enable fast IP forwarding"); /* * Try to forward a packet based on the destination address. * This is a fast path optimized for the plain forwarding case. * If the packet is handled (and consumed) here then we return NULL; * otherwise mbuf is returned and the packet should be delivered * to ip_input for full processing. */ struct mbuf * ip_fastforward(struct mbuf *m) { struct ip *ip; struct mbuf *m0 = NULL; - struct nhop_data nhd, *pnhd; + struct nhop_prepend nhd, *pnhd; struct ifnet *ifp; struct in_addr odest, dest; uint16_t sum, ip_len, ip_off; uint32_t fibnum; int error = 0; int hlen, mtu; struct m_tag *fwd_tag = NULL; /* * Are we active and forwarding packets? */ if (!V_ipfastforward_active || !V_ipforwarding) return m; M_ASSERTVALID(m); M_ASSERTPKTHDR(m); fibnum = M_GETFIB(m); pnhd = NULL; /* * Step 1: check for packet drop conditions (and sanity checks) */ /* * Is entire packet big enough? */ if (m->m_pkthdr.len < sizeof(struct ip)) { IPSTAT_INC(ips_tooshort); goto drop; } /* * Is first mbuf large enough for ip header and is header present? */ if (m->m_len < sizeof (struct ip) && (m = m_pullup(m, sizeof (struct ip))) == NULL) { IPSTAT_INC(ips_toosmall); return NULL; /* mbuf already free'd */ } ip = mtod(m, struct ip *); /* * Is it IPv4? */ if (ip->ip_v != IPVERSION) { IPSTAT_INC(ips_badvers); goto drop; } /* * Is IP header length correct and is it in first mbuf? */ hlen = ip->ip_hl << 2; if (hlen < sizeof(struct ip)) { /* minimum header length */ IPSTAT_INC(ips_badhlen); goto drop; } if (hlen > m->m_len) { if ((m = m_pullup(m, hlen)) == NULL) { IPSTAT_INC(ips_badhlen); return NULL; /* mbuf already free'd */ } ip = mtod(m, struct ip *); } /* * Checksum correct? */ if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); else { if (hlen == sizeof(struct ip)) sum = in_cksum_hdr(ip); else sum = in_cksum(m, hlen); } if (sum) { IPSTAT_INC(ips_badsum); goto drop; } /* * Remember that we have checked the IP header and found it valid. */ m->m_pkthdr.csum_flags |= (CSUM_IP_CHECKED | CSUM_IP_VALID); ip_len = ntohs(ip->ip_len); /* * Is IP length longer than packet we have got? */ if (m->m_pkthdr.len < ip_len) { IPSTAT_INC(ips_tooshort); goto drop; } /* * Is packet longer than IP header tells us? If yes, truncate packet. */ if (m->m_pkthdr.len > ip_len) { if (m->m_len == m->m_pkthdr.len) { m->m_len = ip_len; m->m_pkthdr.len = ip_len; } else m_adj(m, ip_len - m->m_pkthdr.len); } /* * Is packet from or to 127/8? */ if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { IPSTAT_INC(ips_badaddr); goto drop; } #ifdef ALTQ /* * Is packet dropped by traffic conditioner? */ if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) goto drop; #endif /* * Step 2: fallback conditions to normal ip_input path processing */ /* * Only IP packets without options */ if (ip->ip_hl != (sizeof(struct ip) >> 2)) { if (V_ip_doopts == 1) return m; else if (V_ip_doopts == 2) { icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_FILTER_PROHIB, 0, 0); return NULL; /* mbuf already free'd */ } /* else ignore IP options and continue */ } /* * Only unicast IP, not from loopback, no L2 or IP broadcast, * no multicast, no INADDR_ANY * * XXX: Probably some of these checks could be direct drop * conditions. However it is not clear whether there are some * hacks or obscure behaviours which make it neccessary to * let ip_input handle it. We play safe here and let ip_input * deal with it until it is proven that we can directly drop it. */ if ((m->m_flags & (M_BCAST|M_MCAST)) || (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) || ntohl(ip->ip_src.s_addr) == (u_long)INADDR_BROADCAST || ntohl(ip->ip_dst.s_addr) == (u_long)INADDR_BROADCAST || IN_MULTICAST(ntohl(ip->ip_src.s_addr)) || IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) || IN_LINKLOCAL(ntohl(ip->ip_src.s_addr)) || IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr)) || ip->ip_src.s_addr == INADDR_ANY || ip->ip_dst.s_addr == INADDR_ANY ) return m; /* * Is it for a local address on this host? */ if (in_localip(ip->ip_dst)) return m; IPSTAT_INC(ips_total); /* * Step 3: incoming packet firewall processing */ odest.s_addr = dest.s_addr = ip->ip_dst.s_addr; /* * Run through list of ipfilter hooks for input packets */ if (!PFIL_HOOKED(&V_inet_pfil_hook)) goto passin; if (pfil_run_hooks( &V_inet_pfil_hook, &m, m->m_pkthdr.rcvif, PFIL_IN, NULL) || m == NULL) goto drop; M_ASSERTVALID(m); M_ASSERTPKTHDR(m); ip = mtod(m, struct ip *); /* m may have changed by pfil hook */ dest.s_addr = ip->ip_dst.s_addr; /* * Destination address changed? */ if (odest.s_addr != dest.s_addr) { /* * Is it now for a local address on this host? */ if (in_localip(dest)) goto forwardlocal; /* * Go on with new destination address */ } if (m->m_flags & M_FASTFWD_OURS) { /* * ipfw changed it for a local address on this host. */ goto forwardlocal; } passin: /* * Step 4: decrement TTL and look up route */ /* * Check TTL */ #ifdef IPSTEALTH if (!V_ipstealth) { #endif if (ip->ip_ttl <= IPTTLDEC) { icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0); return NULL; /* mbuf already free'd */ } /* * Decrement the TTL and incrementally change the IP header checksum. * Don't bother doing this with hw checksum offloading, it's faster * doing it right here. */ ip->ip_ttl -= IPTTLDEC; if (ip->ip_sum >= (u_int16_t) ~htons(IPTTLDEC << 8)) ip->ip_sum -= ~htons(IPTTLDEC << 8); else ip->ip_sum += htons(IPTTLDEC << 8); #ifdef IPSTEALTH } #endif /* * Find route to destination. */ fibnum = M_GETFIB(m); if (fib4_lookup_prepend(fibnum, dest, m, &nhd, NULL) != 0) { IPSTAT_INC(ips_noroute); IPSTAT_INC(ips_cantforward); icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); return (NULL); } pnhd = &nhd; ifp = NH_LIFP(&nhd); /* * Immediately drop blackholed traffic, and directed broadcasts * for either the all-ones or all-zero subnet addresses on * locally attached networks. */ - if ((nhd.nh_flags & (NHOP_BLACKHOLE|RTF_BROADCAST)) != 0) + if ((nhd.nh_flags & (NHF_BLACKHOLE|NHF_BROADCAST)) != 0) goto drop; /* * Step 5: outgoing firewall packet processing */ /* * Run through list of hooks for output packets. */ if (!PFIL_HOOKED(&V_inet_pfil_hook)) goto passout; if (pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_OUT, NULL) || m == NULL) { goto drop; } M_ASSERTVALID(m); M_ASSERTPKTHDR(m); ip = mtod(m, struct ip *); dest.s_addr = ip->ip_dst.s_addr; /* * Destination address changed? */ if (m->m_flags & M_IP_NEXTHOP) fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL); if (odest.s_addr != dest.s_addr || fwd_tag != NULL) { /* * Is it now for a local address on this host? */ if (m->m_flags & M_FASTFWD_OURS || in_localip(dest)) { forwardlocal: /* * Return packet for processing by ip_input(). */ m->m_flags |= M_FASTFWD_OURS; if (pnhd != NULL) - fib4_free_nh(fibnum, pnhd); + fib4_free_nh_prepend(fibnum, pnhd); return m; } /* * Redo route lookup with new destination address */ if (fwd_tag) { dest.s_addr = ((struct sockaddr_in *) (fwd_tag + 1))->sin_addr.s_addr; m_tag_delete(m, fwd_tag); m->m_flags &= ~M_IP_NEXTHOP; } - fib4_free_nh(fibnum, pnhd); + fib4_free_nh_prepend(fibnum, pnhd); if (fib4_lookup_prepend(fibnum, dest, m, &nhd, NULL) != 0) { IPSTAT_INC(ips_noroute); IPSTAT_INC(ips_cantforward); icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); return (NULL); } pnhd = &nhd; ifp = NH_LIFP(&nhd); } passout: /* * Step 6: send off the packet */ ip_len = ntohs(ip->ip_len); ip_off = ntohs(ip->ip_off); #if 0 /* * Check if route is dampned (when ARP is unable to resolve) */ if ((ro.ro_rt->rt_flags & RTF_REJECT) && (ro.ro_rt->rt_expire == 0 || time_uptime < ro.ro_rt->rt_expire)) { icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); goto consumed; } #endif - if ((nhd.nh_flags & NHOP_REJECT) != 0) { + if ((nhd.nh_flags & NHF_REJECT) != 0) { icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); goto consumed; } /* * Check if media link state of interface is not down */ if (ifp->if_link_state == LINK_STATE_DOWN) { icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); goto consumed; } /* * Check if packet fits MTU or if hardware will fragment for us */ mtu = nhd.nh_mtu; if (ip_len <= mtu) { /* * Avoid confusing lower layers. */ m_clrprotoflags(m); /* * Send off the packet via outgoing interface */ IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL); error = fib4_sendmbuf(ifp, m, &nhd, dest); } else { /* * Handle EMSGSIZE with icmp reply needfrag for TCP MTU discovery */ if (ip_off & IP_DF) { IPSTAT_INC(ips_cantfrag); icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG, 0, mtu); goto consumed; } else { /* * We have to fragment the packet */ m->m_pkthdr.csum_flags |= CSUM_IP; if (ip_fragment(ip, &m, mtu, ifp->if_hwassist)) goto drop; KASSERT(m != NULL, ("null mbuf and no error")); /* * Send off the fragments via outgoing interface */ error = 0; do { m0 = m->m_nextpkt; m->m_nextpkt = NULL; /* * Avoid confusing lower layers. */ m_clrprotoflags(m); IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL); error = fib4_sendmbuf(ifp, m, &nhd, dest); if (error) break; } while ((m = m0) != NULL); if (error) { /* Reclaim remaining fragments */ for (m = m0; m; m = m0) { m0 = m->m_nextpkt; m_freem(m); } } else IPSTAT_INC(ips_fragmented); } } if (error != 0) IPSTAT_INC(ips_odropped); else { IPSTAT_INC(ips_forward); IPSTAT_INC(ips_fastforward); } consumed: - fib4_free_nh(fibnum, &nhd); + fib4_free_nh_prepend(fibnum, &nhd); return NULL; drop: if (m) m_freem(m); if (pnhd != NULL) - fib4_free_nh(fibnum, pnhd); + fib4_free_nh_prepend(fibnum, pnhd); return NULL; } Index: projects/routing/sys/netinet/ip_icmp.c =================================================================== --- projects/routing/sys/netinet/ip_icmp.c (revision 287053) +++ projects/routing/sys/netinet/ip_icmp.c (revision 287054) @@ -1,970 +1,970 @@ /*- * Copyright (c) 1982, 1986, 1988, 1993 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)ip_icmp.c 8.2 (Berkeley) 1/4/94 */ #include __FBSDID("$FreeBSD$"); #include "opt_inet.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef INET #include #include #endif /* INET */ /* * ICMP routines: error generation, receive packet processing, and * routines to turnaround packets back to the originator, and * host table maintenance routines. */ static VNET_DEFINE(int, icmplim) = 200; #define V_icmplim VNET(icmplim) SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(icmplim), 0, "Maximum number of ICMP responses per second"); static VNET_DEFINE(int, icmplim_output) = 1; #define V_icmplim_output VNET(icmplim_output) SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(icmplim_output), 0, "Enable logging of ICMP response rate limiting"); #ifdef INET VNET_PCPUSTAT_DEFINE(struct icmpstat, icmpstat); VNET_PCPUSTAT_SYSINIT(icmpstat); SYSCTL_VNET_PCPUSTAT(_net_inet_icmp, ICMPCTL_STATS, stats, struct icmpstat, icmpstat, "ICMP statistics (struct icmpstat, netinet/icmp_var.h)"); #ifdef VIMAGE VNET_PCPUSTAT_SYSUNINIT(icmpstat); #endif /* VIMAGE */ static VNET_DEFINE(int, icmpmaskrepl) = 0; #define V_icmpmaskrepl VNET(icmpmaskrepl) SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(icmpmaskrepl), 0, "Reply to ICMP Address Mask Request packets."); static VNET_DEFINE(u_int, icmpmaskfake) = 0; #define V_icmpmaskfake VNET(icmpmaskfake) SYSCTL_UINT(_net_inet_icmp, OID_AUTO, maskfake, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(icmpmaskfake), 0, "Fake reply to ICMP Address Mask Request packets."); VNET_DEFINE(int, drop_redirect) = 0; #define V_drop_redirect VNET(drop_redirect) SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(drop_redirect), 0, "Ignore ICMP redirects"); static VNET_DEFINE(int, log_redirect) = 0; #define V_log_redirect VNET(log_redirect) SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(log_redirect), 0, "Log ICMP redirects to the console"); static VNET_DEFINE(char, reply_src[IFNAMSIZ]); #define V_reply_src VNET(reply_src) SYSCTL_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(reply_src), IFNAMSIZ, "icmp reply source for non-local packets."); static VNET_DEFINE(int, icmp_rfi) = 0; #define V_icmp_rfi VNET(icmp_rfi) SYSCTL_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(icmp_rfi), 0, "ICMP reply from incoming interface for non-local packets"); static VNET_DEFINE(int, icmp_quotelen) = 8; #define V_icmp_quotelen VNET(icmp_quotelen) SYSCTL_INT(_net_inet_icmp, OID_AUTO, quotelen, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(icmp_quotelen), 0, "Number of bytes from original packet to quote in ICMP reply"); /* * ICMP broadcast echo sysctl */ static VNET_DEFINE(int, icmpbmcastecho) = 0; #define V_icmpbmcastecho VNET(icmpbmcastecho) SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(icmpbmcastecho), 0, ""); static VNET_DEFINE(int, icmptstamprepl) = 1; #define V_icmptstamprepl VNET(icmptstamprepl) SYSCTL_INT(_net_inet_icmp, OID_AUTO, tstamprepl, CTLFLAG_RW, &VNET_NAME(icmptstamprepl), 0, "Respond to ICMP Timestamp packets"); #ifdef ICMPPRINTFS int icmpprintfs = 0; #endif static void icmp_reflect(struct mbuf *); static void icmp_send(struct mbuf *, struct mbuf *); extern struct protosw inetsw[]; /* * Kernel module interface for updating icmpstat. The argument is an index * into icmpstat treated as an array of u_long. While this encodes the * general layout of icmpstat into the caller, it doesn't encode its * location, so that future changes to add, for example, per-CPU stats * support won't cause binary compatibility problems for kernel modules. */ void kmod_icmpstat_inc(int statnum) { counter_u64_add(VNET(icmpstat)[statnum], 1); } /* * Generate an error packet of type error * in response to bad packet ip. */ void icmp_error(struct mbuf *n, int type, int code, uint32_t dest, int mtu) { register struct ip *oip = mtod(n, struct ip *), *nip; register unsigned oiphlen = oip->ip_hl << 2; register struct icmp *icp; register struct mbuf *m; unsigned icmplen, icmpelen, nlen; KASSERT((u_int)type <= ICMP_MAXTYPE, ("%s: illegal ICMP type", __func__)); #ifdef ICMPPRINTFS if (icmpprintfs) printf("icmp_error(%p, %x, %d)\n", oip, type, code); #endif if (type != ICMP_REDIRECT) ICMPSTAT_INC(icps_error); /* * Don't send error: * if the original packet was encrypted. * if not the first fragment of message. * in response to a multicast or broadcast packet. * if the old packet protocol was an ICMP error message. */ if (n->m_flags & M_DECRYPTED) goto freeit; if (oip->ip_off & htons(~(IP_MF|IP_DF))) goto freeit; if (n->m_flags & (M_BCAST|M_MCAST)) goto freeit; if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT && n->m_len >= oiphlen + ICMP_MINLEN && !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiphlen))->icmp_type)) { ICMPSTAT_INC(icps_oldicmp); goto freeit; } /* Drop if IP header plus 8 bytes is not contignous in first mbuf. */ if (oiphlen + 8 > n->m_len) goto freeit; /* * Calculate length to quote from original packet and * prevent the ICMP mbuf from overflowing. * Unfortunatly this is non-trivial since ip_forward() * sends us truncated packets. */ nlen = m_length(n, NULL); if (oip->ip_p == IPPROTO_TCP) { struct tcphdr *th; int tcphlen; if (oiphlen + sizeof(struct tcphdr) > n->m_len && n->m_next == NULL) goto stdreply; if (n->m_len < oiphlen + sizeof(struct tcphdr) && ((n = m_pullup(n, oiphlen + sizeof(struct tcphdr))) == NULL)) goto freeit; th = (struct tcphdr *)((caddr_t)oip + oiphlen); tcphlen = th->th_off << 2; if (tcphlen < sizeof(struct tcphdr)) goto freeit; if (ntohs(oip->ip_len) < oiphlen + tcphlen) goto freeit; if (oiphlen + tcphlen > n->m_len && n->m_next == NULL) goto stdreply; if (n->m_len < oiphlen + tcphlen && ((n = m_pullup(n, oiphlen + tcphlen)) == NULL)) goto freeit; icmpelen = max(tcphlen, min(V_icmp_quotelen, ntohs(oip->ip_len) - oiphlen)); } else stdreply: icmpelen = max(8, min(V_icmp_quotelen, ntohs(oip->ip_len) - oiphlen)); icmplen = min(oiphlen + icmpelen, nlen); if (icmplen < sizeof(struct ip)) goto freeit; if (MHLEN > sizeof(struct ip) + ICMP_MINLEN + icmplen) m = m_gethdr(M_NOWAIT, MT_DATA); else m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); if (m == NULL) goto freeit; #ifdef MAC mac_netinet_icmp_reply(n, m); #endif icmplen = min(icmplen, M_TRAILINGSPACE(m) - sizeof(struct ip) - ICMP_MINLEN); m_align(m, ICMP_MINLEN + icmplen); m->m_len = ICMP_MINLEN + icmplen; /* XXX MRT make the outgoing packet use the same FIB * that was associated with the incoming packet */ M_SETFIB(m, M_GETFIB(n)); icp = mtod(m, struct icmp *); ICMPSTAT_INC(icps_outhist[type]); icp->icmp_type = type; if (type == ICMP_REDIRECT) icp->icmp_gwaddr.s_addr = dest; else { icp->icmp_void = 0; /* * The following assignments assume an overlay with the * just zeroed icmp_void field. */ if (type == ICMP_PARAMPROB) { icp->icmp_pptr = code; code = 0; } else if (type == ICMP_UNREACH && code == ICMP_UNREACH_NEEDFRAG && mtu) { icp->icmp_nextmtu = htons(mtu); } } icp->icmp_code = code; /* * Copy the quotation into ICMP message and * convert quoted IP header back to network representation. */ m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip); nip = &icp->icmp_ip; /* * Set up ICMP message mbuf and copy old IP header (without options * in front of ICMP message. * If the original mbuf was meant to bypass the firewall, the error * reply should bypass as well. */ m->m_flags |= n->m_flags & M_SKIP_FIREWALL; m->m_data -= sizeof(struct ip); m->m_len += sizeof(struct ip); m->m_pkthdr.len = m->m_len; m->m_pkthdr.rcvif = n->m_pkthdr.rcvif; nip = mtod(m, struct ip *); bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip)); nip->ip_len = htons(m->m_len); nip->ip_v = IPVERSION; nip->ip_hl = 5; nip->ip_p = IPPROTO_ICMP; nip->ip_tos = 0; nip->ip_off = 0; icmp_reflect(m); freeit: m_freem(n); } /* * Process a received ICMP message. */ int icmp_input(struct mbuf **mp, int *offp, int proto) { struct icmp *icp; struct in_ifaddr *ia; struct mbuf *m = *mp; struct ip *ip = mtod(m, struct ip *); struct sockaddr_in icmpsrc, icmpdst, icmpgw; int hlen = *offp; int icmplen = ntohs(ip->ip_len) - *offp; int i, code; void (*ctlfunc)(int, struct sockaddr *, void *); int fibnum; *mp = NULL; /* * Locate icmp structure in mbuf, and check * that not corrupted and of at least minimum length. */ #ifdef ICMPPRINTFS if (icmpprintfs) { char buf[4 * sizeof "123"]; strcpy(buf, inet_ntoa(ip->ip_src)); printf("icmp_input from %s to %s, len %d\n", buf, inet_ntoa(ip->ip_dst), icmplen); } #endif if (icmplen < ICMP_MINLEN) { ICMPSTAT_INC(icps_tooshort); goto freeit; } i = hlen + min(icmplen, ICMP_ADVLENMIN); if (m->m_len < i && (m = m_pullup(m, i)) == NULL) { ICMPSTAT_INC(icps_tooshort); return (IPPROTO_DONE); } ip = mtod(m, struct ip *); m->m_len -= hlen; m->m_data += hlen; icp = mtod(m, struct icmp *); if (in_cksum(m, icmplen)) { ICMPSTAT_INC(icps_checksum); goto freeit; } m->m_len += hlen; m->m_data -= hlen; #ifdef ICMPPRINTFS if (icmpprintfs) printf("icmp_input, type %d code %d\n", icp->icmp_type, icp->icmp_code); #endif /* * Message type specific processing. */ if (icp->icmp_type > ICMP_MAXTYPE) goto raw; /* Initialize */ bzero(&icmpsrc, sizeof(icmpsrc)); icmpsrc.sin_len = sizeof(struct sockaddr_in); icmpsrc.sin_family = AF_INET; bzero(&icmpdst, sizeof(icmpdst)); icmpdst.sin_len = sizeof(struct sockaddr_in); icmpdst.sin_family = AF_INET; bzero(&icmpgw, sizeof(icmpgw)); icmpgw.sin_len = sizeof(struct sockaddr_in); icmpgw.sin_family = AF_INET; ICMPSTAT_INC(icps_inhist[icp->icmp_type]); code = icp->icmp_code; switch (icp->icmp_type) { case ICMP_UNREACH: switch (code) { case ICMP_UNREACH_NET: case ICMP_UNREACH_HOST: case ICMP_UNREACH_SRCFAIL: case ICMP_UNREACH_NET_UNKNOWN: case ICMP_UNREACH_HOST_UNKNOWN: case ICMP_UNREACH_ISOLATED: case ICMP_UNREACH_TOSNET: case ICMP_UNREACH_TOSHOST: case ICMP_UNREACH_HOST_PRECEDENCE: case ICMP_UNREACH_PRECEDENCE_CUTOFF: code = PRC_UNREACH_NET; break; case ICMP_UNREACH_NEEDFRAG: code = PRC_MSGSIZE; break; /* * RFC 1122, Sections 3.2.2.1 and 4.2.3.9. * Treat subcodes 2,3 as immediate RST */ case ICMP_UNREACH_PROTOCOL: case ICMP_UNREACH_PORT: code = PRC_UNREACH_PORT; break; case ICMP_UNREACH_NET_PROHIB: case ICMP_UNREACH_HOST_PROHIB: case ICMP_UNREACH_FILTER_PROHIB: code = PRC_UNREACH_ADMIN_PROHIB; break; default: goto badcode; } goto deliver; case ICMP_TIMXCEED: if (code > 1) goto badcode; code += PRC_TIMXCEED_INTRANS; goto deliver; case ICMP_PARAMPROB: if (code > 1) goto badcode; code = PRC_PARAMPROB; deliver: /* * Problem with datagram; advise higher level routines. */ if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) || icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) { ICMPSTAT_INC(icps_badlen); goto freeit; } /* Discard ICMP's in response to multicast packets */ if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr))) goto badcode; #ifdef ICMPPRINTFS if (icmpprintfs) printf("deliver to protocol %d\n", icp->icmp_ip.ip_p); #endif icmpsrc.sin_addr = icp->icmp_ip.ip_dst; /* * XXX if the packet contains [IPv4 AH TCP], we can't make a * notification to TCP layer. */ ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput; if (ctlfunc) (*ctlfunc)(code, (struct sockaddr *)&icmpsrc, (void *)&icp->icmp_ip); break; badcode: ICMPSTAT_INC(icps_badcode); break; case ICMP_ECHO: if (!V_icmpbmcastecho && (m->m_flags & (M_MCAST | M_BCAST)) != 0) { ICMPSTAT_INC(icps_bmcastecho); break; } icp->icmp_type = ICMP_ECHOREPLY; if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0) goto freeit; else goto reflect; case ICMP_TSTAMP: if (V_icmptstamprepl == 0) break; if (!V_icmpbmcastecho && (m->m_flags & (M_MCAST | M_BCAST)) != 0) { ICMPSTAT_INC(icps_bmcasttstamp); break; } if (icmplen < ICMP_TSLEN) { ICMPSTAT_INC(icps_badlen); break; } icp->icmp_type = ICMP_TSTAMPREPLY; icp->icmp_rtime = iptime(); icp->icmp_ttime = icp->icmp_rtime; /* bogus, do later! */ if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0) goto freeit; else goto reflect; case ICMP_MASKREQ: if (V_icmpmaskrepl == 0) break; /* * We are not able to respond with all ones broadcast * unless we receive it over a point-to-point interface. */ if (icmplen < ICMP_MASKLEN) break; switch (ip->ip_dst.s_addr) { case INADDR_BROADCAST: case INADDR_ANY: icmpdst.sin_addr = ip->ip_src; break; default: icmpdst.sin_addr = ip->ip_dst; } ia = (struct in_ifaddr *)ifaof_ifpforaddr( (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif); if (ia == NULL) break; if (ia->ia_ifp == NULL) { ifa_free(&ia->ia_ifa); break; } icp->icmp_type = ICMP_MASKREPLY; if (V_icmpmaskfake == 0) icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr; else icp->icmp_mask = V_icmpmaskfake; if (ip->ip_src.s_addr == 0) { if (ia->ia_ifp->if_flags & IFF_BROADCAST) ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr; else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT) ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr; } ifa_free(&ia->ia_ifa); reflect: ICMPSTAT_INC(icps_reflect); ICMPSTAT_INC(icps_outhist[icp->icmp_type]); icmp_reflect(m); return (IPPROTO_DONE); case ICMP_REDIRECT: if (V_log_redirect) { u_long src, dst, gw; src = ntohl(ip->ip_src.s_addr); dst = ntohl(icp->icmp_ip.ip_dst.s_addr); gw = ntohl(icp->icmp_gwaddr.s_addr); printf("icmp redirect from %d.%d.%d.%d: " "%d.%d.%d.%d => %d.%d.%d.%d\n", (int)(src >> 24), (int)((src >> 16) & 0xff), (int)((src >> 8) & 0xff), (int)(src & 0xff), (int)(dst >> 24), (int)((dst >> 16) & 0xff), (int)((dst >> 8) & 0xff), (int)(dst & 0xff), (int)(gw >> 24), (int)((gw >> 16) & 0xff), (int)((gw >> 8) & 0xff), (int)(gw & 0xff)); } /* * RFC1812 says we must ignore ICMP redirects if we * are acting as router. */ if (V_drop_redirect || V_ipforwarding) break; if (code > 3) goto badcode; if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) || icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) { ICMPSTAT_INC(icps_badlen); break; } /* * Short circuit routing redirects to force * immediate change in the kernel's routing * tables. The message is also handed to anyone * listening on a raw socket (e.g. the routing * daemon for use in updating its tables). */ icmpgw.sin_addr = ip->ip_src; icmpdst.sin_addr = icp->icmp_gwaddr; #ifdef ICMPPRINTFS if (icmpprintfs) { char buf[4 * sizeof "123"]; strcpy(buf, inet_ntoa(icp->icmp_ip.ip_dst)); printf("redirect dst %s to %s\n", buf, inet_ntoa(icp->icmp_gwaddr)); } #endif icmpsrc.sin_addr = icp->icmp_ip.ip_dst; for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) { in_rtredirect((struct sockaddr *)&icmpsrc, (struct sockaddr *)&icmpdst, (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST, (struct sockaddr *)&icmpgw, fibnum); } pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc); break; /* * No kernel processing for the following; * just fall through to send to raw listener. */ case ICMP_ECHOREPLY: case ICMP_ROUTERADVERT: case ICMP_ROUTERSOLICIT: case ICMP_TSTAMPREPLY: case ICMP_IREQREPLY: case ICMP_MASKREPLY: case ICMP_SOURCEQUENCH: default: break; } raw: *mp = m; rip_input(mp, offp, proto); return (IPPROTO_DONE); freeit: m_freem(m); return (IPPROTO_DONE); } /* * Reflect the ip packet back to the source */ static void icmp_reflect(struct mbuf *m) { struct rm_priotracker in_ifa_tracker; struct ip *ip = mtod(m, struct ip *); struct ifaddr *ifa; struct ifnet *ifp; struct in_ifaddr *ia; struct in_addr t; struct mbuf *opts = 0; struct nhop4_extended nh_ext; int optlen = (ip->ip_hl << 2) - sizeof(struct ip); if (IN_MULTICAST(ntohl(ip->ip_src.s_addr)) || IN_EXPERIMENTAL(ntohl(ip->ip_src.s_addr)) || IN_ZERONET(ntohl(ip->ip_src.s_addr)) ) { m_freem(m); /* Bad return address */ ICMPSTAT_INC(icps_badaddr); goto done; /* Ip_output() will check for broadcast */ } t = ip->ip_dst; ip->ip_dst = ip->ip_src; /* * Source selection for ICMP replies: * * If the incoming packet was addressed directly to one of our * own addresses, use dst as the src for the reply. */ IN_IFADDR_RLOCK(&in_ifa_tracker); LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) { if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) { t = IA_SIN(ia)->sin_addr; IN_IFADDR_RUNLOCK(&in_ifa_tracker); goto match; } } IN_IFADDR_RUNLOCK(&in_ifa_tracker); /* * If the incoming packet was addressed to one of our broadcast * addresses, use the first non-broadcast address which corresponds * to the incoming interface. */ ifp = m->m_pkthdr.rcvif; if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) { IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { if (ifa->ifa_addr->sa_family != AF_INET) continue; ia = ifatoia(ifa); if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == t.s_addr) { t = IA_SIN(ia)->sin_addr; IF_ADDR_RUNLOCK(ifp); goto match; } } IF_ADDR_RUNLOCK(ifp); } /* * If the packet was transiting through us, use the address of * the interface the packet came through in. If that interface * doesn't have a suitable IP address, the normal selection * criteria apply. */ if (V_icmp_rfi && ifp != NULL) { IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { if (ifa->ifa_addr->sa_family != AF_INET) continue; ia = ifatoia(ifa); t = IA_SIN(ia)->sin_addr; IF_ADDR_RUNLOCK(ifp); goto match; } IF_ADDR_RUNLOCK(ifp); } /* * If the incoming packet was not addressed directly to us, use * designated interface for icmp replies specified by sysctl * net.inet.icmp.reply_src (default not set). Otherwise continue * with normal source selection. */ if (V_reply_src[0] != '\0' && (ifp = ifunit(V_reply_src))) { IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { if (ifa->ifa_addr->sa_family != AF_INET) continue; ia = ifatoia(ifa); t = IA_SIN(ia)->sin_addr; IF_ADDR_RUNLOCK(ifp); goto match; } IF_ADDR_RUNLOCK(ifp); } /* * If the packet was transiting through us, use the address of * the interface that is the closest to the packet source. * When we don't have a route back to the packet source, stop here * and drop the packet. */ - if (fib4_lookup_nh_extended(M_GETFIB(m), ip->ip_dst, 0, &nh_ext) != 0) { + if (fib4_lookup_nh_ext(M_GETFIB(m), ip->ip_dst, 0, 0, &nh_ext) != 0) { m_freem(m); ICMPSTAT_INC(icps_noroute); goto done; } t = nh_ext.nh_src; match: #ifdef MAC mac_netinet_icmp_replyinplace(m); #endif ip->ip_src = t; ip->ip_ttl = V_ip_defttl; if (optlen > 0) { register u_char *cp; int opt, cnt; u_int len; /* * Retrieve any source routing from the incoming packet; * add on any record-route or timestamp options. */ cp = (u_char *) (ip + 1); if ((opts = ip_srcroute(m)) == 0 && (opts = m_gethdr(M_NOWAIT, MT_DATA))) { opts->m_len = sizeof(struct in_addr); mtod(opts, struct in_addr *)->s_addr = 0; } if (opts) { #ifdef ICMPPRINTFS if (icmpprintfs) printf("icmp_reflect optlen %d rt %d => ", optlen, opts->m_len); #endif for (cnt = optlen; cnt > 0; cnt -= len, cp += len) { opt = cp[IPOPT_OPTVAL]; if (opt == IPOPT_EOL) break; if (opt == IPOPT_NOP) len = 1; else { if (cnt < IPOPT_OLEN + sizeof(*cp)) break; len = cp[IPOPT_OLEN]; if (len < IPOPT_OLEN + sizeof(*cp) || len > cnt) break; } /* * Should check for overflow, but it "can't happen" */ if (opt == IPOPT_RR || opt == IPOPT_TS || opt == IPOPT_SECURITY) { bcopy((caddr_t)cp, mtod(opts, caddr_t) + opts->m_len, len); opts->m_len += len; } } /* Terminate & pad, if necessary */ cnt = opts->m_len % 4; if (cnt) { for (; cnt < 4; cnt++) { *(mtod(opts, caddr_t) + opts->m_len) = IPOPT_EOL; opts->m_len++; } } #ifdef ICMPPRINTFS if (icmpprintfs) printf("%d\n", opts->m_len); #endif } ip_stripoptions(m); } m_tag_delete_nonpersistent(m); m->m_flags &= ~(M_BCAST|M_MCAST); icmp_send(m, opts); done: if (opts) (void)m_free(opts); } /* * Send an icmp packet back to the ip level, * after supplying a checksum. */ static void icmp_send(struct mbuf *m, struct mbuf *opts) { register struct ip *ip = mtod(m, struct ip *); register int hlen; register struct icmp *icp; hlen = ip->ip_hl << 2; m->m_data += hlen; m->m_len -= hlen; icp = mtod(m, struct icmp *); icp->icmp_cksum = 0; icp->icmp_cksum = in_cksum(m, ntohs(ip->ip_len) - hlen); m->m_data -= hlen; m->m_len += hlen; m->m_pkthdr.rcvif = (struct ifnet *)0; #ifdef ICMPPRINTFS if (icmpprintfs) { char buf[4 * sizeof "123"]; strcpy(buf, inet_ntoa(ip->ip_dst)); printf("icmp_send dst %s src %s\n", buf, inet_ntoa(ip->ip_src)); } #endif (void) ip_output(m, opts, NULL, 0, NULL, NULL); } /* * Return milliseconds since 00:00 UTC in network format. */ uint32_t iptime(void) { struct timeval atv; u_long t; getmicrotime(&atv); t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000; return (htonl(t)); } /* * Return the next larger or smaller MTU plateau (table from RFC 1191) * given current value MTU. If DIR is less than zero, a larger plateau * is returned; otherwise, a smaller value is returned. */ int ip_next_mtu(int mtu, int dir) { static int mtutab[] = { 65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508, 296, 68, 0 }; int i, size; size = (sizeof mtutab) / (sizeof mtutab[0]); if (dir >= 0) { for (i = 0; i < size; i++) if (mtu > mtutab[i]) return mtutab[i]; } else { for (i = size - 1; i >= 0; i--) if (mtu < mtutab[i]) return mtutab[i]; if (mtu == mtutab[0]) return mtutab[0]; } return 0; } #endif /* INET */ /* * badport_bandlim() - check for ICMP bandwidth limit * * Return 0 if it is ok to send an ICMP error response, -1 if we have * hit our bandwidth limit and it is not ok. * * If icmplim is <= 0, the feature is disabled and 0 is returned. * * For now we separate the TCP and UDP subsystems w/ different 'which' * values. We may eventually remove this separation (and simplify the * code further). * * Note that the printing of the error message is delayed so we can * properly print the icmp error rate that the system was trying to do * (i.e. 22000/100 pps, etc...). This can cause long delays in printing * the 'final' error, but it doesn't make sense to solve the printing * delay with more complex code. */ int badport_bandlim(int which) { #define N(a) (sizeof (a) / sizeof (a[0])) static struct rate { const char *type; struct timeval lasttime; int curpps; } rates[BANDLIM_MAX+1] = { { "icmp unreach response" }, { "icmp ping response" }, { "icmp tstamp response" }, { "closed port RST response" }, { "open port RST response" }, { "icmp6 unreach response" }, { "sctp ootb response" } }; /* * Return ok status if feature disabled or argument out of range. */ if (V_icmplim > 0 && (u_int) which < N(rates)) { struct rate *r = &rates[which]; int opps = r->curpps; if (!ppsratecheck(&r->lasttime, &r->curpps, V_icmplim)) return -1; /* discard packet */ /* * If we've dropped below the threshold after having * rate-limited traffic print the message. This preserves * the previous behaviour at the expense of added complexity. */ if (V_icmplim_output && opps > V_icmplim) log(LOG_NOTICE, "Limiting %s from %d to %d packets/sec\n", r->type, opps, V_icmplim); } return 0; /* okay to send packet */ #undef N } Index: projects/routing/sys/netinet/ip_input.c =================================================================== --- projects/routing/sys/netinet/ip_input.c (revision 287053) +++ projects/routing/sys/netinet/ip_input.c (revision 287054) @@ -1,1258 +1,1259 @@ /*- * Copyright (c) 1982, 1986, 1988, 1993 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 */ #include __FBSDID("$FreeBSD$"); #include "opt_bootp.h" #include "opt_ipfw.h" #include "opt_ipstealth.h" #include "opt_ipsec.h" #include "opt_route.h" #include "opt_rss.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef IPSEC #include #endif /* IPSEC */ #include #include #include #include #ifdef CTASSERT CTASSERT(sizeof(struct ip) == 20); #endif /* IP reassembly functions are defined in ip_reass.c. */ extern void ipreass_init(void); extern void ipreass_drain(void); extern void ipreass_slowtimo(void); #ifdef VIMAGE extern void ipreass_destroy(void); #endif struct rmlock in_ifaddr_lock; RM_SYSINIT(in_ifaddr_lock, &in_ifaddr_lock, "in_ifaddr_lock"); VNET_DEFINE(int, rsvp_on); VNET_DEFINE(int, ipforwarding); SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipforwarding), 0, "Enable IP forwarding between interfaces"); static VNET_DEFINE(int, ipsendredirects) = 1; /* XXX */ #define V_ipsendredirects VNET(ipsendredirects) SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsendredirects), 0, "Enable sending IP redirects"); /* * XXX - Setting ip_checkinterface mostly implements the receive side of * the Strong ES model described in RFC 1122, but since the routing table * and transmit implementation do not implement the Strong ES model, * setting this to 1 results in an odd hybrid. * * XXX - ip_checkinterface currently must be disabled if you use ipnat * to translate the destination address to another local interface. * * XXX - ip_checkinterface must be disabled if you add IP aliases * to the loopback interface instead of the interface where the * packets for those addresses are received. */ static VNET_DEFINE(int, ip_checkinterface); #define V_ip_checkinterface VNET(ip_checkinterface) SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_checkinterface), 0, "Verify packet arrives on correct interface"); VNET_DEFINE(struct pfil_head, inet_pfil_hook); /* Packet filter hooks */ static struct netisr_handler ip_nh = { .nh_name = "ip", .nh_handler = ip_input, .nh_proto = NETISR_IP, #ifdef RSS .nh_m2cpuid = rss_soft_m2cpuid, .nh_policy = NETISR_POLICY_CPU, .nh_dispatch = NETISR_DISPATCH_HYBRID, #else .nh_policy = NETISR_POLICY_FLOW, #endif }; #ifdef RSS /* * Directly dispatched frames are currently assumed * to have a flowid already calculated. * * It should likely have something that assert it * actually has valid flow details. */ static struct netisr_handler ip_direct_nh = { .nh_name = "ip_direct", .nh_handler = ip_direct_input, .nh_proto = NETISR_IP_DIRECT, .nh_m2cpuid = rss_m2cpuid, .nh_policy = NETISR_POLICY_CPU, .nh_dispatch = NETISR_DISPATCH_HYBRID, }; #endif extern struct domain inetdomain; extern struct protosw inetsw[]; u_char ip_protox[IPPROTO_MAX]; VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead); /* first inet address */ VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table */ VNET_DEFINE(u_long, in_ifaddrhmask); /* mask for hash table */ #ifdef IPCTL_DEFMTU SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, &ip_mtu, 0, "Default MTU"); #endif #ifdef IPSTEALTH VNET_DEFINE(int, ipstealth); SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipstealth), 0, "IP stealth mode, no TTL decrementation on forwarding"); #endif /* * IP statistics are stored in the "array" of counter(9)s. */ VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat); VNET_PCPUSTAT_SYSINIT(ipstat); SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)"); #ifdef VIMAGE VNET_PCPUSTAT_SYSUNINIT(ipstat); #endif /* VIMAGE */ /* * Kernel module interface for updating ipstat. The argument is an index * into ipstat treated as an array. */ void kmod_ipstat_inc(int statnum) { counter_u64_add(VNET(ipstat)[statnum], 1); } void kmod_ipstat_dec(int statnum) { counter_u64_add(VNET(ipstat)[statnum], -1); } static int sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS) { int error, qlimit; netisr_getqlimit(&ip_nh, &qlimit); error = sysctl_handle_int(oidp, &qlimit, 0, req); if (error || !req->newptr) return (error); if (qlimit < 1) return (EINVAL); return (netisr_setqlimit(&ip_nh, qlimit)); } SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_queue_maxlen, "I", "Maximum size of the IP input queue"); static int sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS) { u_int64_t qdrops_long; int error, qdrops; netisr_getqdrops(&ip_nh, &qdrops_long); qdrops = qdrops_long; error = sysctl_handle_int(oidp, &qdrops, 0, req); if (error || !req->newptr) return (error); if (qdrops != 0) return (EINVAL); netisr_clearqdrops(&ip_nh); return (0); } SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_queue_drops, "I", "Number of packets dropped from the IP input queue"); #ifdef RSS static int sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS) { int error, qlimit; netisr_getqlimit(&ip_direct_nh, &qlimit); error = sysctl_handle_int(oidp, &qlimit, 0, req); if (error || !req->newptr) return (error); if (qlimit < 1) return (EINVAL); return (netisr_setqlimit(&ip_direct_nh, qlimit)); } SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_direct_queue_maxlen, CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_direct_queue_maxlen, "I", "Maximum size of the IP direct input queue"); static int sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS) { u_int64_t qdrops_long; int error, qdrops; netisr_getqdrops(&ip_direct_nh, &qdrops_long); qdrops = qdrops_long; error = sysctl_handle_int(oidp, &qdrops, 0, req); if (error || !req->newptr) return (error); if (qdrops != 0) return (EINVAL); netisr_clearqdrops(&ip_direct_nh); return (0); } SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_direct_queue_drops, CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_direct_queue_drops, "I", "Number of packets dropped from the IP direct input queue"); #endif /* RSS */ /* * IP initialization: fill in IP protocol switch table. * All protocols not implemented in kernel go to raw IP protocol handler. */ void ip_init(void) { struct protosw *pr; int i; TAILQ_INIT(&V_in_ifaddrhead); V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask); /* Initialize IP reassembly queue. */ ipreass_init(); /* Initialize packet filter hooks. */ V_inet_pfil_hook.ph_type = PFIL_TYPE_AF; V_inet_pfil_hook.ph_af = AF_INET; if ((i = pfil_head_register(&V_inet_pfil_hook)) != 0) printf("%s: WARNING: unable to register pfil hook, " "error %d\n", __func__, i); /* Skip initialization of globals for non-default instances. */ if (!IS_DEFAULT_VNET(curvnet)) return; pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); if (pr == NULL) panic("ip_init: PF_INET not found"); /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */ for (i = 0; i < IPPROTO_MAX; i++) ip_protox[i] = pr - inetsw; /* * Cycle through IP protocols and put them into the appropriate place * in ip_protox[]. */ for (pr = inetdomain.dom_protosw; pr < inetdomain.dom_protoswNPROTOSW; pr++) if (pr->pr_domain->dom_family == PF_INET && pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) { /* Be careful to only index valid IP protocols. */ if (pr->pr_protocol < IPPROTO_MAX) ip_protox[pr->pr_protocol] = pr - inetsw; } netisr_register(&ip_nh); #ifdef RSS netisr_register(&ip_direct_nh); #endif } #ifdef VIMAGE void ip_destroy(void) { int i; if ((i = pfil_head_unregister(&V_inet_pfil_hook)) != 0) printf("%s: WARNING: unable to unregister pfil hook, " "error %d\n", __func__, i); /* Cleanup in_ifaddr hash table; should be empty. */ hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask); /* Destroy IP reassembly queue. */ ipreass_destroy(); } #endif #ifdef RSS /* * IP direct input routine. * * This is called when reinjecting completed fragments where * all of the previous checking and book-keeping has been done. */ void ip_direct_input(struct mbuf *m) { struct ip *ip; int hlen; ip = mtod(m, struct ip *); hlen = ip->ip_hl << 2; IPSTAT_INC(ips_delivered); (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p); return; } #endif /* * Ip input routine. Checksum and byte swap header. If fragmented * try to reassemble. Process options. Pass to next level. */ void ip_input(struct mbuf *m) { struct ip *ip = NULL; struct in_ifaddr *ia = NULL; struct ifaddr *ifa; struct ifnet *ifp; int checkif, hlen = 0; uint16_t sum, ip_len; int dchg = 0; /* dest changed after fw */ struct in_addr odst; /* original dst address */ M_ASSERTPKTHDR(m); if (m->m_flags & M_FASTFWD_OURS) { m->m_flags &= ~M_FASTFWD_OURS; /* Set up some basics that will be used later. */ ip = mtod(m, struct ip *); hlen = ip->ip_hl << 2; ip_len = ntohs(ip->ip_len); goto ours; } IPSTAT_INC(ips_total); if (m->m_pkthdr.len < sizeof(struct ip)) goto tooshort; if (m->m_len < sizeof (struct ip) && (m = m_pullup(m, sizeof (struct ip))) == NULL) { IPSTAT_INC(ips_toosmall); return; } ip = mtod(m, struct ip *); if (ip->ip_v != IPVERSION) { IPSTAT_INC(ips_badvers); goto bad; } hlen = ip->ip_hl << 2; if (hlen < sizeof(struct ip)) { /* minimum header length */ IPSTAT_INC(ips_badhlen); goto bad; } if (hlen > m->m_len) { if ((m = m_pullup(m, hlen)) == NULL) { IPSTAT_INC(ips_badhlen); return; } ip = mtod(m, struct ip *); } IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL); /* 127/8 must not appear on wire - RFC1122 */ ifp = m->m_pkthdr.rcvif; if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { if ((ifp->if_flags & IFF_LOOPBACK) == 0) { IPSTAT_INC(ips_badaddr); goto bad; } } if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); } else { if (hlen == sizeof(struct ip)) { sum = in_cksum_hdr(ip); } else { sum = in_cksum(m, hlen); } } if (sum) { IPSTAT_INC(ips_badsum); goto bad; } #ifdef ALTQ if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) /* packet is dropped by traffic conditioner */ return; #endif ip_len = ntohs(ip->ip_len); if (ip_len < hlen) { IPSTAT_INC(ips_badlen); goto bad; } /* * Check that the amount of data in the buffers * is as at least much as the IP header would have us expect. * Trim mbufs if longer than we expect. * Drop packet if shorter than we expect. */ if (m->m_pkthdr.len < ip_len) { tooshort: IPSTAT_INC(ips_tooshort); goto bad; } if (m->m_pkthdr.len > ip_len) { if (m->m_len == m->m_pkthdr.len) { m->m_len = ip_len; m->m_pkthdr.len = ip_len; } else m_adj(m, ip_len - m->m_pkthdr.len); } #ifdef IPSEC /* * Bypass packet filtering for packets previously handled by IPsec. */ if (ip_ipsec_filtertunnel(m)) goto passin; #endif /* IPSEC */ /* * Run through list of hooks for input packets. * * NB: Beware of the destination address changing (e.g. * by NAT rewriting). When this happens, tell * ip_forward to do the right thing. */ /* Jump over all PFIL processing if hooks are not active. */ if (!PFIL_HOOKED(&V_inet_pfil_hook)) goto passin; odst = ip->ip_dst; if (pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_IN, NULL) != 0) return; if (m == NULL) /* consumed by filter */ return; ip = mtod(m, struct ip *); dchg = (odst.s_addr != ip->ip_dst.s_addr); ifp = m->m_pkthdr.rcvif; if (m->m_flags & M_FASTFWD_OURS) { m->m_flags &= ~M_FASTFWD_OURS; goto ours; } if (m->m_flags & M_IP_NEXTHOP) { dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL); if (dchg != 0) { /* * Directly ship the packet on. This allows * forwarding packets originally destined to us * to some other directly connected host. */ ip_forward(m, 1); return; } } passin: /* * Process options and, if not destined for us, * ship it on. ip_dooptions returns 1 when an * error was detected (causing an icmp message * to be sent and the original packet to be freed). */ if (hlen > sizeof (struct ip) && ip_dooptions(m, 0)) return; /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no * matter if it is destined to another node, or whether it is * a multicast one, RSVP wants it! and prevents it from being forwarded * anywhere else. Also checks if the rsvp daemon is running before * grabbing the packet. */ if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP) goto ours; /* * Check our list of addresses, to see if the packet is for us. * If we don't have any addresses, assume any unicast packet * we receive might be for us (and let the upper layers deal * with it). */ if (TAILQ_EMPTY(&V_in_ifaddrhead) && (m->m_flags & (M_MCAST|M_BCAST)) == 0) goto ours; /* * Enable a consistency check between the destination address * and the arrival interface for a unicast packet (the RFC 1122 * strong ES model) if IP forwarding is disabled and the packet * is not locally generated and the packet is not subject to * 'ipfw fwd'. * * XXX - Checking also should be disabled if the destination * address is ipnat'ed to a different interface. * * XXX - Checking is incompatible with IP aliases added * to the loopback interface instead of the interface where * the packets are received. * * XXX - This is the case for carp vhost IPs as well so we * insert a workaround. If the packet got here, we already * checked with carp_iamatch() and carp_forus(). */ checkif = V_ip_checkinterface && (V_ipforwarding == 0) && ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) && ifp->if_carp == NULL && (dchg == 0); /* * Check for exact addresses in the hash bucket. */ /* IN_IFADDR_RLOCK(); */ LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) { /* * If the address matches, verify that the packet * arrived via the correct interface if checking is * enabled. */ if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr && (!checkif || ia->ia_ifp == ifp)) { counter_u64_add(ia->ia_ifa.ifa_ipackets, 1); counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len); /* IN_IFADDR_RUNLOCK(); */ goto ours; } } /* IN_IFADDR_RUNLOCK(); */ /* * Check for broadcast addresses. * * Only accept broadcast packets that arrive via the matching * interface. Reception of forwarded directed broadcasts would * be handled via ip_forward() and ether_output() with the loopback * into the stack for SIMPLEX interfaces handled by ether_output(). */ if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) { IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { if (ifa->ifa_addr->sa_family != AF_INET) continue; ia = ifatoia(ifa); if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == ip->ip_dst.s_addr) { counter_u64_add(ia->ia_ifa.ifa_ipackets, 1); counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len); IF_ADDR_RUNLOCK(ifp); goto ours; } #ifdef BOOTP_COMPAT if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) { counter_u64_add(ia->ia_ifa.ifa_ipackets, 1); counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len); IF_ADDR_RUNLOCK(ifp); goto ours; } #endif } IF_ADDR_RUNLOCK(ifp); ia = NULL; } /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */ if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) { IPSTAT_INC(ips_cantforward); m_freem(m); return; } if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { if (V_ip_mrouter) { /* * If we are acting as a multicast router, all * incoming multicast packets are passed to the * kernel-level multicast forwarding function. * The packet is returned (relatively) intact; if * ip_mforward() returns a non-zero value, the packet * must be discarded, else it may be accepted below. */ if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) { IPSTAT_INC(ips_cantforward); m_freem(m); return; } /* * The process-level routing daemon needs to receive * all multicast IGMP packets, whether or not this * host belongs to their destination groups. */ if (ip->ip_p == IPPROTO_IGMP) goto ours; IPSTAT_INC(ips_forward); } /* * Assume the packet is for us, to avoid prematurely taking * a lock on the in_multi hash. Protocols must perform * their own filtering and update statistics accordingly. */ goto ours; } if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) goto ours; if (ip->ip_dst.s_addr == INADDR_ANY) goto ours; /* * Not for us; forward if possible and desirable. */ if (V_ipforwarding == 0) { IPSTAT_INC(ips_cantforward); m_freem(m); } else { ip_forward(m, dchg); } return; ours: #ifdef IPSTEALTH /* * IPSTEALTH: Process non-routing options only * if the packet is destined for us. */ if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1)) return; #endif /* IPSTEALTH */ /* * Attempt reassembly; if it succeeds, proceed. * ip_reass() will return a different mbuf. */ if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) { /* XXXGL: shouldn't we save & set m_flags? */ m = ip_reass(m); if (m == NULL) return; ip = mtod(m, struct ip *); /* Get the header length of the reassembled packet */ hlen = ip->ip_hl << 2; } #ifdef IPSEC /* * enforce IPsec policy checking if we are seeing last header. * note that we do not visit this with protocols with pcb layer * code - like udp/tcp/raw ip. */ if (ip_ipsec_input(m, ip->ip_p) != 0) goto bad; #endif /* IPSEC */ /* * Switch out to protocol's input routine. */ IPSTAT_INC(ips_delivered); (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p); return; bad: m_freem(m); } /* * IP timer processing; * if a timer expires on a reassembly * queue, discard it. */ void ip_slowtimo(void) { VNET_ITERATOR_DECL(vnet_iter); VNET_LIST_RLOCK_NOSLEEP(); VNET_FOREACH(vnet_iter) { CURVNET_SET(vnet_iter); ipreass_slowtimo(); CURVNET_RESTORE(); } VNET_LIST_RUNLOCK_NOSLEEP(); } void ip_drain(void) { VNET_ITERATOR_DECL(vnet_iter); VNET_LIST_RLOCK_NOSLEEP(); VNET_FOREACH(vnet_iter) { CURVNET_SET(vnet_iter); ipreass_drain(); CURVNET_RESTORE(); } VNET_LIST_RUNLOCK_NOSLEEP(); } /* * The protocol to be inserted into ip_protox[] must be already registered * in inetsw[], either statically or through pf_proto_register(). */ int ipproto_register(short ipproto) { struct protosw *pr; /* Sanity checks. */ if (ipproto <= 0 || ipproto >= IPPROTO_MAX) return (EPROTONOSUPPORT); /* * The protocol slot must not be occupied by another protocol * already. An index pointing to IPPROTO_RAW is unused. */ pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); if (pr == NULL) return (EPFNOSUPPORT); if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */ return (EEXIST); /* Find the protocol position in inetsw[] and set the index. */ for (pr = inetdomain.dom_protosw; pr < inetdomain.dom_protoswNPROTOSW; pr++) { if (pr->pr_domain->dom_family == PF_INET && pr->pr_protocol && pr->pr_protocol == ipproto) { ip_protox[pr->pr_protocol] = pr - inetsw; return (0); } } return (EPROTONOSUPPORT); } int ipproto_unregister(short ipproto) { struct protosw *pr; /* Sanity checks. */ if (ipproto <= 0 || ipproto >= IPPROTO_MAX) return (EPROTONOSUPPORT); /* Check if the protocol was indeed registered. */ pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); if (pr == NULL) return (EPFNOSUPPORT); if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */ return (ENOENT); /* Reset the protocol slot to IPPROTO_RAW. */ ip_protox[ipproto] = pr - inetsw; return (0); } u_char inetctlerrmap[PRC_NCMDS] = { 0, 0, 0, 0, 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, EMSGSIZE, EHOSTUNREACH, 0, 0, 0, 0, EHOSTUNREACH, 0, ENOPROTOOPT, ECONNREFUSED }; /* * Forward a packet. If some error occurs return the sender * an icmp packet. Note we can't always generate a meaningful * icmp message because icmp doesn't have a large enough repertoire * of codes and types. * * If not forwarding, just drop the packet. This could be confusing * if ipforwarding was zero but some routing protocol was advancing * us as a gateway to somewhere. However, we must let the routing * protocol deal with that. * * The srcrt parameter indicates whether the packet is being forwarded * via a source route. */ void ip_forward(struct mbuf *m, int srcrt) { struct ip *ip = mtod(m, struct ip *); struct mbuf *mcopy; struct in_addr dest; struct nhop4_basic nh4, *pnh4; struct route_info ri; int error, type = 0, code = 0, mtu = 0; if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { IPSTAT_INC(ips_cantforward); m_freem(m); return; } #ifdef IPSEC if (ip_ipsec_fwd(m) != 0) { IPSTAT_INC(ips_cantforward); m_freem(m); return; } #endif /* IPSEC */ #ifdef IPSTEALTH if (!V_ipstealth) { #endif if (ip->ip_ttl <= IPTTLDEC) { icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0); return; } #ifdef IPSTEALTH } #endif pnh4 = &nh4; if (fib4_lookup_nh_basic(M_GETFIB(m), ip->ip_dst, 0, &nh4) != 0) pnh4 = NULL; #ifndef IPSEC /* * 'ia' may be NULL if there is no route for this destination. * In case of IPsec, Don't discard it just yet, but pass it to * ip_output in case of outgoing IPsec policy. */ if (!srcrt && pnh4 == NULL) { icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); RO_RTFREE(&ro); return; } #endif /* * Save the IP header and at most 8 bytes of the payload, * in case we need to generate an ICMP message to the src. * * XXX this can be optimized a lot by saving the data in a local * buffer on the stack (72 bytes at most), and only allocating the * mbuf if really necessary. The vast majority of the packets * are forwarded without having to send an ICMP back (either * because unnecessary, or because rate limited), so we are * really we are wasting a lot of work here. * * We don't use m_copy() because it might return a reference * to a shared cluster. Both this function and ip_output() * assume exclusive access to the IP header in `m', so any * data in a cluster may change before we reach icmp_error(). */ mcopy = m_gethdr(M_NOWAIT, m->m_type); if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) { /* * It's probably ok if the pkthdr dup fails (because * the deep copy of the tag chain failed), but for now * be conservative and just discard the copy since * code below may some day want the tags. */ m_free(mcopy); mcopy = NULL; } if (mcopy != NULL) { mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy)); mcopy->m_pkthdr.len = mcopy->m_len; m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); } #ifdef IPSTEALTH if (!V_ipstealth) { #endif ip->ip_ttl -= IPTTLDEC; #ifdef IPSTEALTH } #endif /* * If forwarding packet using same interface that it came in on, * perhaps should send a redirect to sender to shortcut a hop. * Only send redirect if source is sending directly to us, * and if packet was not source routed (or has any options). * Also, don't send redirect if forwarding using a default route * or a route modified by a redirect. */ dest.s_addr = 0; if (!srcrt && V_ipsendredirects && pnh4 != NULL && nh4.nh_ifp == m->m_pkthdr.rcvif) { - if ((nh4.nh_flags & (NHOP_REDIRECT|NHOP_DEFAULT)) == 0) { + + if ((nh4.nh_flags & (NHF_REDIRECT|NHF_DEFAULT)) == 0) { dest = nh4.nh_addr; /*Router requirements says to only send host redirects*/ type = ICMP_REDIRECT; code = ICMP_REDIRECT_HOST; } } bzero(&ri, sizeof(ri)); error = ip_output(m, NULL, &ri, IP_FORWARDING, NULL, NULL); if (error == EMSGSIZE) mtu = ri.ri_mtu; if (error) IPSTAT_INC(ips_cantforward); else { IPSTAT_INC(ips_forward); if (type) IPSTAT_INC(ips_redirectsent); else { if (mcopy) m_freem(mcopy); return; } } if (mcopy == NULL) { return; } switch (error) { case 0: /* forwarded, but need redirect */ /* type, code set above */ break; case ENETUNREACH: case EHOSTUNREACH: case ENETDOWN: case EHOSTDOWN: default: type = ICMP_UNREACH; code = ICMP_UNREACH_HOST; break; case EMSGSIZE: type = ICMP_UNREACH; code = ICMP_UNREACH_NEEDFRAG; #ifdef IPSEC /* * If IPsec is configured for this path, * override any possibly mtu value set by ip_output. */ mtu = min(ri.ri_mtu, ip_ipsec_mtu(mcopy, mtu)); #endif /* IPSEC */ IPSTAT_INC(ips_cantfrag); break; case ENOBUFS: case EACCES: /* ipfw denied packet */ m_freem(mcopy); return; } icmp_error(mcopy, type, code, dest.s_addr, mtu); } void ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, struct mbuf *m) { if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) { struct bintime bt; bintime(&bt); if (inp->inp_socket->so_options & SO_BINTIME) { *mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt), SCM_BINTIME, SOL_SOCKET); if (*mp) mp = &(*mp)->m_next; } if (inp->inp_socket->so_options & SO_TIMESTAMP) { struct timeval tv; bintime2timeval(&bt, &tv); *mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv), SCM_TIMESTAMP, SOL_SOCKET); if (*mp) mp = &(*mp)->m_next; } } if (inp->inp_flags & INP_RECVDSTADDR) { *mp = sbcreatecontrol((caddr_t)&ip->ip_dst, sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); if (*mp) mp = &(*mp)->m_next; } if (inp->inp_flags & INP_RECVTTL) { *mp = sbcreatecontrol((caddr_t)&ip->ip_ttl, sizeof(u_char), IP_RECVTTL, IPPROTO_IP); if (*mp) mp = &(*mp)->m_next; } #ifdef notyet /* XXX * Moving these out of udp_input() made them even more broken * than they already were. */ /* options were tossed already */ if (inp->inp_flags & INP_RECVOPTS) { *mp = sbcreatecontrol((caddr_t)opts_deleted_above, sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); if (*mp) mp = &(*mp)->m_next; } /* ip_srcroute doesn't do what we want here, need to fix */ if (inp->inp_flags & INP_RECVRETOPTS) { *mp = sbcreatecontrol((caddr_t)ip_srcroute(m), sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); if (*mp) mp = &(*mp)->m_next; } #endif if (inp->inp_flags & INP_RECVIF) { struct ifnet *ifp; struct sdlbuf { struct sockaddr_dl sdl; u_char pad[32]; } sdlbuf; struct sockaddr_dl *sdp; struct sockaddr_dl *sdl2 = &sdlbuf.sdl; if ((ifp = m->m_pkthdr.rcvif) && ifp->if_index && ifp->if_index <= V_if_index) { sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr; /* * Change our mind and don't try copy. */ if (sdp->sdl_family != AF_LINK || sdp->sdl_len > sizeof(sdlbuf)) { goto makedummy; } bcopy(sdp, sdl2, sdp->sdl_len); } else { makedummy: sdl2->sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]); sdl2->sdl_family = AF_LINK; sdl2->sdl_index = 0; sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; } *mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len, IP_RECVIF, IPPROTO_IP); if (*mp) mp = &(*mp)->m_next; } if (inp->inp_flags & INP_RECVTOS) { *mp = sbcreatecontrol((caddr_t)&ip->ip_tos, sizeof(u_char), IP_RECVTOS, IPPROTO_IP); if (*mp) mp = &(*mp)->m_next; } if (inp->inp_flags2 & INP_RECVFLOWID) { uint32_t flowid, flow_type; flowid = m->m_pkthdr.flowid; flow_type = M_HASHTYPE_GET(m); /* * XXX should handle the failure of one or the * other - don't populate both? */ *mp = sbcreatecontrol((caddr_t) &flowid, sizeof(uint32_t), IP_FLOWID, IPPROTO_IP); if (*mp) mp = &(*mp)->m_next; *mp = sbcreatecontrol((caddr_t) &flow_type, sizeof(uint32_t), IP_FLOWTYPE, IPPROTO_IP); if (*mp) mp = &(*mp)->m_next; } #ifdef RSS if (inp->inp_flags2 & INP_RECVRSSBUCKETID) { uint32_t flowid, flow_type; uint32_t rss_bucketid; flowid = m->m_pkthdr.flowid; flow_type = M_HASHTYPE_GET(m); if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) { *mp = sbcreatecontrol((caddr_t) &rss_bucketid, sizeof(uint32_t), IP_RSSBUCKETID, IPPROTO_IP); if (*mp) mp = &(*mp)->m_next; } } #endif } /* * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on * locking. This code remains in ip_input.c as ip_mroute.c is optionally * compiled. */ static VNET_DEFINE(int, ip_rsvp_on); VNET_DEFINE(struct socket *, ip_rsvpd); #define V_ip_rsvp_on VNET(ip_rsvp_on) int ip_rsvp_init(struct socket *so) { if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) return EOPNOTSUPP; if (V_ip_rsvpd != NULL) return EADDRINUSE; V_ip_rsvpd = so; /* * This may seem silly, but we need to be sure we don't over-increment * the RSVP counter, in case something slips up. */ if (!V_ip_rsvp_on) { V_ip_rsvp_on = 1; V_rsvp_on++; } return 0; } int ip_rsvp_done(void) { V_ip_rsvpd = NULL; /* * This may seem silly, but we need to be sure we don't over-decrement * the RSVP counter, in case something slips up. */ if (V_ip_rsvp_on) { V_ip_rsvp_on = 0; V_rsvp_on--; } return 0; } int rsvp_input(struct mbuf **mp, int *offp, int proto) { struct mbuf *m; m = *mp; *mp = NULL; if (rsvp_input_p) { /* call the real one if loaded */ *mp = m; rsvp_input_p(mp, offp, proto); return (IPPROTO_DONE); } /* Can still get packets with rsvp_on = 0 if there is a local member * of the group to which the RSVP packet is addressed. But in this * case we want to throw the packet away. */ if (!V_rsvp_on) { m_freem(m); return (IPPROTO_DONE); } if (V_ip_rsvpd != NULL) { *mp = m; rip_input(mp, offp, proto); return (IPPROTO_DONE); } /* Drop the packet */ m_freem(m); return (IPPROTO_DONE); } Index: projects/routing/sys/netinet/ip_options.c =================================================================== --- projects/routing/sys/netinet/ip_options.c (revision 287053) +++ projects/routing/sys/netinet/ip_options.c (revision 287054) @@ -1,760 +1,760 @@ /* * Copyright (c) 1982, 1986, 1988, 1993 * The Regents of the University of California. * Copyright (c) 2005 Andre Oppermann, Internet Business Solutions AG. * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include "opt_ipstealth.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include static VNET_DEFINE(int, ip_dosourceroute); SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_dosourceroute), 0, "Enable forwarding source routed IP packets"); #define V_ip_dosourceroute VNET(ip_dosourceroute) static VNET_DEFINE(int, ip_acceptsourceroute); SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_acceptsourceroute), 0, "Enable accepting source routed IP packets"); #define V_ip_acceptsourceroute VNET(ip_acceptsourceroute) VNET_DEFINE(int, ip_doopts) = 1; /* 0 = ignore, 1 = process, 2 = reject */ SYSCTL_INT(_net_inet_ip, OID_AUTO, process_options, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_doopts), 0, "Enable IP options processing ([LS]SRR, RR, TS)"); static void save_rte(struct mbuf *m, u_char *, struct in_addr); /* * Do option processing on a datagram, possibly discarding it if bad options * are encountered, or forwarding it if source-routed. * * The pass argument is used when operating in the IPSTEALTH mode to tell * what options to process: [LS]SRR (pass 0) or the others (pass 1). The * reason for as many as two passes is that when doing IPSTEALTH, non-routing * options should be processed only if the packet is for us. * * Returns 1 if packet has been forwarded/freed, 0 if the packet should be * processed further. */ int ip_dooptions(struct mbuf *m, int pass) { struct ip *ip = mtod(m, struct ip *); u_char *cp; struct in_ifaddr *ia; int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; struct in_addr *sin, dst; uint32_t ntime; struct nhop4_extended nh_ext; struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; /* Ignore or reject packets with IP options. */ if (V_ip_doopts == 0) return 0; else if (V_ip_doopts == 2) { type = ICMP_UNREACH; code = ICMP_UNREACH_FILTER_PROHIB; goto bad; } dst = ip->ip_dst; cp = (u_char *)(ip + 1); cnt = (ip->ip_hl << 2) - sizeof (struct ip); for (; cnt > 0; cnt -= optlen, cp += optlen) { opt = cp[IPOPT_OPTVAL]; if (opt == IPOPT_EOL) break; if (opt == IPOPT_NOP) optlen = 1; else { if (cnt < IPOPT_OLEN + sizeof(*cp)) { code = &cp[IPOPT_OLEN] - (u_char *)ip; goto bad; } optlen = cp[IPOPT_OLEN]; if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { code = &cp[IPOPT_OLEN] - (u_char *)ip; goto bad; } } switch (opt) { default: break; /* * Source routing with record. Find interface with current * destination address. If none on this machine then drop if * strictly routed, or do nothing if loosely routed. Record * interface address and bring up next address component. If * strictly routed make sure next address is on directly * accessible net. */ case IPOPT_LSRR: case IPOPT_SSRR: #ifdef IPSTEALTH if (V_ipstealth && pass > 0) break; #endif if (optlen < IPOPT_OFFSET + sizeof(*cp)) { code = &cp[IPOPT_OLEN] - (u_char *)ip; goto bad; } if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { code = &cp[IPOPT_OFFSET] - (u_char *)ip; goto bad; } ipaddr.sin_addr = ip->ip_dst; if (ifa_ifwithaddr_check((struct sockaddr *)&ipaddr) == 0) { if (opt == IPOPT_SSRR) { type = ICMP_UNREACH; code = ICMP_UNREACH_SRCFAIL; goto bad; } if (!V_ip_dosourceroute) goto nosourcerouting; /* * Loose routing, and not at next destination * yet; nothing to do except forward. */ break; } off--; /* 0 origin */ if (off > optlen - (int)sizeof(struct in_addr)) { /* * End of source route. Should be for us. */ if (!V_ip_acceptsourceroute) goto nosourcerouting; save_rte(m, cp, ip->ip_src); break; } #ifdef IPSTEALTH if (V_ipstealth) goto dropit; #endif if (!V_ip_dosourceroute) { if (V_ipforwarding) { char buf[16]; /* aaa.bbb.ccc.ddd\0 */ /* * Acting as a router, so generate * ICMP */ nosourcerouting: 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; } else { /* * Not acting as a router, so * silently drop. */ #ifdef IPSTEALTH dropit: #endif IPSTAT_INC(ips_cantforward); m_freem(m); return (1); } } /* * locate outgoing interface */ (void)memcpy(&ipaddr.sin_addr, cp + off, sizeof(ipaddr.sin_addr)); if (opt == IPOPT_SSRR) { #define INA struct in_ifaddr * #define SA struct sockaddr * ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr, RT_ALL_FIBS); if (ia == NULL) ia = (INA)ifa_ifwithnet((SA)&ipaddr, 0, RT_ALL_FIBS); if (ia == NULL) { type = ICMP_UNREACH; code = ICMP_UNREACH_SRCFAIL; goto bad; } memcpy(cp + off, &(IA_SIN(ia)->sin_addr), sizeof(struct in_addr)); ifa_free(&ia->ia_ifa); } else { /* XXX MRT 0 for routing */ - if (fib4_lookup_nh_extended(M_GETFIB(m), - ipaddr.sin_addr, 0, &nh_ext) != 0) { + if (fib4_lookup_nh_ext(M_GETFIB(m), + ipaddr.sin_addr, 0, 0, &nh_ext) != 0) { type = ICMP_UNREACH; code = ICMP_UNREACH_SRCFAIL; goto bad; } memcpy(cp + off, &nh_ext.nh_src, sizeof(struct in_addr)); } ip->ip_dst = ipaddr.sin_addr; cp[IPOPT_OFFSET] += sizeof(struct in_addr); /* * Let ip_intr's mcast routing check handle mcast pkts */ forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr)); break; case IPOPT_RR: #ifdef IPSTEALTH if (V_ipstealth && pass == 0) break; #endif if (optlen < IPOPT_OFFSET + sizeof(*cp)) { code = &cp[IPOPT_OFFSET] - (u_char *)ip; goto bad; } if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { code = &cp[IPOPT_OFFSET] - (u_char *)ip; goto bad; } /* * If no space remains, ignore. */ off--; /* 0 origin */ if (off > optlen - (int)sizeof(struct in_addr)) break; (void)memcpy(&ipaddr.sin_addr, &ip->ip_dst, sizeof(ipaddr.sin_addr)); /* * Locate outgoing interface; if we're the * destination, use the incoming interface (should be * same). */ if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) != NULL) { memcpy(cp + off, &(IA_SIN(ia)->sin_addr), sizeof(struct in_addr)); ifa_free(&ia->ia_ifa); - } else if (fib4_lookup_nh_extended(M_GETFIB(m), - ipaddr.sin_addr, 0, &nh_ext) == 0) { + } else if (fib4_lookup_nh_ext(M_GETFIB(m), + ipaddr.sin_addr, 0, 0, &nh_ext) == 0) { memcpy(cp + off, &nh_ext.nh_src, sizeof(struct in_addr)); } else { type = ICMP_UNREACH; code = ICMP_UNREACH_HOST; goto bad; } cp[IPOPT_OFFSET] += sizeof(struct in_addr); break; case IPOPT_TS: #ifdef IPSTEALTH if (V_ipstealth && pass == 0) break; #endif code = cp - (u_char *)ip; if (optlen < 4 || optlen > 40) { code = &cp[IPOPT_OLEN] - (u_char *)ip; goto bad; } if ((off = cp[IPOPT_OFFSET]) < 5) { code = &cp[IPOPT_OLEN] - (u_char *)ip; goto bad; } if (off > optlen - (int)sizeof(int32_t)) { cp[IPOPT_OFFSET + 1] += (1 << 4); if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) { code = &cp[IPOPT_OFFSET] - (u_char *)ip; goto bad; } break; } off--; /* 0 origin */ sin = (struct in_addr *)(cp + off); switch (cp[IPOPT_OFFSET + 1] & 0x0f) { case IPOPT_TS_TSONLY: break; case IPOPT_TS_TSANDADDR: if (off + sizeof(uint32_t) + sizeof(struct in_addr) > optlen) { code = &cp[IPOPT_OFFSET] - (u_char *)ip; goto bad; } ipaddr.sin_addr = dst; ia = (INA)ifaof_ifpforaddr((SA)&ipaddr, m->m_pkthdr.rcvif); if (ia == NULL) continue; (void)memcpy(sin, &IA_SIN(ia)->sin_addr, sizeof(struct in_addr)); ifa_free(&ia->ia_ifa); cp[IPOPT_OFFSET] += sizeof(struct in_addr); off += sizeof(struct in_addr); break; case IPOPT_TS_PRESPEC: if (off + sizeof(uint32_t) + sizeof(struct in_addr) > optlen) { code = &cp[IPOPT_OFFSET] - (u_char *)ip; goto bad; } (void)memcpy(&ipaddr.sin_addr, sin, sizeof(struct in_addr)); if (ifa_ifwithaddr_check((SA)&ipaddr) == 0) continue; cp[IPOPT_OFFSET] += sizeof(struct in_addr); off += sizeof(struct in_addr); break; default: code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip; goto bad; } ntime = iptime(); (void)memcpy(cp + off, &ntime, sizeof(uint32_t)); cp[IPOPT_OFFSET] += sizeof(uint32_t); } } if (forward && V_ipforwarding) { ip_forward(m, 1); return (1); } return (0); bad: icmp_error(m, type, code, 0, 0); IPSTAT_INC(ips_badoptions); return (1); } /* * Save incoming source route for use in replies, to be picked up later by * ip_srcroute if the receiver is interested. */ static void save_rte(struct mbuf *m, u_char *option, struct in_addr dst) { unsigned olen; struct ipopt_tag *opts; opts = (struct ipopt_tag *)m_tag_get(PACKET_TAG_IPOPTIONS, sizeof(struct ipopt_tag), M_NOWAIT); if (opts == NULL) return; olen = option[IPOPT_OLEN]; if (olen > sizeof(opts->ip_srcrt) - (1 + sizeof(dst))) { m_tag_free((struct m_tag *)opts); return; } bcopy(option, opts->ip_srcrt.srcopt, olen); opts->ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); opts->ip_srcrt.dst = dst; m_tag_prepend(m, (struct m_tag *)opts); } /* * Retrieve incoming source route for use in replies, in the same form used * by setsockopt. The first hop is placed before the options, will be * removed later. */ struct mbuf * ip_srcroute(struct mbuf *m0) { struct in_addr *p, *q; struct mbuf *m; struct ipopt_tag *opts; opts = (struct ipopt_tag *)m_tag_find(m0, PACKET_TAG_IPOPTIONS, NULL); if (opts == NULL) return (NULL); if (opts->ip_nhops == 0) return (NULL); m = m_get(M_NOWAIT, MT_DATA); if (m == NULL) return (NULL); #define OPTSIZ (sizeof(opts->ip_srcrt.nop) + sizeof(opts->ip_srcrt.srcopt)) /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ m->m_len = opts->ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + OPTSIZ; /* * First, save first hop for return route. */ p = &(opts->ip_srcrt.route[opts->ip_nhops - 1]); *(mtod(m, struct in_addr *)) = *p--; /* * Copy option fields and padding (nop) to mbuf. */ opts->ip_srcrt.nop = IPOPT_NOP; opts->ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; (void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), &(opts->ip_srcrt.nop), OPTSIZ); q = (struct in_addr *)(mtod(m, caddr_t) + sizeof(struct in_addr) + OPTSIZ); #undef OPTSIZ /* * Record return path as an IP source route, reversing the path * (pointers are now aligned). */ while (p >= opts->ip_srcrt.route) { *q++ = *p--; } /* * Last hop goes to final destination. */ *q = opts->ip_srcrt.dst; m_tag_delete(m0, (struct m_tag *)opts); return (m); } /* * Strip out IP options, at higher level protocol in the kernel. */ void ip_stripoptions(struct mbuf *m) { struct ip *ip = mtod(m, struct ip *); int olen; olen = (ip->ip_hl << 2) - sizeof(struct ip); m->m_len -= olen; if (m->m_flags & M_PKTHDR) m->m_pkthdr.len -= olen; ip->ip_len = htons(ntohs(ip->ip_len) - olen); ip->ip_hl = sizeof(struct ip) >> 2; bcopy((char *)ip + sizeof(struct ip) + olen, (ip + 1), (size_t )(m->m_len - sizeof(struct ip))); } /* * Insert IP options into preformed packet. Adjust IP destination as * required for IP source routing, as indicated by a non-zero in_addr at the * start of the options. * * XXX This routine assumes that the packet has no options in place. */ struct mbuf * ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen) { struct ipoption *p = mtod(opt, struct ipoption *); struct mbuf *n; struct ip *ip = mtod(m, struct ip *); unsigned optlen; optlen = opt->m_len - sizeof(p->ipopt_dst); if (optlen + ntohs(ip->ip_len) > IP_MAXPACKET) { *phlen = 0; return (m); /* XXX should fail */ } if (p->ipopt_dst.s_addr) ip->ip_dst = p->ipopt_dst; if (!M_WRITABLE(m) || M_LEADINGSPACE(m) < optlen) { n = m_gethdr(M_NOWAIT, MT_DATA); if (n == NULL) { *phlen = 0; return (m); } m_move_pkthdr(n, m); n->m_pkthdr.rcvif = NULL; n->m_pkthdr.len += optlen; m->m_len -= sizeof(struct ip); m->m_data += sizeof(struct ip); n->m_next = m; m = n; m->m_len = optlen + sizeof(struct ip); m->m_data += max_linkhdr; bcopy(ip, mtod(m, void *), sizeof(struct ip)); } else { m->m_data -= optlen; m->m_len += optlen; m->m_pkthdr.len += optlen; bcopy(ip, mtod(m, void *), sizeof(struct ip)); } ip = mtod(m, struct ip *); bcopy(p->ipopt_list, ip + 1, optlen); *phlen = sizeof(struct ip) + optlen; ip->ip_v = IPVERSION; ip->ip_hl = *phlen >> 2; ip->ip_len = htons(ntohs(ip->ip_len) + optlen); return (m); } /* * Copy options from ip to jp, omitting those not copied during * fragmentation. */ int ip_optcopy(struct ip *ip, struct ip *jp) { u_char *cp, *dp; int opt, optlen, cnt; cp = (u_char *)(ip + 1); dp = (u_char *)(jp + 1); cnt = (ip->ip_hl << 2) - sizeof (struct ip); for (; cnt > 0; cnt -= optlen, cp += optlen) { opt = cp[0]; if (opt == IPOPT_EOL) break; if (opt == IPOPT_NOP) { /* Preserve for IP mcast tunnel's LSRR alignment. */ *dp++ = IPOPT_NOP; optlen = 1; continue; } KASSERT(cnt >= IPOPT_OLEN + sizeof(*cp), ("ip_optcopy: malformed ipv4 option")); optlen = cp[IPOPT_OLEN]; KASSERT(optlen >= IPOPT_OLEN + sizeof(*cp) && optlen <= cnt, ("ip_optcopy: malformed ipv4 option")); /* Bogus lengths should have been caught by ip_dooptions. */ if (optlen > cnt) optlen = cnt; if (IPOPT_COPIED(opt)) { bcopy(cp, dp, optlen); dp += optlen; } } for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) *dp++ = IPOPT_EOL; return (optlen); } /* * Set up IP options in pcb for insertion in output packets. Store in mbuf * with pointer in pcbopt, adding pseudo-option with destination address if * source routed. */ int ip_pcbopts(struct inpcb *inp, int optname, struct mbuf *m) { int cnt, optlen; u_char *cp; struct mbuf **pcbopt; u_char opt; INP_WLOCK_ASSERT(inp); pcbopt = &inp->inp_options; /* turn off any old options */ if (*pcbopt) (void)m_free(*pcbopt); *pcbopt = 0; if (m == NULL || m->m_len == 0) { /* * Only turning off any previous options. */ if (m != NULL) (void)m_free(m); return (0); } if (m->m_len % sizeof(int32_t)) goto bad; /* * IP first-hop destination address will be stored before actual * options; move other options back and clear it when none present. */ if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN]) goto bad; cnt = m->m_len; m->m_len += sizeof(struct in_addr); cp = mtod(m, u_char *) + sizeof(struct in_addr); bcopy(mtod(m, void *), cp, (unsigned)cnt); bzero(mtod(m, void *), sizeof(struct in_addr)); for (; cnt > 0; cnt -= optlen, cp += optlen) { opt = cp[IPOPT_OPTVAL]; if (opt == IPOPT_EOL) break; if (opt == IPOPT_NOP) optlen = 1; else { if (cnt < IPOPT_OLEN + sizeof(*cp)) goto bad; optlen = cp[IPOPT_OLEN]; if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) goto bad; } switch (opt) { default: break; case IPOPT_LSRR: case IPOPT_SSRR: /* * User process specifies route as: * * ->A->B->C->D * * D must be our final destination (but we can't * check that since we may not have connected yet). * A is first hop destination, which doesn't appear * in actual IP option, but is stored before the * options. */ /* XXX-BZ PRIV_NETINET_SETHDROPTS? */ if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr)) goto bad; m->m_len -= sizeof(struct in_addr); cnt -= sizeof(struct in_addr); optlen -= sizeof(struct in_addr); cp[IPOPT_OLEN] = optlen; /* * Move first hop before start of options. */ bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t), sizeof(struct in_addr)); /* * Then copy rest of options back * to close up the deleted entry. */ bcopy((&cp[IPOPT_OFFSET+1] + sizeof(struct in_addr)), &cp[IPOPT_OFFSET+1], (unsigned)cnt - (IPOPT_MINOFF - 1)); break; } } if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr)) goto bad; *pcbopt = m; return (0); bad: (void)m_free(m); return (EINVAL); } /* * Check for the presence of the IP Router Alert option [RFC2113] * in the header of an IPv4 datagram. * * This call is not intended for use from the forwarding path; it is here * so that protocol domains may check for the presence of the option. * Given how FreeBSD's IPv4 stack is currently structured, the Router Alert * option does not have much relevance to the implementation, though this * may change in future. * Router alert options SHOULD be passed if running in IPSTEALTH mode and * we are not the endpoint. * Length checks on individual options should already have been peformed * by ip_dooptions() therefore they are folded under INVARIANTS here. * * Return zero if not present or options are invalid, non-zero if present. */ int ip_checkrouteralert(struct mbuf *m) { struct ip *ip = mtod(m, struct ip *); u_char *cp; int opt, optlen, cnt, found_ra; found_ra = 0; cp = (u_char *)(ip + 1); cnt = (ip->ip_hl << 2) - sizeof (struct ip); for (; cnt > 0; cnt -= optlen, cp += optlen) { opt = cp[IPOPT_OPTVAL]; if (opt == IPOPT_EOL) break; if (opt == IPOPT_NOP) optlen = 1; else { #ifdef INVARIANTS if (cnt < IPOPT_OLEN + sizeof(*cp)) break; #endif optlen = cp[IPOPT_OLEN]; #ifdef INVARIANTS if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) break; #endif } switch (opt) { case IPOPT_RA: #ifdef INVARIANTS if (optlen != IPOPT_OFFSET + sizeof(uint16_t) || (*((uint16_t *)&cp[IPOPT_OFFSET]) != 0)) break; else #endif found_ra = 1; break; default: break; } } return (found_ra); } Index: projects/routing/sys/netinet/ip_output.c =================================================================== --- projects/routing/sys/netinet/ip_output.c (revision 287053) +++ projects/routing/sys/netinet/ip_output.c (revision 287054) @@ -1,1396 +1,1395 @@ /*- * Copyright (c) 1982, 1986, 1988, 1990, 1993 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)ip_output.c 8.3 (Berkeley) 1/21/94 */ #include __FBSDID("$FreeBSD$"); #include "opt_inet.h" #include "opt_ipfw.h" #include "opt_ipsec.h" #include "opt_mbuf_stress_test.h" #include "opt_mpath.h" #include "opt_route.h" #include "opt_sctp.h" #include "opt_rss.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef RADIX_MPATH #include #endif #include #include #include #include #include #include #include #include #include #include #include #ifdef SCTP #include #include #endif #include #ifdef IPSEC #include #include #endif /* IPSEC*/ #include #include #ifdef MBUF_STRESS_TEST static int mbuf_frag_size = 0; SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW, &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size"); #endif static void ip_mloopback (struct ifnet *, struct mbuf *, int); extern int in_mcast_loop; extern struct protosw inetsw[]; static inline int ip_output_pfil(struct mbuf *m, struct ifnet *ifp, struct inpcb *inp, struct in_addr *dst, int *fibnum, int *error) { struct m_tag *fwd_tag = NULL; struct in_addr odst; struct ip *ip; struct sockaddr_in *dst_sa; ip = mtod(m, struct ip *); /* Run through list of hooks for output packets. */ odst.s_addr = ip->ip_dst.s_addr; *error = pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_OUT, inp); if ((*error) != 0 || m == NULL) return 1; /* Finished */ ip = mtod(m, struct ip *); /* See if destination IP address was changed by packet filter. */ if (odst.s_addr != ip->ip_dst.s_addr) { m->m_flags |= M_SKIP_FIREWALL; /* If destination is now ourself drop to ip_input(). */ if (in_localip(ip->ip_dst)) { m->m_flags |= M_FASTFWD_OURS; if (m->m_pkthdr.rcvif == NULL) m->m_pkthdr.rcvif = V_loif; if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { m->m_pkthdr.csum_flags |= CSUM_DATA_VALID | CSUM_PSEUDO_HDR; m->m_pkthdr.csum_data = 0xffff; } m->m_pkthdr.csum_flags |= CSUM_IP_CHECKED | CSUM_IP_VALID; #ifdef SCTP if (m->m_pkthdr.csum_flags & CSUM_SCTP) m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID; #endif *error = netisr_queue(NETISR_IP, m); return 1; /* Finished */ } *dst = ip->ip_dst; return -1; /* Reloop */ } /* See if fib was changed by packet filter. */ if ((*fibnum) != M_GETFIB(m)) { m->m_flags |= M_SKIP_FIREWALL; *fibnum = M_GETFIB(m); return -1; /* Reloop for FIB change */ } /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */ if (m->m_flags & M_FASTFWD_OURS) { if (m->m_pkthdr.rcvif == NULL) m->m_pkthdr.rcvif = V_loif; if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { m->m_pkthdr.csum_flags |= CSUM_DATA_VALID | CSUM_PSEUDO_HDR; m->m_pkthdr.csum_data = 0xffff; } #ifdef SCTP if (m->m_pkthdr.csum_flags & CSUM_SCTP) m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID; #endif m->m_pkthdr.csum_flags |= CSUM_IP_CHECKED | CSUM_IP_VALID; *error = netisr_queue(NETISR_IP, m); return 1; /* Finished */ } /* Or forward to some other address? */ if ((m->m_flags & M_IP_NEXTHOP) && ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) { dst_sa = (struct sockaddr_in *)(fwd_tag + 1); bzero(dst_sa, sizeof(*dst_sa)); dst_sa->sin_family = AF_INET; dst_sa->sin_len = sizeof(*dst_sa); dst_sa->sin_addr = *dst; m->m_flags |= M_SKIP_FIREWALL; m->m_flags &= ~M_IP_NEXTHOP; m_tag_delete(m, fwd_tag); return -1; /* Reloop for CHANGE of dst */ } return 0; } /* * IP output. The packet in mbuf chain m contains a skeletal IP * header (with len, off, ttl, proto, tos, src, dst). * The mbuf chain containing the packet will be freed. * The mbuf opt, if present, will not be freed. * If route ro is present and has ro_rt initialized, route lookup would be * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL, * then result of route lookup is stored in ro->ro_rt. * * In the IP forwarding case, the packet will arrive with options already * inserted, so must have a NULL opt pointer. */ int ip_output(struct mbuf *m, struct mbuf *opt, struct route_info *ri, int flags, struct ip_moptions *imo, struct inpcb *inp) { struct rm_priotracker in_ifa_tracker; struct ip *ip; struct ifnet *ifp = NULL; /* keep compiler happy */ struct mbuf *m0; int hlen = sizeof (struct ip); int mtu; int error = 0; struct in_addr dst, local_addr; struct sockaddr_in gw_out; struct in_ifaddr *ia; int isbroadcast; uint16_t ip_len, ip_off; - struct nhop_data local_nh, *nh; + struct nhop_prepend local_nh, *nh; struct nhop4_extended nhe, *pnhe; uint32_t fibnum; #ifdef IPSEC int no_route_but_check_spd = 0; #endif M_ASSERTPKTHDR(m); if (inp != NULL) { INP_LOCK_ASSERT(inp); M_SETFIB(m, inp->inp_inc.inc_fibnum); if ((flags & IP_NODEFAULTFLOWID) == 0) { m->m_pkthdr.flowid = inp->inp_flowid; M_HASHTYPE_SET(m, inp->inp_flowtype); } } nh = NULL; #ifdef FLOWTABLE if (ri->ri_nh == NULL) (void )flowtable_lookup(AF_INET, m, ri); #endif if (opt) { int len = 0; m = ip_insertoptions(m, opt, &len); if (len != 0) hlen = len; /* ip->ip_hl is updated above */ } ip = mtod(m, struct ip *); ip_len = ntohs(ip->ip_len); ip_off = ntohs(ip->ip_off); if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { ip->ip_v = IPVERSION; ip->ip_hl = hlen >> 2; ip_fillid(ip); IPSTAT_INC(ips_localout); } else { /* Header already set, fetch hlen from there */ hlen = ip->ip_hl << 2; } /* * dst/gw handling: * * We start with @dst poiting to IPv4 destination. * Pfil output hook can force us to repeat IPv4 lookup, * in that case we either re-read IPv4 header or use * provided @dst from ip_output_pfil. * * While sending, we have 2 cases: * either we are doing broadcast/L2 multicast/route to IF - * in that case there is no special gateway. * or routing lookup - in that case we either have full * prepend header so we don't need any L3 info anymore, * or non-existent/imcomplete ARP entry - in that case * we have GW info inside @nh. */ /* Make compiler happy */ dst.s_addr = INADDR_ANY; fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m); again: local_addr = ip->ip_src; /* * If routing to interface only, short circuit routing lookup. * The use of an all-ones broadcast address implies this; an * interface is specified by the broadcast address of an interface, * or the destination address of a ptp interface. */ if (flags & IP_SENDONES) { /* Use sockaddr API */ memset(&gw_out, 0, sizeof(gw_out)); gw_out.sin_len = sizeof(gw_out); gw_out.sin_family = AF_INET; gw_out.sin_addr = dst; if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(&gw_out), M_GETFIB(m)))) == NULL && (ia = ifatoia(ifa_ifwithdstaddr(sintosa(&gw_out), M_GETFIB(m)))) == NULL) { IPSTAT_INC(ips_noroute); error = ENETUNREACH; goto bad; } local_addr = IA_SIN(ia)->sin_addr; ifp = ia->ia_ifp; mtu = ifp->if_mtu; ifa_free(&ia->ia_ifa); ia = NULL; ip->ip_dst.s_addr = INADDR_BROADCAST; dst = ip->ip_dst; ip->ip_ttl = 1; isbroadcast = 1; } else if (flags & IP_ROUTETOIF) { /* Use sockaddr API */ memset(&gw_out, 0, sizeof(gw_out)); gw_out.sin_len = sizeof(gw_out); gw_out.sin_family = AF_INET; gw_out.sin_addr = dst; if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(&gw_out), M_GETFIB(m)))) == NULL && (ia = ifatoia(ifa_ifwithnet(sintosa(&gw_out), 0, M_GETFIB(m)))) == NULL) { IPSTAT_INC(ips_noroute); error = ENETUNREACH; goto bad; } local_addr = IA_SIN(ia)->sin_addr; ifp = ia->ia_ifp; mtu = ifp->if_mtu; ifa_free(&ia->ia_ifa); ia = NULL; ip->ip_ttl = 1; isbroadcast = in_broadcast(dst, ifp); } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) && imo != NULL && imo->imo_multicast_ifp != NULL) { /* * Bypass the normal routing lookup for multicast * packets if the interface is specified. */ ifp = imo->imo_multicast_ifp; mtu = ifp->if_mtu; IFP_TO_IA(ifp, ia, &in_ifa_tracker); if (ia != NULL) { local_addr = IA_SIN(ia)->sin_addr; ifa_free(&ia->ia_ifa); ia = NULL; } isbroadcast = 0; /* fool gcc */ } else { /* * We have to either use provided cached nexthop * or do route lookup. */ if (local_addr.s_addr != INADDR_ANY) pnhe = NULL; else pnhe = &nhe; if (ri == NULL || ri->ri_nh == NULL) { /* * Do resolve ourselves */ error = fib4_lookup_prepend(fibnum, dst, m, &local_nh, pnhe); } else { /* * Got cached nexhop. Note that nexthop might * be multipath so we need to need to "convert" * it first. Assume caller provide locking guarantees * for that. */ fib4_choose_prepend(fibnum, ri->ri_nh, m->m_pkthdr.flowid, &local_nh, pnhe); } if (error != 0) { #ifdef IPSEC /* * There is no route for this packet, but it is * possible that a matching SPD entry exists. */ no_route_but_check_spd = 1; mtu = 0; /* Silence GCC warning. */ goto sendit; #endif IPSTAT_INC(ips_noroute); goto bad; } /* Get local address from extended nexthop info */ if (pnhe != NULL) local_addr = nhe.nh_src; nh = &local_nh; ifp = NH_LIFP(nh); mtu = nh->nh_mtu; if (nh->nh_flags & (RTF_HOST|RTF_GATEWAY)) { /* XXX: Set RTF_BROADCAST if GW address is broadcast */ isbroadcast = (nh->nh_flags & RTF_BROADCAST); } else isbroadcast = in_broadcast(dst, ifp); } /* * XXX: Move somewhere to sendit */ if (ri != NULL) ri->ri_mtu = mtu; /* Catch a possible divide by zero later. */ KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p", __func__, mtu, nh, (nh != NULL) ? nh->nh_flags : 0, ifp)); /* * If source address not specified yet, use address * of outgoing interface. */ if (ip->ip_src.s_addr == INADDR_ANY) ip->ip_src = local_addr; if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { m->m_flags |= M_MCAST; /* IP destination address is multicast. */ /* * See if the caller provided any multicast options */ if (imo != NULL) { ip->ip_ttl = imo->imo_multicast_ttl; if (imo->imo_multicast_vif != -1) ip->ip_src.s_addr = ip_mcast_src ? ip_mcast_src(imo->imo_multicast_vif) : INADDR_ANY; } else ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; /* * Confirm that the outgoing interface supports multicast. */ if ((imo == NULL) || (imo->imo_multicast_vif == -1)) { if ((ifp->if_flags & IFF_MULTICAST) == 0) { IPSTAT_INC(ips_noroute); error = ENETUNREACH; goto bad; } } if ((imo == NULL && in_mcast_loop) || (imo && imo->imo_multicast_loop)) { /* * Loop back multicast datagram if not expressly * forbidden to do so, even if we are not a member * of the group; ip_input() will filter it later, * thus deferring a hash lookup and mutex acquisition * at the expense of a cheap copy using m_copym(). */ ip_mloopback(ifp, m, hlen); } else { /* * If we are acting as a multicast router, perform * multicast forwarding as if the packet had just * arrived on the interface to which we are about * to send. The multicast forwarding function * recursively calls this function, using the * IP_FORWARDING flag to prevent infinite recursion. * * Multicasts that are looped back by ip_mloopback(), * above, will be forwarded by the ip_input() routine, * if necessary. */ if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) { /* * If rsvp daemon is not running, do not * set ip_moptions. This ensures that the packet * is multicast and not just sent down one link * as prescribed by rsvpd. */ if (!V_rsvp_on) imo = NULL; if (ip_mforward && ip_mforward(ip, ifp, m, imo) != 0) { m_freem(m); goto done; } } } /* * Multicasts with a time-to-live of zero may be looped- * back, above, but must not be transmitted on a network. * Also, multicasts addressed to the loopback interface * are not sent -- the above call to ip_mloopback() will * loop back a copy. ip_input() will drop the copy if * this host does not belong to the destination group on * the loopback interface. */ if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) { m_freem(m); goto done; } goto sendit; } /* * Look for broadcast address and * verify user is allowed to send * such a packet. */ if (isbroadcast) { if ((ifp->if_flags & IFF_BROADCAST) == 0) { error = EADDRNOTAVAIL; goto bad; } if ((flags & IP_ALLOWBROADCAST) == 0) { error = EACCES; goto bad; } /* don't allow broadcast messages to be fragmented */ if (ip_len > mtu) { error = EMSGSIZE; goto bad; } m->m_flags |= M_BCAST; } else { m->m_flags &= ~M_BCAST; } sendit: #ifdef IPSEC switch(ip_ipsec_output(&m, inp, &error)) { case 1: goto bad; case -1: goto done; case 0: default: break; /* Continue with packet processing. */ } /* * Check if there was a route for this packet; return error if not. */ if (no_route_but_check_spd) { IPSTAT_INC(ips_noroute); error = EHOSTUNREACH; goto bad; } /* Update variables that are affected by ipsec4_output(). */ ip = mtod(m, struct ip *); hlen = ip->ip_hl << 2; #endif /* IPSEC */ /* Jump over all PFIL processing if hooks are not active. */ if (PFIL_HOOKED(&V_inet_pfil_hook)) { switch (ip_output_pfil(m, ifp, inp, &dst, &fibnum, &error)) { case 1: /* Finished */ goto done; case 0: /* Continue normally */ ip = mtod(m, struct ip *); break; case -1: /* Need to try again */ /* Reset everything for a new round */ /* TODO: Carefully inspect multipath cached route case */ if (nh != NULL) { - fib4_free_nh(fibnum, nh); + fib4_free_nh_prepend(fibnum, nh); nh = NULL; } ri = NULL; ip = mtod(m, struct ip *); goto again; - } } /* 127/8 must not appear on wire - RFC1122. */ if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { if ((ifp->if_flags & IFF_LOOPBACK) == 0) { IPSTAT_INC(ips_badaddr); error = EADDRNOTAVAIL; goto bad; } } m->m_pkthdr.csum_flags |= CSUM_IP; if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) { in_delayed_cksum(m); m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; } #ifdef SCTP if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) { sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2)); m->m_pkthdr.csum_flags &= ~CSUM_SCTP; } #endif /* * If small enough for interface, or the interface will take * care of the fragmentation for us, we can just send directly. */ if (ip_len <= mtu || (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0) { ip->ip_sum = 0; if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) { ip->ip_sum = in_cksum(m, hlen); m->m_pkthdr.csum_flags &= ~CSUM_IP; } #if 0 /* * Record statistics for this interface address. * With CSUM_TSO the byte/packet count will be slightly * incorrect because we count the IP+TCP headers only * once instead of for every generated packet. */ if (!(flags & IP_FORWARDING) && ia) { if (m->m_pkthdr.csum_flags & CSUM_TSO) counter_u64_add(ia->ia_ifa.ifa_opackets, m->m_pkthdr.len / m->m_pkthdr.tso_segsz); else counter_u64_add(ia->ia_ifa.ifa_opackets, 1); counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len); } #endif #ifdef MBUF_STRESS_TEST if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size) m = m_fragment(m, M_NOWAIT, mbuf_frag_size); #endif /* * Reset layer specific mbuf flags * to avoid confusing lower layers. */ m_clrprotoflags(m); IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL); error = fib4_sendmbuf(ifp, m, nh, dst); goto done; } /* Balk when DF bit is set or the interface didn't support TSO. */ if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) { error = EMSGSIZE; IPSTAT_INC(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, mtu, ifp->if_hwassist); if (error) goto bad; for (; m; m = m0) { m0 = m->m_nextpkt; m->m_nextpkt = 0; if (error == 0) { #if 0 /* Record statistics for this interface address. */ if (ia != NULL) { counter_u64_add(ia->ia_ifa.ifa_opackets, 1); counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len); } #endif /* * Reset layer specific mbuf flags * to avoid confusing upper layers. */ m_clrprotoflags(m); IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL); error = fib4_sendmbuf(ifp, m, nh, dst); } else m_freem(m); } if (error == 0) IPSTAT_INC(ips_fragmented); done: /* TODO: Carefully inspect multipath cached route case */ if (nh != NULL) - fib4_free_nh(fibnum, nh); + fib4_free_nh_prepend(fibnum, nh); 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 error, m_frag may contain a partially built * chain of fragments that should be freed by the caller. * * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist) */ int ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu, u_long if_hwassist_flags) { 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; uint16_t ip_len, ip_off; ip_len = ntohs(ip->ip_len); ip_off = ntohs(ip->ip_off); if (ip_off & IP_DF) { /* Fragmentation not allowed */ IPSTAT_INC(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) { in_delayed_cksum(m0); m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; } #ifdef SCTP if (m0->m_pkthdr.csum_flags & CSUM_SCTP) { sctp_delayed_cksum(m0, hlen); m0->m_pkthdr.csum_flags &= ~CSUM_SCTP; } #endif if (len > PAGE_SIZE) { /* * Fragment 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; off = MIN(mtu, m0->m_pkthdr.len); /* * firstlen (off - hlen) must be aligned on an * 8-byte boundary */ if (off < hlen) goto smart_frag_failure; off = ((off - hlen) & ~7) + hlen; newlen = (~PAGE_MASK) & mtu; if ((newlen + sizeof (struct ip)) > mtu) { /* we failed, go back the default */ smart_frag_failure: newlen = len; off = hlen + len; } len = newlen; } else { off = hlen + len; } firstlen = off - hlen; mnext = &m0->m_nextpkt; /* pointer to next packet */ /* * 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_len; off += len, nfrags++) { struct ip *mhip; /* ip header on the fragment */ struct mbuf *m; int mhlen = sizeof (struct ip); m = m_gethdr(M_NOWAIT, MT_DATA); if (m == NULL) { error = ENOBUFS; IPSTAT_INC(ips_odropped); goto done; } /* * Make sure the complete packet header gets copied * from the originating mbuf to the newly created * mbuf. This also ensures that existing firewall * classification(s), VLAN tags and so on get copied * to the resulting fragmented packet(s): */ if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) { m_free(m); error = ENOBUFS; IPSTAT_INC(ips_odropped); goto done; } /* * 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_copym(). */ m->m_data += max_linkhdr; mhip = mtod(m, struct ip *); *mhip = *ip; if (hlen > sizeof (struct ip)) { mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); mhip->ip_v = IPVERSION; mhip->ip_hl = mhlen >> 2; } m->m_len = mhlen; /* XXX do we need to add ip_off below ? */ mhip->ip_off = ((off - hlen) >> 3) + ip_off; if (off + len >= ip_len) len = ip_len - off; else mhip->ip_off |= IP_MF; mhip->ip_len = htons((u_short)(len + mhlen)); m->m_next = m_copym(m0, off, len, M_NOWAIT); if (m->m_next == NULL) { /* copy failed */ m_free(m); error = ENOBUFS; /* ??? */ IPSTAT_INC(ips_odropped); goto done; } m->m_pkthdr.len = mhlen + len; #ifdef MAC mac_netinet_fragment(m0, m); #endif mhip->ip_off = htons(mhip->ip_off); mhip->ip_sum = 0; if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) { mhip->ip_sum = in_cksum(m, mhlen); m->m_pkthdr.csum_flags &= ~CSUM_IP; } *mnext = m; mnext = &m->m_nextpkt; } IPSTAT_ADD(ips_ofragments, nfrags); /* * Update first fragment by trimming what's been copied out * and updating header. */ m_adj(m0, hlen + firstlen - ip_len); m0->m_pkthdr.len = hlen + firstlen; ip->ip_len = htons((u_short)m0->m_pkthdr.len); ip->ip_off = htons(ip_off | IP_MF); ip->ip_sum = 0; if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) { ip->ip_sum = in_cksum(m0, hlen); m0->m_pkthdr.csum_flags &= ~CSUM_IP; } done: *m_frag = m0; return error; } void in_delayed_cksum(struct mbuf *m) { struct ip *ip; uint16_t csum, offset, ip_len; ip = mtod(m, struct ip *); offset = ip->ip_hl << 2 ; ip_len = ntohs(ip->ip_len); csum = in_cksum_skip(m, ip_len, offset); if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0) csum = 0xffff; offset += m->m_pkthdr.csum_data; /* checksum offset */ /* find the mbuf in the chain where the checksum starts*/ while ((m != NULL) && (offset >= m->m_len)) { offset -= m->m_len; m = m->m_next; } KASSERT(m != NULL, ("in_delayed_cksum: checksum outside mbuf chain.")); KASSERT(offset + sizeof(u_short) <= m->m_len, ("in_delayed_cksum: checksum split between mbufs.")); *(u_short *)(m->m_data + offset) = csum; } /* * IP socket option processing. */ int ip_ctloutput(struct socket *so, struct sockopt *sopt) { struct inpcb *inp = sotoinpcb(so); int error, optval; #ifdef RSS uint32_t rss_bucket; int retval; #endif error = optval = 0; if (sopt->sopt_level != IPPROTO_IP) { error = EINVAL; if (sopt->sopt_level == SOL_SOCKET && sopt->sopt_dir == SOPT_SET) { switch (sopt->sopt_name) { case SO_REUSEADDR: INP_WLOCK(inp); if ((so->so_options & SO_REUSEADDR) != 0) inp->inp_flags2 |= INP_REUSEADDR; else inp->inp_flags2 &= ~INP_REUSEADDR; INP_WUNLOCK(inp); error = 0; break; case SO_REUSEPORT: INP_WLOCK(inp); if ((so->so_options & SO_REUSEPORT) != 0) inp->inp_flags2 |= INP_REUSEPORT; else inp->inp_flags2 &= ~INP_REUSEPORT; INP_WUNLOCK(inp); error = 0; break; case SO_SETFIB: INP_WLOCK(inp); inp->inp_inc.inc_fibnum = so->so_fibnum; INP_WUNLOCK(inp); error = 0; break; default: break; } } return (error); } switch (sopt->sopt_dir) { case SOPT_SET: switch (sopt->sopt_name) { case IP_OPTIONS: #ifdef notyet case IP_RETOPTS: #endif { struct mbuf *m; if (sopt->sopt_valsize > MLEN) { error = EMSGSIZE; break; } m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA); if (m == NULL) { error = ENOBUFS; break; } m->m_len = sopt->sopt_valsize; error = sooptcopyin(sopt, mtod(m, char *), m->m_len, m->m_len); if (error) { m_free(m); break; } INP_WLOCK(inp); error = ip_pcbopts(inp, sopt->sopt_name, m); INP_WUNLOCK(inp); return (error); } case IP_BINDANY: if (sopt->sopt_td != NULL) { error = priv_check(sopt->sopt_td, PRIV_NETINET_BINDANY); if (error) break; } /* FALLTHROUGH */ case IP_BINDMULTI: #ifdef RSS case IP_RSS_LISTEN_BUCKET: #endif case IP_TOS: case IP_TTL: case IP_MINTTL: case IP_RECVOPTS: case IP_RECVRETOPTS: case IP_RECVDSTADDR: case IP_RECVTTL: case IP_RECVIF: case IP_ONESBCAST: case IP_DONTFRAG: case IP_RECVTOS: case IP_RECVFLOWID: #ifdef RSS case IP_RECVRSSBUCKETID: #endif error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); if (error) break; switch (sopt->sopt_name) { case IP_TOS: inp->inp_ip_tos = optval; break; case IP_TTL: inp->inp_ip_ttl = optval; break; case IP_MINTTL: if (optval >= 0 && optval <= MAXTTL) inp->inp_ip_minttl = optval; else error = EINVAL; break; #define OPTSET(bit) do { \ INP_WLOCK(inp); \ if (optval) \ inp->inp_flags |= bit; \ else \ inp->inp_flags &= ~bit; \ INP_WUNLOCK(inp); \ } while (0) #define OPTSET2(bit, val) do { \ INP_WLOCK(inp); \ if (val) \ inp->inp_flags2 |= bit; \ else \ inp->inp_flags2 &= ~bit; \ INP_WUNLOCK(inp); \ } while (0) case IP_RECVOPTS: OPTSET(INP_RECVOPTS); break; case IP_RECVRETOPTS: OPTSET(INP_RECVRETOPTS); break; case IP_RECVDSTADDR: OPTSET(INP_RECVDSTADDR); break; case IP_RECVTTL: OPTSET(INP_RECVTTL); break; case IP_RECVIF: OPTSET(INP_RECVIF); break; case IP_ONESBCAST: OPTSET(INP_ONESBCAST); break; case IP_DONTFRAG: OPTSET(INP_DONTFRAG); break; case IP_BINDANY: OPTSET(INP_BINDANY); break; case IP_RECVTOS: OPTSET(INP_RECVTOS); break; case IP_BINDMULTI: OPTSET2(INP_BINDMULTI, optval); break; case IP_RECVFLOWID: OPTSET2(INP_RECVFLOWID, optval); break; #ifdef RSS case IP_RSS_LISTEN_BUCKET: if ((optval >= 0) && (optval < rss_getnumbuckets())) { inp->inp_rss_listen_bucket = optval; OPTSET2(INP_RSS_BUCKET_SET, 1); } else { error = EINVAL; } break; case IP_RECVRSSBUCKETID: OPTSET2(INP_RECVRSSBUCKETID, optval); break; #endif } break; #undef OPTSET #undef OPTSET2 /* * Multicast socket options are processed by the in_mcast * module. */ case IP_MULTICAST_IF: case IP_MULTICAST_VIF: case IP_MULTICAST_TTL: case IP_MULTICAST_LOOP: case IP_ADD_MEMBERSHIP: case IP_DROP_MEMBERSHIP: case IP_ADD_SOURCE_MEMBERSHIP: case IP_DROP_SOURCE_MEMBERSHIP: case IP_BLOCK_SOURCE: case IP_UNBLOCK_SOURCE: case IP_MSFILTER: case MCAST_JOIN_GROUP: case MCAST_LEAVE_GROUP: case MCAST_JOIN_SOURCE_GROUP: case MCAST_LEAVE_SOURCE_GROUP: case MCAST_BLOCK_SOURCE: case MCAST_UNBLOCK_SOURCE: error = inp_setmoptions(inp, sopt); break; case IP_PORTRANGE: error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); if (error) break; INP_WLOCK(inp); switch (optval) { case IP_PORTRANGE_DEFAULT: inp->inp_flags &= ~(INP_LOWPORT); inp->inp_flags &= ~(INP_HIGHPORT); break; case IP_PORTRANGE_HIGH: inp->inp_flags &= ~(INP_LOWPORT); inp->inp_flags |= INP_HIGHPORT; break; case IP_PORTRANGE_LOW: inp->inp_flags &= ~(INP_HIGHPORT); inp->inp_flags |= INP_LOWPORT; break; default: error = EINVAL; break; } INP_WUNLOCK(inp); break; #ifdef IPSEC case IP_IPSEC_POLICY: { caddr_t req; struct mbuf *m; if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */ break; if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */ break; req = mtod(m, caddr_t); error = ipsec_set_policy(inp, sopt->sopt_name, req, m->m_len, (sopt->sopt_td != NULL) ? sopt->sopt_td->td_ucred : NULL); m_freem(m); break; } #endif /* IPSEC */ default: error = ENOPROTOOPT; break; } break; case SOPT_GET: switch (sopt->sopt_name) { case IP_OPTIONS: case IP_RETOPTS: if (inp->inp_options) error = sooptcopyout(sopt, mtod(inp->inp_options, char *), inp->inp_options->m_len); else sopt->sopt_valsize = 0; break; case IP_TOS: case IP_TTL: case IP_MINTTL: case IP_RECVOPTS: case IP_RECVRETOPTS: case IP_RECVDSTADDR: case IP_RECVTTL: case IP_RECVIF: case IP_PORTRANGE: case IP_ONESBCAST: case IP_DONTFRAG: case IP_BINDANY: case IP_RECVTOS: case IP_BINDMULTI: case IP_FLOWID: case IP_FLOWTYPE: case IP_RECVFLOWID: #ifdef RSS case IP_RSSBUCKETID: case IP_RECVRSSBUCKETID: #endif switch (sopt->sopt_name) { case IP_TOS: optval = inp->inp_ip_tos; break; case IP_TTL: optval = inp->inp_ip_ttl; break; case IP_MINTTL: optval = inp->inp_ip_minttl; break; #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) #define OPTBIT2(bit) (inp->inp_flags2 & bit ? 1 : 0) case IP_RECVOPTS: optval = OPTBIT(INP_RECVOPTS); break; case IP_RECVRETOPTS: optval = OPTBIT(INP_RECVRETOPTS); break; case IP_RECVDSTADDR: optval = OPTBIT(INP_RECVDSTADDR); break; case IP_RECVTTL: optval = OPTBIT(INP_RECVTTL); break; case IP_RECVIF: optval = OPTBIT(INP_RECVIF); break; case IP_PORTRANGE: if (inp->inp_flags & INP_HIGHPORT) optval = IP_PORTRANGE_HIGH; else if (inp->inp_flags & INP_LOWPORT) optval = IP_PORTRANGE_LOW; else optval = 0; break; case IP_ONESBCAST: optval = OPTBIT(INP_ONESBCAST); break; case IP_DONTFRAG: optval = OPTBIT(INP_DONTFRAG); break; case IP_BINDANY: optval = OPTBIT(INP_BINDANY); break; case IP_RECVTOS: optval = OPTBIT(INP_RECVTOS); break; case IP_FLOWID: optval = inp->inp_flowid; break; case IP_FLOWTYPE: optval = inp->inp_flowtype; break; case IP_RECVFLOWID: optval = OPTBIT2(INP_RECVFLOWID); break; #ifdef RSS case IP_RSSBUCKETID: retval = rss_hash2bucket(inp->inp_flowid, inp->inp_flowtype, &rss_bucket); if (retval == 0) optval = rss_bucket; else error = EINVAL; break; case IP_RECVRSSBUCKETID: optval = OPTBIT2(INP_RECVRSSBUCKETID); break; #endif case IP_BINDMULTI: optval = OPTBIT2(INP_BINDMULTI); break; } error = sooptcopyout(sopt, &optval, sizeof optval); break; /* * Multicast socket options are processed by the in_mcast * module. */ case IP_MULTICAST_IF: case IP_MULTICAST_VIF: case IP_MULTICAST_TTL: case IP_MULTICAST_LOOP: case IP_MSFILTER: error = inp_getmoptions(inp, sopt); break; #ifdef IPSEC case IP_IPSEC_POLICY: { struct mbuf *m = NULL; caddr_t req = NULL; size_t len = 0; if (m != 0) { req = mtod(m, caddr_t); len = m->m_len; } error = ipsec_get_policy(sotoinpcb(so), req, len, &m); if (error == 0) error = soopt_mcopyout(sopt, m); /* XXX */ if (error == 0) m_freem(m); break; } #endif /* IPSEC */ default: error = ENOPROTOOPT; break; } break; } return (error); } /* * Routine called from ip_output() to loop back a copy of an IP multicast * packet to the input queue of a specified interface. Note that this * calls the output routine of the loopback "driver", but with an interface * pointer that might NOT be a loopback interface -- evil, but easier than * replicating that code here. */ static void ip_mloopback(struct ifnet *ifp, struct mbuf *m, int hlen) { register struct ip *ip; struct mbuf *copym; /* * Make a deep copy of the packet because we're going to * modify the pack in order to generate checksums. */ copym = m_dup(m, M_NOWAIT); if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen)) copym = m_pullup(copym, hlen); if (copym != NULL) { /* If needed, compute the checksum and mark it as valid. */ if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { in_delayed_cksum(copym); copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID | CSUM_PSEUDO_HDR; copym->m_pkthdr.csum_data = 0xffff; } /* * We don't bother to fragment if the IP length is greater * than the interface's MTU. Can this possibly matter? */ ip = mtod(copym, struct ip *); ip->ip_sum = 0; ip->ip_sum = in_cksum(copym, hlen); if_simloop(ifp, copym, AF_INET, 0); } } Index: projects/routing/sys/netinet/tcp_subr.c =================================================================== --- projects/routing/sys/netinet/tcp_subr.c (revision 287053) +++ projects/routing/sys/netinet/tcp_subr.c (revision 287054) @@ -1,2517 +1,2517 @@ /*- * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)tcp_subr.c 8.2 (Berkeley) 5/24/95 */ #include __FBSDID("$FreeBSD$"); #include "opt_compat.h" #include "opt_inet.h" #include "opt_inet6.h" #include "opt_ipsec.h" #include "opt_tcpdebug.h" #include #include #include #include #include #include #include #include #include #include #ifdef INET6 #include #endif #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef INET6 #include #include #include #include #include #endif #include #include #include #include #include #include #ifdef INET6 #include #endif #include #ifdef TCPDEBUG #include #endif #ifdef INET6 #include #endif #ifdef TCP_OFFLOAD #include #endif #ifdef IPSEC #include #include #ifdef INET6 #include #endif #include #include #endif /*IPSEC*/ #include #include #include VNET_DEFINE(int, tcp_mssdflt) = TCP_MSS; #ifdef INET6 VNET_DEFINE(int, tcp_v6mssdflt) = TCP6_MSS; #endif static int sysctl_net_inet_tcp_mss_check(SYSCTL_HANDLER_ARGS) { int error, new; new = V_tcp_mssdflt; error = sysctl_handle_int(oidp, &new, 0, req); if (error == 0 && req->newptr) { if (new < TCP_MINMSS) error = EINVAL; else V_tcp_mssdflt = new; } return (error); } SYSCTL_PROC(_net_inet_tcp, TCPCTL_MSSDFLT, mssdflt, CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW, &VNET_NAME(tcp_mssdflt), 0, &sysctl_net_inet_tcp_mss_check, "I", "Default TCP Maximum Segment Size"); #ifdef INET6 static int sysctl_net_inet_tcp_mss_v6_check(SYSCTL_HANDLER_ARGS) { int error, new; new = V_tcp_v6mssdflt; error = sysctl_handle_int(oidp, &new, 0, req); if (error == 0 && req->newptr) { if (new < TCP_MINMSS) error = EINVAL; else V_tcp_v6mssdflt = new; } return (error); } SYSCTL_PROC(_net_inet_tcp, TCPCTL_V6MSSDFLT, v6mssdflt, CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW, &VNET_NAME(tcp_v6mssdflt), 0, &sysctl_net_inet_tcp_mss_v6_check, "I", "Default TCP Maximum Segment Size for IPv6"); #endif /* INET6 */ /* * Minimum MSS we accept and use. This prevents DoS attacks where * we are forced to a ridiculous low MSS like 20 and send hundreds * of packets instead of one. The effect scales with the available * bandwidth and quickly saturates the CPU and network interface * with packet generation and sending. Set to zero to disable MINMSS * checking. This setting prevents us from sending too small packets. */ VNET_DEFINE(int, tcp_minmss) = TCP_MINMSS; SYSCTL_INT(_net_inet_tcp, OID_AUTO, minmss, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(tcp_minmss), 0, "Minimum TCP Maximum Segment Size"); VNET_DEFINE(int, tcp_do_rfc1323) = 1; SYSCTL_INT(_net_inet_tcp, TCPCTL_DO_RFC1323, rfc1323, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(tcp_do_rfc1323), 0, "Enable rfc1323 (high performance TCP) extensions"); static int tcp_log_debug = 0; SYSCTL_INT(_net_inet_tcp, OID_AUTO, log_debug, CTLFLAG_RW, &tcp_log_debug, 0, "Log errors caused by incoming TCP segments"); static int tcp_tcbhashsize; SYSCTL_INT(_net_inet_tcp, OID_AUTO, tcbhashsize, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &tcp_tcbhashsize, 0, "Size of TCP control-block hashtable"); static int do_tcpdrain = 1; SYSCTL_INT(_net_inet_tcp, OID_AUTO, do_tcpdrain, CTLFLAG_RW, &do_tcpdrain, 0, "Enable tcp_drain routine for extra help when low on mbufs"); SYSCTL_UINT(_net_inet_tcp, OID_AUTO, pcbcount, CTLFLAG_VNET | CTLFLAG_RD, &VNET_NAME(tcbinfo.ipi_count), 0, "Number of active PCBs"); static VNET_DEFINE(int, icmp_may_rst) = 1; #define V_icmp_may_rst VNET(icmp_may_rst) SYSCTL_INT(_net_inet_tcp, OID_AUTO, icmp_may_rst, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(icmp_may_rst), 0, "Certain ICMP unreachable messages may abort connections in SYN_SENT"); static VNET_DEFINE(int, tcp_isn_reseed_interval) = 0; #define V_tcp_isn_reseed_interval VNET(tcp_isn_reseed_interval) SYSCTL_INT(_net_inet_tcp, OID_AUTO, isn_reseed_interval, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(tcp_isn_reseed_interval), 0, "Seconds between reseeding of ISN secret"); static int tcp_soreceive_stream; SYSCTL_INT(_net_inet_tcp, OID_AUTO, soreceive_stream, CTLFLAG_RDTUN, &tcp_soreceive_stream, 0, "Using soreceive_stream for TCP sockets"); #ifdef TCP_SIGNATURE static int tcp_sig_checksigs = 1; SYSCTL_INT(_net_inet_tcp, OID_AUTO, signature_verify_input, CTLFLAG_RW, &tcp_sig_checksigs, 0, "Verify RFC2385 digests on inbound traffic"); #endif VNET_DEFINE(uma_zone_t, sack_hole_zone); #define V_sack_hole_zone VNET(sack_hole_zone) VNET_DEFINE(struct hhook_head *, tcp_hhh[HHOOK_TCP_LAST+1]); static struct inpcb *tcp_notify(struct inpcb *, int); static struct inpcb *tcp_mtudisc_notify(struct inpcb *, int); static char * tcp_log_addr(struct in_conninfo *inc, struct tcphdr *th, void *ip4hdr, const void *ip6hdr); static void tcp_timer_discard(struct tcpcb *, uint32_t); /* * Target size of TCP PCB hash tables. Must be a power of two. * * Note that this can be overridden by the kernel environment * variable net.inet.tcp.tcbhashsize */ #ifndef TCBHASHSIZE #define TCBHASHSIZE 0 #endif /* * XXX * Callouts should be moved into struct tcp directly. They are currently * separate because the tcpcb structure is exported to userland for sysctl * parsing purposes, which do not know about callouts. */ struct tcpcb_mem { struct tcpcb tcb; struct tcp_timer tt; struct cc_var ccv; struct osd osd; }; static VNET_DEFINE(uma_zone_t, tcpcb_zone); #define V_tcpcb_zone VNET(tcpcb_zone) MALLOC_DEFINE(M_TCPLOG, "tcplog", "TCP address and flags print buffers"); static struct mtx isn_mtx; #define ISN_LOCK_INIT() mtx_init(&isn_mtx, "isn_mtx", NULL, MTX_DEF) #define ISN_LOCK() mtx_lock(&isn_mtx) #define ISN_UNLOCK() mtx_unlock(&isn_mtx) /* * TCP initialization. */ static void tcp_zone_change(void *tag) { uma_zone_set_max(V_tcbinfo.ipi_zone, maxsockets); uma_zone_set_max(V_tcpcb_zone, maxsockets); tcp_tw_zone_change(); } static int tcp_inpcb_init(void *mem, int size, int flags) { struct inpcb *inp = mem; INP_LOCK_INIT(inp, "inp", "tcpinp"); return (0); } /* * Take a value and get the next power of 2 that doesn't overflow. * Used to size the tcp_inpcb hash buckets. */ static int maketcp_hashsize(int size) { int hashsize; /* * auto tune. * get the next power of 2 higher than maxsockets. */ hashsize = 1 << fls(size); /* catch overflow, and just go one power of 2 smaller */ if (hashsize < size) { hashsize = 1 << (fls(size) - 1); } return (hashsize); } void tcp_init(void) { const char *tcbhash_tuneable; int hashsize; tcbhash_tuneable = "net.inet.tcp.tcbhashsize"; if (hhook_head_register(HHOOK_TYPE_TCP, HHOOK_TCP_EST_IN, &V_tcp_hhh[HHOOK_TCP_EST_IN], HHOOK_NOWAIT|HHOOK_HEADISINVNET) != 0) printf("%s: WARNING: unable to register helper hook\n", __func__); if (hhook_head_register(HHOOK_TYPE_TCP, HHOOK_TCP_EST_OUT, &V_tcp_hhh[HHOOK_TCP_EST_OUT], HHOOK_NOWAIT|HHOOK_HEADISINVNET) != 0) printf("%s: WARNING: unable to register helper hook\n", __func__); hashsize = TCBHASHSIZE; TUNABLE_INT_FETCH(tcbhash_tuneable, &hashsize); if (hashsize == 0) { /* * Auto tune the hash size based on maxsockets. * A perfect hash would have a 1:1 mapping * (hashsize = maxsockets) however it's been * suggested that O(2) average is better. */ hashsize = maketcp_hashsize(maxsockets / 4); /* * Our historical default is 512, * do not autotune lower than this. */ if (hashsize < 512) hashsize = 512; if (bootverbose) printf("%s: %s auto tuned to %d\n", __func__, tcbhash_tuneable, hashsize); } /* * We require a hashsize to be a power of two. * Previously if it was not a power of two we would just reset it * back to 512, which could be a nasty surprise if you did not notice * the error message. * Instead what we do is clip it to the closest power of two lower * than the specified hash value. */ if (!powerof2(hashsize)) { int oldhashsize = hashsize; hashsize = maketcp_hashsize(hashsize); /* prevent absurdly low value */ if (hashsize < 16) hashsize = 16; printf("%s: WARNING: TCB hash size not a power of 2, " "clipped from %d to %d.\n", __func__, oldhashsize, hashsize); } in_pcbinfo_init(&V_tcbinfo, "tcp", &V_tcb, hashsize, hashsize, "tcp_inpcb", tcp_inpcb_init, NULL, UMA_ZONE_NOFREE, IPI_HASHFIELDS_4TUPLE); /* * These have to be type stable for the benefit of the timers. */ V_tcpcb_zone = uma_zcreate("tcpcb", sizeof(struct tcpcb_mem), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); uma_zone_set_max(V_tcpcb_zone, maxsockets); uma_zone_set_warning(V_tcpcb_zone, "kern.ipc.maxsockets limit reached"); tcp_tw_init(); syncache_init(); tcp_hc_init(); TUNABLE_INT_FETCH("net.inet.tcp.sack.enable", &V_tcp_do_sack); V_sack_hole_zone = uma_zcreate("sackhole", sizeof(struct sackhole), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); /* Skip initialization of globals for non-default instances. */ if (!IS_DEFAULT_VNET(curvnet)) return; tcp_reass_global_init(); /* XXX virtualize those bellow? */ tcp_delacktime = TCPTV_DELACK; tcp_keepinit = TCPTV_KEEP_INIT; tcp_keepidle = TCPTV_KEEP_IDLE; tcp_keepintvl = TCPTV_KEEPINTVL; tcp_maxpersistidle = TCPTV_KEEP_IDLE; tcp_msl = TCPTV_MSL; tcp_rexmit_min = TCPTV_MIN; if (tcp_rexmit_min < 1) tcp_rexmit_min = 1; tcp_rexmit_slop = TCPTV_CPU_VAR; tcp_finwait2_timeout = TCPTV_FINWAIT2_TIMEOUT; tcp_tcbhashsize = hashsize; if (tcp_soreceive_stream) { #ifdef INET tcp_usrreqs.pru_soreceive = soreceive_stream; #endif #ifdef INET6 tcp6_usrreqs.pru_soreceive = soreceive_stream; #endif /* INET6 */ } #ifdef INET6 #define TCP_MINPROTOHDR (sizeof(struct ip6_hdr) + sizeof(struct tcphdr)) #else /* INET6 */ #define TCP_MINPROTOHDR (sizeof(struct tcpiphdr)) #endif /* INET6 */ if (max_protohdr < TCP_MINPROTOHDR) max_protohdr = TCP_MINPROTOHDR; if (max_linkhdr + TCP_MINPROTOHDR > MHLEN) panic("tcp_init"); #undef TCP_MINPROTOHDR ISN_LOCK_INIT(); EVENTHANDLER_REGISTER(shutdown_pre_sync, tcp_fini, NULL, SHUTDOWN_PRI_DEFAULT); EVENTHANDLER_REGISTER(maxsockets_change, tcp_zone_change, NULL, EVENTHANDLER_PRI_ANY); } #ifdef VIMAGE void tcp_destroy(void) { int error; tcp_hc_destroy(); syncache_destroy(); tcp_tw_destroy(); in_pcbinfo_destroy(&V_tcbinfo); uma_zdestroy(V_sack_hole_zone); uma_zdestroy(V_tcpcb_zone); error = hhook_head_deregister(V_tcp_hhh[HHOOK_TCP_EST_IN]); if (error != 0) { printf("%s: WARNING: unable to deregister helper hook " "type=%d, id=%d: error %d returned\n", __func__, HHOOK_TYPE_TCP, HHOOK_TCP_EST_IN, error); } error = hhook_head_deregister(V_tcp_hhh[HHOOK_TCP_EST_OUT]); if (error != 0) { printf("%s: WARNING: unable to deregister helper hook " "type=%d, id=%d: error %d returned\n", __func__, HHOOK_TYPE_TCP, HHOOK_TCP_EST_OUT, error); } } #endif void tcp_fini(void *xtp) { } /* * Fill in the IP and TCP headers for an outgoing packet, given the tcpcb. * tcp_template used to store this data in mbufs, but we now recopy it out * of the tcpcb each time to conserve mbufs. */ void tcpip_fillheaders(struct inpcb *inp, void *ip_ptr, void *tcp_ptr) { struct tcphdr *th = (struct tcphdr *)tcp_ptr; INP_WLOCK_ASSERT(inp); #ifdef INET6 if ((inp->inp_vflag & INP_IPV6) != 0) { struct ip6_hdr *ip6; ip6 = (struct ip6_hdr *)ip_ptr; ip6->ip6_flow = (ip6->ip6_flow & ~IPV6_FLOWINFO_MASK) | (inp->inp_flow & IPV6_FLOWINFO_MASK); ip6->ip6_vfc = (ip6->ip6_vfc & ~IPV6_VERSION_MASK) | (IPV6_VERSION & IPV6_VERSION_MASK); ip6->ip6_nxt = IPPROTO_TCP; ip6->ip6_plen = htons(sizeof(struct tcphdr)); ip6->ip6_src = inp->in6p_laddr; ip6->ip6_dst = inp->in6p_faddr; } #endif /* INET6 */ #if defined(INET6) && defined(INET) else #endif #ifdef INET { struct ip *ip; ip = (struct ip *)ip_ptr; ip->ip_v = IPVERSION; ip->ip_hl = 5; ip->ip_tos = inp->inp_ip_tos; ip->ip_len = 0; ip->ip_id = 0; ip->ip_off = 0; ip->ip_ttl = inp->inp_ip_ttl; ip->ip_sum = 0; ip->ip_p = IPPROTO_TCP; ip->ip_src = inp->inp_laddr; ip->ip_dst = inp->inp_faddr; } #endif /* INET */ th->th_sport = inp->inp_lport; th->th_dport = inp->inp_fport; th->th_seq = 0; th->th_ack = 0; th->th_x2 = 0; th->th_off = 5; th->th_flags = 0; th->th_win = 0; th->th_urp = 0; th->th_sum = 0; /* in_pseudo() is called later for ipv4 */ } /* * Create template to be used to send tcp packets on a connection. * Allocates an mbuf and fills in a skeletal tcp/ip header. The only * use for this function is in keepalives, which use tcp_respond. */ struct tcptemp * tcpip_maketemplate(struct inpcb *inp) { struct tcptemp *t; t = malloc(sizeof(*t), M_TEMP, M_NOWAIT); if (t == NULL) return (NULL); tcpip_fillheaders(inp, (void *)&t->tt_ipgen, (void *)&t->tt_t); return (t); } /* * Send a single message to the TCP at address specified by * the given TCP/IP header. If m == NULL, then we make a copy * of the tcpiphdr at th and send directly to the addressed host. * This is used to force keep alive messages out using the TCP * template for a connection. If flags are given then we send * a message back to the TCP which originated the segment th, * and discard the mbuf containing it and any other attached mbufs. * * In any case the ack and sequence number of the transmitted * segment are as specified by the parameters. * * NOTE: If m != NULL, then th must point to *inside* the mbuf. */ void tcp_respond(struct tcpcb *tp, void *ipgen, struct tcphdr *th, struct mbuf *m, tcp_seq ack, tcp_seq seq, int flags) { int tlen; int win = 0; struct ip *ip; struct tcphdr *nth; #ifdef INET6 struct ip6_hdr *ip6; int isipv6; #endif /* INET6 */ int ipflags = 0; struct inpcb *inp; KASSERT(tp != NULL || m != NULL, ("tcp_respond: tp and m both NULL")); #ifdef INET6 isipv6 = ((struct ip *)ipgen)->ip_v == (IPV6_VERSION >> 4); ip6 = ipgen; #endif /* INET6 */ ip = ipgen; if (tp != NULL) { inp = tp->t_inpcb; KASSERT(inp != NULL, ("tcp control block w/o inpcb")); INP_WLOCK_ASSERT(inp); } else inp = NULL; if (tp != NULL) { if (!(flags & TH_RST)) { win = sbspace(&inp->inp_socket->so_rcv); if (win > (long)TCP_MAXWIN << tp->rcv_scale) win = (long)TCP_MAXWIN << tp->rcv_scale; } } if (m == NULL) { m = m_gethdr(M_NOWAIT, MT_DATA); if (m == NULL) return; tlen = 0; m->m_data += max_linkhdr; #ifdef INET6 if (isipv6) { bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(struct ip6_hdr)); ip6 = mtod(m, struct ip6_hdr *); nth = (struct tcphdr *)(ip6 + 1); } else #endif /* INET6 */ { bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip)); ip = mtod(m, struct ip *); nth = (struct tcphdr *)(ip + 1); } bcopy((caddr_t)th, (caddr_t)nth, sizeof(struct tcphdr)); flags = TH_ACK; } else { /* * reuse the mbuf. * XXX MRT We inherrit the FIB, which is lucky. */ m_freem(m->m_next); m->m_next = NULL; m->m_data = (caddr_t)ipgen; /* m_len is set later */ tlen = 0; #define xchg(a,b,type) { type t; t=a; a=b; b=t; } #ifdef INET6 if (isipv6) { xchg(ip6->ip6_dst, ip6->ip6_src, struct in6_addr); nth = (struct tcphdr *)(ip6 + 1); } else #endif /* INET6 */ { xchg(ip->ip_dst.s_addr, ip->ip_src.s_addr, uint32_t); nth = (struct tcphdr *)(ip + 1); } if (th != nth) { /* * this is usually a case when an extension header * exists between the IPv6 header and the * TCP header. */ nth->th_sport = th->th_sport; nth->th_dport = th->th_dport; } xchg(nth->th_dport, nth->th_sport, uint16_t); #undef xchg } #ifdef INET6 if (isipv6) { ip6->ip6_flow = 0; ip6->ip6_vfc = IPV6_VERSION; ip6->ip6_nxt = IPPROTO_TCP; tlen += sizeof (struct ip6_hdr) + sizeof (struct tcphdr); ip6->ip6_plen = htons(tlen - sizeof(*ip6)); } #endif #if defined(INET) && defined(INET6) else #endif #ifdef INET { tlen += sizeof (struct tcpiphdr); ip->ip_len = htons(tlen); ip->ip_ttl = V_ip_defttl; if (V_path_mtu_discovery) ip->ip_off |= htons(IP_DF); } #endif m->m_len = tlen; m->m_pkthdr.len = tlen; m->m_pkthdr.rcvif = NULL; #ifdef MAC if (inp != NULL) { /* * Packet is associated with a socket, so allow the * label of the response to reflect the socket label. */ INP_WLOCK_ASSERT(inp); mac_inpcb_create_mbuf(inp, m); } else { /* * Packet is not associated with a socket, so possibly * update the label in place. */ mac_netinet_tcp_reply(m); } #endif nth->th_seq = htonl(seq); nth->th_ack = htonl(ack); nth->th_x2 = 0; nth->th_off = sizeof (struct tcphdr) >> 2; nth->th_flags = flags; if (tp != NULL) nth->th_win = htons((u_short) (win >> tp->rcv_scale)); else nth->th_win = htons((u_short)win); nth->th_urp = 0; m->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum); #ifdef INET6 if (isipv6) { m->m_pkthdr.csum_flags = CSUM_TCP_IPV6; nth->th_sum = in6_cksum_pseudo(ip6, tlen - sizeof(struct ip6_hdr), IPPROTO_TCP, 0); ip6->ip6_hlim = in6_selecthlim(tp != NULL ? tp->t_inpcb : NULL, NULL); } #endif /* INET6 */ #if defined(INET6) && defined(INET) else #endif #ifdef INET { m->m_pkthdr.csum_flags = CSUM_TCP; 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))); } #endif /* INET */ #ifdef TCPDEBUG if (tp == NULL || (inp->inp_socket->so_options & SO_DEBUG)) tcp_trace(TA_OUTPUT, 0, tp, mtod(m, void *), th, 0); #endif if (flags & TH_RST) TCP_PROBE5(accept__refused, NULL, NULL, mtod(m, const char *), tp, nth); TCP_PROBE5(send, NULL, tp, mtod(m, const char *), tp, nth); #ifdef INET6 if (isipv6) (void) ip6_output(m, NULL, NULL, ipflags, NULL, NULL, inp); #endif /* INET6 */ #if defined(INET) && defined(INET6) else #endif #ifdef INET (void) ip_output(m, NULL, NULL, ipflags, NULL, inp); #endif } /* * Create a new TCP control block, making an * empty reassembly queue and hooking it to the argument * protocol control block. The `inp' parameter must have * come from the zone allocator set up in tcp_init(). */ struct tcpcb * tcp_newtcpcb(struct inpcb *inp) { struct tcpcb_mem *tm; struct tcpcb *tp; #ifdef INET6 int isipv6 = (inp->inp_vflag & INP_IPV6) != 0; #endif /* INET6 */ tm = uma_zalloc(V_tcpcb_zone, M_NOWAIT | M_ZERO); if (tm == NULL) return (NULL); tp = &tm->tcb; /* Initialise cc_var struct for this tcpcb. */ tp->ccv = &tm->ccv; tp->ccv->type = IPPROTO_TCP; tp->ccv->ccvc.tcp = tp; /* * Use the current system default CC algorithm. */ CC_LIST_RLOCK(); KASSERT(!STAILQ_EMPTY(&cc_list), ("cc_list is empty!")); CC_ALGO(tp) = CC_DEFAULT(); CC_LIST_RUNLOCK(); if (CC_ALGO(tp)->cb_init != NULL) if (CC_ALGO(tp)->cb_init(tp->ccv) > 0) { uma_zfree(V_tcpcb_zone, tm); return (NULL); } tp->osd = &tm->osd; if (khelp_init_osd(HELPER_CLASS_TCP, tp->osd)) { uma_zfree(V_tcpcb_zone, tm); return (NULL); } #ifdef VIMAGE tp->t_vnet = inp->inp_vnet; #endif tp->t_timers = &tm->tt; /* LIST_INIT(&tp->t_segq); */ /* XXX covered by M_ZERO */ tp->t_maxseg = tp->t_maxopd = #ifdef INET6 isipv6 ? V_tcp_v6mssdflt : #endif /* INET6 */ V_tcp_mssdflt; /* Set up our timeouts. */ callout_init(&tp->t_timers->tt_rexmt, 1); callout_init(&tp->t_timers->tt_persist, 1); callout_init(&tp->t_timers->tt_keep, 1); callout_init(&tp->t_timers->tt_2msl, 1); callout_init(&tp->t_timers->tt_delack, 1); if (V_tcp_do_rfc1323) tp->t_flags = (TF_REQ_SCALE|TF_REQ_TSTMP); if (V_tcp_do_sack) tp->t_flags |= TF_SACK_PERMIT; TAILQ_INIT(&tp->snd_holes); /* * The tcpcb will hold a reference on its inpcb until tcp_discardcb() * is called. */ in_pcbref(inp); /* Reference for tcpcb */ tp->t_inpcb = inp; /* * Init srtt to TCPTV_SRTTBASE (0), so we can tell that we have no * rtt estimate. Set rttvar so that srtt + 4 * rttvar gives * reasonable initial retransmit time. */ tp->t_srtt = TCPTV_SRTTBASE; tp->t_rttvar = ((TCPTV_RTOBASE - TCPTV_SRTTBASE) << TCP_RTTVAR_SHIFT) / 4; tp->t_rttmin = tcp_rexmit_min; tp->t_rxtcur = TCPTV_RTOBASE; tp->snd_cwnd = TCP_MAXWIN << TCP_MAX_WINSHIFT; tp->snd_ssthresh = TCP_MAXWIN << TCP_MAX_WINSHIFT; tp->t_rcvtime = ticks; /* * IPv4 TTL initialization is necessary for an IPv6 socket as well, * because the socket may be bound to an IPv6 wildcard address, * which may match an IPv4-mapped IPv6 address. */ inp->inp_ip_ttl = V_ip_defttl; inp->inp_ppcb = tp; return (tp); /* XXX */ } /* * Switch the congestion control algorithm back to NewReno for any active * control blocks using an algorithm which is about to go away. * This ensures the CC framework can allow the unload to proceed without leaving * any dangling pointers which would trigger a panic. * Returning non-zero would inform the CC framework that something went wrong * and it would be unsafe to allow the unload to proceed. However, there is no * way for this to occur with this implementation so we always return zero. */ int tcp_ccalgounload(struct cc_algo *unload_algo) { struct cc_algo *tmpalgo; struct inpcb *inp; struct tcpcb *tp; VNET_ITERATOR_DECL(vnet_iter); /* * Check all active control blocks across all network stacks and change * any that are using "unload_algo" back to NewReno. If "unload_algo" * requires cleanup code to be run, call it. */ VNET_LIST_RLOCK(); VNET_FOREACH(vnet_iter) { CURVNET_SET(vnet_iter); INP_INFO_RLOCK(&V_tcbinfo); /* * New connections already part way through being initialised * with the CC algo we're removing will not race with this code * because the INP_INFO_WLOCK is held during initialisation. We * therefore don't enter the loop below until the connection * list has stabilised. */ LIST_FOREACH(inp, &V_tcb, inp_list) { INP_WLOCK(inp); /* Important to skip tcptw structs. */ if (!(inp->inp_flags & INP_TIMEWAIT) && (tp = intotcpcb(inp)) != NULL) { /* * By holding INP_WLOCK here, we are assured * that the connection is not currently * executing inside the CC module's functions * i.e. it is safe to make the switch back to * NewReno. */ if (CC_ALGO(tp) == unload_algo) { tmpalgo = CC_ALGO(tp); /* NewReno does not require any init. */ CC_ALGO(tp) = &newreno_cc_algo; if (tmpalgo->cb_destroy != NULL) tmpalgo->cb_destroy(tp->ccv); } } INP_WUNLOCK(inp); } INP_INFO_RUNLOCK(&V_tcbinfo); CURVNET_RESTORE(); } VNET_LIST_RUNLOCK(); return (0); } /* * Drop a TCP connection, reporting * the specified error. If connection is synchronized, * then send a RST to peer. */ struct tcpcb * tcp_drop(struct tcpcb *tp, int errno) { struct socket *so = tp->t_inpcb->inp_socket; INP_INFO_WLOCK_ASSERT(&V_tcbinfo); INP_WLOCK_ASSERT(tp->t_inpcb); if (TCPS_HAVERCVDSYN(tp->t_state)) { tcp_state_change(tp, TCPS_CLOSED); (void) tcp_output(tp); TCPSTAT_INC(tcps_drops); } else TCPSTAT_INC(tcps_conndrops); if (errno == ETIMEDOUT && tp->t_softerror) errno = tp->t_softerror; so->so_error = errno; return (tcp_close(tp)); } void tcp_discardcb(struct tcpcb *tp) { struct inpcb *inp = tp->t_inpcb; struct socket *so = inp->inp_socket; #ifdef INET6 int isipv6 = (inp->inp_vflag & INP_IPV6) != 0; #endif /* INET6 */ int released; INP_WLOCK_ASSERT(inp); /* * Make sure that all of our timers are stopped before we delete the * PCB. * * If stopping a timer fails, we schedule a discard function in same * callout, and the last discard function called will take care of * deleting the tcpcb. */ tcp_timer_stop(tp, TT_REXMT); tcp_timer_stop(tp, TT_PERSIST); tcp_timer_stop(tp, TT_KEEP); tcp_timer_stop(tp, TT_2MSL); tcp_timer_stop(tp, TT_DELACK); /* * If we got enough samples through the srtt filter, * save the rtt and rttvar in the routing entry. * 'Enough' is arbitrarily defined as 4 rtt samples. * 4 samples is enough for the srtt filter to converge * to within enough % of the correct value; fewer samples * and we could save a bogus rtt. The danger is not high * as tcp quickly recovers from everything. * XXX: Works very well but needs some more statistics! */ if (tp->t_rttupdated >= 4) { struct hc_metrics_lite metrics; u_long ssthresh; bzero(&metrics, sizeof(metrics)); /* * Update the ssthresh always when the conditions below * are satisfied. This gives us better new start value * for the congestion avoidance for new connections. * ssthresh is only set if packet loss occured on a session. * * XXXRW: 'so' may be NULL here, and/or socket buffer may be * being torn down. Ideally this code would not use 'so'. */ ssthresh = tp->snd_ssthresh; if (ssthresh != 0 && ssthresh < so->so_snd.sb_hiwat / 2) { /* * convert the limit from user data bytes to * packets then to packet data bytes. */ ssthresh = (ssthresh + tp->t_maxseg / 2) / tp->t_maxseg; if (ssthresh < 2) ssthresh = 2; ssthresh *= (u_long)(tp->t_maxseg + #ifdef INET6 (isipv6 ? sizeof (struct ip6_hdr) + sizeof (struct tcphdr) : #endif sizeof (struct tcpiphdr) #ifdef INET6 ) #endif ); } else ssthresh = 0; metrics.rmx_ssthresh = ssthresh; metrics.rmx_rtt = tp->t_srtt; metrics.rmx_rttvar = tp->t_rttvar; metrics.rmx_cwnd = tp->snd_cwnd; metrics.rmx_sendpipe = 0; metrics.rmx_recvpipe = 0; tcp_hc_update(&inp->inp_inc, &metrics); } /* free the reassembly queue, if any */ tcp_reass_flush(tp); #ifdef TCP_OFFLOAD /* Disconnect offload device, if any. */ if (tp->t_flags & TF_TOE) tcp_offload_detach(tp); #endif tcp_free_sackholes(tp); /* Allow the CC algorithm to clean up after itself. */ if (CC_ALGO(tp)->cb_destroy != NULL) CC_ALGO(tp)->cb_destroy(tp->ccv); khelp_destroy_osd(tp->osd); CC_ALGO(tp) = NULL; inp->inp_ppcb = NULL; if ((tp->t_timers->tt_flags & TT_MASK) == 0) { /* We own the last reference on tcpcb, let's free it. */ tp->t_inpcb = NULL; uma_zfree(V_tcpcb_zone, tp); released = in_pcbrele_wlocked(inp); KASSERT(!released, ("%s: inp %p should not have been released " "here", __func__, inp)); } } void tcp_timer_2msl_discard(void *xtp) { tcp_timer_discard((struct tcpcb *)xtp, TT_2MSL); } void tcp_timer_keep_discard(void *xtp) { tcp_timer_discard((struct tcpcb *)xtp, TT_KEEP); } void tcp_timer_persist_discard(void *xtp) { tcp_timer_discard((struct tcpcb *)xtp, TT_PERSIST); } void tcp_timer_rexmt_discard(void *xtp) { tcp_timer_discard((struct tcpcb *)xtp, TT_REXMT); } void tcp_timer_delack_discard(void *xtp) { tcp_timer_discard((struct tcpcb *)xtp, TT_DELACK); } void tcp_timer_discard(struct tcpcb *tp, uint32_t timer_type) { struct inpcb *inp; CURVNET_SET(tp->t_vnet); INP_INFO_WLOCK(&V_tcbinfo); inp = tp->t_inpcb; KASSERT(inp != NULL, ("%s: tp %p tp->t_inpcb == NULL", __func__, tp)); INP_WLOCK(inp); KASSERT((tp->t_timers->tt_flags & TT_STOPPED) != 0, ("%s: tcpcb has to be stopped here", __func__)); KASSERT((tp->t_timers->tt_flags & timer_type) != 0, ("%s: discard callout should be running", __func__)); tp->t_timers->tt_flags &= ~timer_type; if ((tp->t_timers->tt_flags & TT_MASK) == 0) { /* We own the last reference on this tcpcb, let's free it. */ tp->t_inpcb = NULL; uma_zfree(V_tcpcb_zone, tp); if (in_pcbrele_wlocked(inp)) { INP_INFO_WUNLOCK(&V_tcbinfo); CURVNET_RESTORE(); return; } } INP_WUNLOCK(inp); INP_INFO_WUNLOCK(&V_tcbinfo); CURVNET_RESTORE(); } /* * Attempt to close a TCP control block, marking it as dropped, and freeing * the socket if we hold the only reference. */ struct tcpcb * tcp_close(struct tcpcb *tp) { struct inpcb *inp = tp->t_inpcb; struct socket *so; INP_INFO_WLOCK_ASSERT(&V_tcbinfo); INP_WLOCK_ASSERT(inp); #ifdef TCP_OFFLOAD if (tp->t_state == TCPS_LISTEN) tcp_offload_listen_stop(tp); #endif in_pcbdrop(inp); TCPSTAT_INC(tcps_closed); KASSERT(inp->inp_socket != NULL, ("tcp_close: inp_socket NULL")); so = inp->inp_socket; soisdisconnected(so); if (inp->inp_flags & INP_SOCKREF) { KASSERT(so->so_state & SS_PROTOREF, ("tcp_close: !SS_PROTOREF")); inp->inp_flags &= ~INP_SOCKREF; INP_WUNLOCK(inp); ACCEPT_LOCK(); SOCK_LOCK(so); so->so_state &= ~SS_PROTOREF; sofree(so); return (NULL); } return (tp); } void tcp_drain(void) { VNET_ITERATOR_DECL(vnet_iter); if (!do_tcpdrain) return; VNET_LIST_RLOCK_NOSLEEP(); VNET_FOREACH(vnet_iter) { CURVNET_SET(vnet_iter); struct inpcb *inpb; struct tcpcb *tcpb; /* * Walk the tcpbs, if existing, and flush the reassembly queue, * if there is one... * XXX: The "Net/3" implementation doesn't imply that the TCP * reassembly queue should be flushed, but in a situation * where we're really low on mbufs, this is potentially * useful. */ INP_INFO_RLOCK(&V_tcbinfo); LIST_FOREACH(inpb, V_tcbinfo.ipi_listhead, inp_list) { if (inpb->inp_flags & INP_TIMEWAIT) continue; INP_WLOCK(inpb); if ((tcpb = intotcpcb(inpb)) != NULL) { tcp_reass_flush(tcpb); tcp_clean_sackreport(tcpb); } INP_WUNLOCK(inpb); } INP_INFO_RUNLOCK(&V_tcbinfo); CURVNET_RESTORE(); } VNET_LIST_RUNLOCK_NOSLEEP(); } /* * Notify a tcp user of an asynchronous error; * store error as soft error, but wake up user * (for now, won't do anything until can select for soft error). * * Do not wake up user since there currently is no mechanism for * reporting soft errors (yet - a kqueue filter may be added). */ static struct inpcb * tcp_notify(struct inpcb *inp, int error) { struct tcpcb *tp; INP_INFO_WLOCK_ASSERT(&V_tcbinfo); INP_WLOCK_ASSERT(inp); if ((inp->inp_flags & INP_TIMEWAIT) || (inp->inp_flags & INP_DROPPED)) return (inp); tp = intotcpcb(inp); KASSERT(tp != NULL, ("tcp_notify: tp == NULL")); /* * Ignore some errors if we are hooked up. * If connection hasn't completed, has retransmitted several times, * and receives a second error, give up now. This is better * than waiting a long time to establish a connection that * can never complete. */ if (tp->t_state == TCPS_ESTABLISHED && (error == EHOSTUNREACH || error == ENETUNREACH || error == EHOSTDOWN)) { return (inp); } else if (tp->t_state < TCPS_ESTABLISHED && tp->t_rxtshift > 3 && tp->t_softerror) { tp = tcp_drop(tp, error); if (tp != NULL) return (inp); else return (NULL); } else { tp->t_softerror = error; return (inp); } #if 0 wakeup( &so->so_timeo); sorwakeup(so); sowwakeup(so); #endif } static int tcp_pcblist(SYSCTL_HANDLER_ARGS) { int error, i, m, n, pcb_count; struct inpcb *inp, **inp_list; inp_gen_t gencnt; struct xinpgen xig; /* * The process of preparing the TCB list is too time-consuming and * resource-intensive to repeat twice on every request. */ if (req->oldptr == NULL) { n = V_tcbinfo.ipi_count + syncache_pcbcount(); n += imax(n / 8, 10); req->oldidx = 2 * (sizeof xig) + n * sizeof(struct xtcpcb); return (0); } if (req->newptr != NULL) return (EPERM); /* * OK, now we're committed to doing something. */ INP_INFO_RLOCK(&V_tcbinfo); gencnt = V_tcbinfo.ipi_gencnt; n = V_tcbinfo.ipi_count; INP_INFO_RUNLOCK(&V_tcbinfo); m = syncache_pcbcount(); error = sysctl_wire_old_buffer(req, 2 * (sizeof xig) + (n + m) * sizeof(struct xtcpcb)); if (error != 0) return (error); xig.xig_len = sizeof xig; xig.xig_count = n + m; xig.xig_gen = gencnt; xig.xig_sogen = so_gencnt; error = SYSCTL_OUT(req, &xig, sizeof xig); if (error) return (error); error = syncache_pcblist(req, m, &pcb_count); if (error) return (error); inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK); if (inp_list == NULL) return (ENOMEM); INP_INFO_RLOCK(&V_tcbinfo); for (inp = LIST_FIRST(V_tcbinfo.ipi_listhead), i = 0; inp != NULL && i < n; inp = LIST_NEXT(inp, inp_list)) { INP_WLOCK(inp); if (inp->inp_gencnt <= gencnt) { /* * XXX: This use of cr_cansee(), introduced with * TCP state changes, is not quite right, but for * now, better than nothing. */ if (inp->inp_flags & INP_TIMEWAIT) { if (intotw(inp) != NULL) error = cr_cansee(req->td->td_ucred, intotw(inp)->tw_cred); else error = EINVAL; /* Skip this inp. */ } else error = cr_canseeinpcb(req->td->td_ucred, inp); if (error == 0) { in_pcbref(inp); inp_list[i++] = inp; } } INP_WUNLOCK(inp); } INP_INFO_RUNLOCK(&V_tcbinfo); n = i; error = 0; for (i = 0; i < n; i++) { inp = inp_list[i]; INP_RLOCK(inp); if (inp->inp_gencnt <= gencnt) { struct xtcpcb xt; void *inp_ppcb; bzero(&xt, sizeof(xt)); xt.xt_len = sizeof xt; /* XXX should avoid extra copy */ bcopy(inp, &xt.xt_inp, sizeof *inp); inp_ppcb = inp->inp_ppcb; if (inp_ppcb == NULL) bzero((char *) &xt.xt_tp, sizeof xt.xt_tp); else if (inp->inp_flags & INP_TIMEWAIT) { bzero((char *) &xt.xt_tp, sizeof xt.xt_tp); xt.xt_tp.t_state = TCPS_TIME_WAIT; } else { bcopy(inp_ppcb, &xt.xt_tp, sizeof xt.xt_tp); if (xt.xt_tp.t_timers) tcp_timer_to_xtimer(&xt.xt_tp, xt.xt_tp.t_timers, &xt.xt_timer); } if (inp->inp_socket != NULL) sotoxsocket(inp->inp_socket, &xt.xt_socket); else { bzero(&xt.xt_socket, sizeof xt.xt_socket); xt.xt_socket.xso_protocol = IPPROTO_TCP; } xt.xt_inp.inp_gencnt = inp->inp_gencnt; INP_RUNLOCK(inp); error = SYSCTL_OUT(req, &xt, sizeof xt); } else INP_RUNLOCK(inp); } INP_INFO_WLOCK(&V_tcbinfo); for (i = 0; i < n; i++) { inp = inp_list[i]; INP_RLOCK(inp); if (!in_pcbrele_rlocked(inp)) INP_RUNLOCK(inp); } INP_INFO_WUNLOCK(&V_tcbinfo); if (!error) { /* * Give the user an updated idea of our state. * If the generation differs from what we told * her before, she knows that something happened * while we were processing this request, and it * might be necessary to retry. */ INP_INFO_RLOCK(&V_tcbinfo); xig.xig_gen = V_tcbinfo.ipi_gencnt; xig.xig_sogen = so_gencnt; xig.xig_count = V_tcbinfo.ipi_count + pcb_count; INP_INFO_RUNLOCK(&V_tcbinfo); error = SYSCTL_OUT(req, &xig, sizeof xig); } free(inp_list, M_TEMP); return (error); } SYSCTL_PROC(_net_inet_tcp, TCPCTL_PCBLIST, pcblist, CTLTYPE_OPAQUE | CTLFLAG_RD, NULL, 0, tcp_pcblist, "S,xtcpcb", "List of active TCP connections"); #ifdef INET static int tcp_getcred(SYSCTL_HANDLER_ARGS) { struct xucred xuc; struct sockaddr_in addrs[2]; struct inpcb *inp; int error; error = priv_check(req->td, PRIV_NETINET_GETCRED); if (error) return (error); error = SYSCTL_IN(req, addrs, sizeof(addrs)); if (error) return (error); inp = in_pcblookup(&V_tcbinfo, addrs[1].sin_addr, addrs[1].sin_port, addrs[0].sin_addr, addrs[0].sin_port, INPLOOKUP_RLOCKPCB, NULL); if (inp != NULL) { if (inp->inp_socket == NULL) error = ENOENT; if (error == 0) error = cr_canseeinpcb(req->td->td_ucred, inp); if (error == 0) cru2x(inp->inp_cred, &xuc); INP_RUNLOCK(inp); } else error = ENOENT; if (error == 0) error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred)); return (error); } SYSCTL_PROC(_net_inet_tcp, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW|CTLFLAG_PRISON, 0, 0, tcp_getcred, "S,xucred", "Get the xucred of a TCP connection"); #endif /* INET */ #ifdef INET6 static int tcp6_getcred(SYSCTL_HANDLER_ARGS) { struct xucred xuc; struct sockaddr_in6 addrs[2]; struct inpcb *inp; int error; #ifdef INET int mapped = 0; #endif error = priv_check(req->td, PRIV_NETINET_GETCRED); if (error) return (error); error = SYSCTL_IN(req, addrs, sizeof(addrs)); if (error) return (error); if ((error = sa6_embedscope(&addrs[0], V_ip6_use_defzone)) != 0 || (error = sa6_embedscope(&addrs[1], V_ip6_use_defzone)) != 0) { return (error); } if (IN6_IS_ADDR_V4MAPPED(&addrs[0].sin6_addr)) { #ifdef INET if (IN6_IS_ADDR_V4MAPPED(&addrs[1].sin6_addr)) mapped = 1; else #endif return (EINVAL); } #ifdef INET if (mapped == 1) inp = in_pcblookup(&V_tcbinfo, *(struct in_addr *)&addrs[1].sin6_addr.s6_addr[12], addrs[1].sin6_port, *(struct in_addr *)&addrs[0].sin6_addr.s6_addr[12], addrs[0].sin6_port, INPLOOKUP_RLOCKPCB, NULL); else #endif inp = in6_pcblookup(&V_tcbinfo, &addrs[1].sin6_addr, addrs[1].sin6_port, &addrs[0].sin6_addr, addrs[0].sin6_port, INPLOOKUP_RLOCKPCB, NULL); if (inp != NULL) { if (inp->inp_socket == NULL) error = ENOENT; if (error == 0) error = cr_canseeinpcb(req->td->td_ucred, inp); if (error == 0) cru2x(inp->inp_cred, &xuc); INP_RUNLOCK(inp); } else error = ENOENT; if (error == 0) error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred)); return (error); } SYSCTL_PROC(_net_inet6_tcp6, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW|CTLFLAG_PRISON, 0, 0, tcp6_getcred, "S,xucred", "Get the xucred of a TCP6 connection"); #endif /* INET6 */ #ifdef INET void tcp_ctlinput(int cmd, struct sockaddr *sa, void *vip) { struct ip *ip = vip; struct tcphdr *th; struct in_addr faddr; struct inpcb *inp; struct tcpcb *tp; struct inpcb *(*notify)(struct inpcb *, int) = tcp_notify; struct icmp *icp; struct in_conninfo inc; tcp_seq icmp_tcp_seq; int mtu; faddr = ((struct sockaddr_in *)sa)->sin_addr; if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY) return; if (cmd == PRC_MSGSIZE) notify = tcp_mtudisc_notify; else if (V_icmp_may_rst && (cmd == PRC_UNREACH_ADMIN_PROHIB || cmd == PRC_UNREACH_PORT || cmd == PRC_TIMXCEED_INTRANS) && ip) notify = tcp_drop_syn_sent; /* * Redirects don't need to be handled up here. */ else if (PRC_IS_REDIRECT(cmd)) return; /* * Hostdead is ugly because it goes linearly through all PCBs. * XXX: We never get this from ICMP, otherwise it makes an * excellent DoS attack on machines with many connections. */ else if (cmd == PRC_HOSTDEAD) ip = NULL; else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0) return; if (ip != NULL) { icp = (struct icmp *)((caddr_t)ip - offsetof(struct icmp, icmp_ip)); th = (struct tcphdr *)((caddr_t)ip + (ip->ip_hl << 2)); INP_INFO_WLOCK(&V_tcbinfo); inp = in_pcblookup(&V_tcbinfo, faddr, th->th_dport, ip->ip_src, th->th_sport, INPLOOKUP_WLOCKPCB, NULL); if (inp != NULL) { if (!(inp->inp_flags & INP_TIMEWAIT) && !(inp->inp_flags & INP_DROPPED) && !(inp->inp_socket == NULL)) { icmp_tcp_seq = htonl(th->th_seq); tp = intotcpcb(inp); if (SEQ_GEQ(icmp_tcp_seq, tp->snd_una) && SEQ_LT(icmp_tcp_seq, tp->snd_max)) { if (cmd == PRC_MSGSIZE) { /* * MTU discovery: * If we got a needfrag set the MTU * in the route to the suggested new * value (if given) and then notify. */ bzero(&inc, sizeof(inc)); inc.inc_faddr = faddr; inc.inc_fibnum = inp->inp_inc.inc_fibnum; mtu = ntohs(icp->icmp_nextmtu); /* * If no alternative MTU was * proposed, try the next smaller * one. */ if (!mtu) mtu = ip_next_mtu( ntohs(ip->ip_len), 1); if (mtu < V_tcp_minmss + sizeof(struct tcpiphdr)) mtu = V_tcp_minmss + sizeof(struct tcpiphdr); /* * Only cache the MTU if it * is smaller than the interface * or route MTU. tcp_mtudisc() * will do right thing by itself. */ if (mtu <= tcp_maxmtu(&inc, NULL)) tcp_hc_updatemtu(&inc, mtu); tcp_mtudisc(inp, mtu); } else inp = (*notify)(inp, inetctlerrmap[cmd]); } } if (inp != NULL) INP_WUNLOCK(inp); } else { bzero(&inc, sizeof(inc)); inc.inc_fport = th->th_dport; inc.inc_lport = th->th_sport; inc.inc_faddr = faddr; inc.inc_laddr = ip->ip_src; syncache_unreach(&inc, th); } INP_INFO_WUNLOCK(&V_tcbinfo); } else in_pcbnotifyall(&V_tcbinfo, faddr, inetctlerrmap[cmd], notify); } #endif /* INET */ #ifdef INET6 void tcp6_ctlinput(int cmd, struct sockaddr *sa, void *d) { struct tcphdr th; struct inpcb *(*notify)(struct inpcb *, int) = tcp_notify; struct ip6_hdr *ip6; struct mbuf *m; struct ip6ctlparam *ip6cp = NULL; const struct sockaddr_in6 *sa6_src = NULL; int off; struct tcp_portonly { u_int16_t th_sport; u_int16_t th_dport; } *thp; if (sa->sa_family != AF_INET6 || sa->sa_len != sizeof(struct sockaddr_in6)) return; if (cmd == PRC_MSGSIZE) notify = tcp_mtudisc_notify; else if (!PRC_IS_REDIRECT(cmd) && ((unsigned)cmd >= PRC_NCMDS || inet6ctlerrmap[cmd] == 0)) return; /* if the parameter is from icmp6, decode it. */ if (d != NULL) { ip6cp = (struct ip6ctlparam *)d; m = ip6cp->ip6c_m; ip6 = ip6cp->ip6c_ip6; off = ip6cp->ip6c_off; sa6_src = ip6cp->ip6c_src; } else { m = NULL; ip6 = NULL; off = 0; /* fool gcc */ sa6_src = &sa6_any; } if (ip6 != NULL) { struct in_conninfo inc; /* * XXX: We assume that when IPV6 is non NULL, * M and OFF are valid. */ /* check if we can safely examine src and dst ports */ if (m->m_pkthdr.len < off + sizeof(*thp)) return; bzero(&th, sizeof(th)); m_copydata(m, off, sizeof(*thp), (caddr_t)&th); in6_pcbnotify(&V_tcbinfo, sa, th.th_dport, (struct sockaddr *)ip6cp->ip6c_src, th.th_sport, cmd, NULL, notify); bzero(&inc, sizeof(inc)); inc.inc_fport = th.th_dport; inc.inc_lport = th.th_sport; inc.inc6_faddr = ((struct sockaddr_in6 *)sa)->sin6_addr; inc.inc6_laddr = ip6cp->ip6c_src->sin6_addr; inc.inc_flags |= INC_ISIPV6; INP_INFO_WLOCK(&V_tcbinfo); syncache_unreach(&inc, &th); INP_INFO_WUNLOCK(&V_tcbinfo); } else in6_pcbnotify(&V_tcbinfo, sa, 0, (const struct sockaddr *)sa6_src, 0, cmd, NULL, notify); } #endif /* INET6 */ /* * Following is where TCP initial sequence number generation occurs. * * There are two places where we must use initial sequence numbers: * 1. In SYN-ACK packets. * 2. In SYN packets. * * All ISNs for SYN-ACK packets are generated by the syncache. See * tcp_syncache.c for details. * * The ISNs in SYN packets must be monotonic; TIME_WAIT recycling * depends on this property. In addition, these ISNs should be * unguessable so as to prevent connection hijacking. To satisfy * the requirements of this situation, the algorithm outlined in * RFC 1948 is used, with only small modifications. * * Implementation details: * * Time is based off the system timer, and is corrected so that it * increases by one megabyte per second. This allows for proper * recycling on high speed LANs while still leaving over an hour * before rollover. * * As reading the *exact* system time is too expensive to be done * whenever setting up a TCP connection, we increment the time * offset in two ways. First, a small random positive increment * is added to isn_offset for each connection that is set up. * Second, the function tcp_isn_tick fires once per clock tick * and increments isn_offset as necessary so that sequence numbers * are incremented at approximately ISN_BYTES_PER_SECOND. The * random positive increments serve only to ensure that the same * exact sequence number is never sent out twice (as could otherwise * happen when a port is recycled in less than the system tick * interval.) * * net.inet.tcp.isn_reseed_interval controls the number of seconds * between seeding of isn_secret. This is normally set to zero, * as reseeding should not be necessary. * * Locking of the global variables isn_secret, isn_last_reseed, isn_offset, * isn_offset_old, and isn_ctx is performed using the TCP pcbinfo lock. In * general, this means holding an exclusive (write) lock. */ #define ISN_BYTES_PER_SECOND 1048576 #define ISN_STATIC_INCREMENT 4096 #define ISN_RANDOM_INCREMENT (4096 - 1) static VNET_DEFINE(u_char, isn_secret[32]); static VNET_DEFINE(int, isn_last); static VNET_DEFINE(int, isn_last_reseed); static VNET_DEFINE(u_int32_t, isn_offset); static VNET_DEFINE(u_int32_t, isn_offset_old); #define V_isn_secret VNET(isn_secret) #define V_isn_last VNET(isn_last) #define V_isn_last_reseed VNET(isn_last_reseed) #define V_isn_offset VNET(isn_offset) #define V_isn_offset_old VNET(isn_offset_old) tcp_seq tcp_new_isn(struct tcpcb *tp) { MD5_CTX isn_ctx; u_int32_t md5_buffer[4]; tcp_seq new_isn; u_int32_t projected_offset; INP_WLOCK_ASSERT(tp->t_inpcb); ISN_LOCK(); /* Seed if this is the first use, reseed if requested. */ if ((V_isn_last_reseed == 0) || ((V_tcp_isn_reseed_interval > 0) && (((u_int)V_isn_last_reseed + (u_int)V_tcp_isn_reseed_interval*hz) < (u_int)ticks))) { read_random(&V_isn_secret, sizeof(V_isn_secret)); V_isn_last_reseed = ticks; } /* Compute the md5 hash and return the ISN. */ MD5Init(&isn_ctx); MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_fport, sizeof(u_short)); MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_lport, sizeof(u_short)); #ifdef INET6 if ((tp->t_inpcb->inp_vflag & INP_IPV6) != 0) { MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->in6p_faddr, sizeof(struct in6_addr)); MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->in6p_laddr, sizeof(struct in6_addr)); } else #endif { MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_faddr, sizeof(struct in_addr)); MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_laddr, sizeof(struct in_addr)); } MD5Update(&isn_ctx, (u_char *) &V_isn_secret, sizeof(V_isn_secret)); MD5Final((u_char *) &md5_buffer, &isn_ctx); new_isn = (tcp_seq) md5_buffer[0]; V_isn_offset += ISN_STATIC_INCREMENT + (arc4random() & ISN_RANDOM_INCREMENT); if (ticks != V_isn_last) { projected_offset = V_isn_offset_old + ISN_BYTES_PER_SECOND / hz * (ticks - V_isn_last); if (SEQ_GT(projected_offset, V_isn_offset)) V_isn_offset = projected_offset; V_isn_offset_old = V_isn_offset; V_isn_last = ticks; } new_isn += V_isn_offset; ISN_UNLOCK(); return (new_isn); } /* * When a specific ICMP unreachable message is received and the * connection state is SYN-SENT, drop the connection. This behavior * is controlled by the icmp_may_rst sysctl. */ struct inpcb * tcp_drop_syn_sent(struct inpcb *inp, int errno) { struct tcpcb *tp; INP_INFO_WLOCK_ASSERT(&V_tcbinfo); INP_WLOCK_ASSERT(inp); if ((inp->inp_flags & INP_TIMEWAIT) || (inp->inp_flags & INP_DROPPED)) return (inp); tp = intotcpcb(inp); if (tp->t_state != TCPS_SYN_SENT) return (inp); tp = tcp_drop(tp, errno); if (tp != NULL) return (inp); else return (NULL); } /* * When `need fragmentation' ICMP is received, update our idea of the MSS * based on the new value. Also nudge TCP to send something, since we * know the packet we just sent was dropped. * This duplicates some code in the tcp_mss() function in tcp_input.c. */ static struct inpcb * tcp_mtudisc_notify(struct inpcb *inp, int error) { return (tcp_mtudisc(inp, -1)); } struct inpcb * tcp_mtudisc(struct inpcb *inp, int mtuoffer) { struct tcpcb *tp; struct socket *so; INP_WLOCK_ASSERT(inp); if ((inp->inp_flags & INP_TIMEWAIT) || (inp->inp_flags & INP_DROPPED)) return (inp); tp = intotcpcb(inp); KASSERT(tp != NULL, ("tcp_mtudisc: tp == NULL")); tcp_mss_update(tp, -1, mtuoffer, NULL, NULL); so = inp->inp_socket; SOCKBUF_LOCK(&so->so_snd); /* If the mss is larger than the socket buffer, decrease the mss. */ if (so->so_snd.sb_hiwat < tp->t_maxseg) tp->t_maxseg = so->so_snd.sb_hiwat; SOCKBUF_UNLOCK(&so->so_snd); TCPSTAT_INC(tcps_mturesent); tp->t_rtttime = 0; tp->snd_nxt = tp->snd_una; tcp_free_sackholes(tp); tp->snd_recover = tp->snd_max; if (tp->t_flags & TF_SACK_PERMIT) EXIT_FASTRECOVERY(tp->t_flags); tcp_output(tp); return (inp); } #ifdef INET /* * Look-up the routing entry to the peer of this inpcb. If no route * is found and it cannot be allocated, then return 0. This routine * is called by TCP routines that access the rmx structure and by * tcp_mss_update to get the peer/interface MTU. */ u_long tcp_maxmtu(struct in_conninfo *inc, struct tcp_ifcap *cap) { struct nhop4_extended nh_ext; struct ifnet *ifp; int error; u_long maxmtu = 0; KASSERT(inc != NULL, ("tcp_maxmtu with NULL in_conninfo pointer")); if (inc->inc_faddr.s_addr == INADDR_ANY) return (0); memset(&nh_ext, 0, sizeof(nh_ext)); - error = fib4_lookup_nh_extended(inc->inc_fibnum, inc->inc_faddr, 0, - &nh_ext); + error = fib4_lookup_nh_ext(inc->inc_fibnum, inc->inc_faddr, 0, + NHOP_LOOKUP_REF, &nh_ext); if (error == 0) { maxmtu = nh_ext.nh_mtu; /* Report additional interface capabilities. */ if (cap != NULL) { ifp = nh_ext.nh_ifp; if (ifp->if_capenable & IFCAP_TSO4 && ifp->if_hwassist & CSUM_TSO) { cap->ifcap |= CSUM_TSO; cap->tsomax = ifp->if_hw_tsomax; cap->tsomaxsegcount = ifp->if_hw_tsomaxsegcount; cap->tsomaxsegsize = ifp->if_hw_tsomaxsegsize; } } } return (maxmtu); } #endif /* INET */ #ifdef INET6 u_long tcp_maxmtu6(struct in_conninfo *inc, struct tcp_ifcap *cap) { struct route_in6 sro6; struct ifnet *ifp; u_long maxmtu = 0; KASSERT(inc != NULL, ("tcp_maxmtu6 with NULL in_conninfo pointer")); bzero(&sro6, sizeof(sro6)); if (!IN6_IS_ADDR_UNSPECIFIED(&inc->inc6_faddr)) { sro6.ro_dst.sin6_family = AF_INET6; sro6.ro_dst.sin6_len = sizeof(struct sockaddr_in6); sro6.ro_dst.sin6_addr = inc->inc6_faddr; in6_rtalloc_ign(&sro6, 0, inc->inc_fibnum); } if (sro6.ro_rt != NULL) { ifp = sro6.ro_rt->rt_ifp; if (sro6.ro_rt->rt_mtu == 0) maxmtu = IN6_LINKMTU(sro6.ro_rt->rt_ifp); else maxmtu = min(sro6.ro_rt->rt_mtu, IN6_LINKMTU(sro6.ro_rt->rt_ifp)); /* Report additional interface capabilities. */ if (cap != NULL) { if (ifp->if_capenable & IFCAP_TSO6 && ifp->if_hwassist & CSUM_TSO) { cap->ifcap |= CSUM_TSO; cap->tsomax = ifp->if_hw_tsomax; cap->tsomaxsegcount = ifp->if_hw_tsomaxsegcount; cap->tsomaxsegsize = ifp->if_hw_tsomaxsegsize; } } RTFREE(sro6.ro_rt); } return (maxmtu); } #endif /* INET6 */ #ifdef IPSEC /* compute ESP/AH header size for TCP, including outer IP header. */ size_t ipsec_hdrsiz_tcp(struct tcpcb *tp) { struct inpcb *inp; struct mbuf *m; size_t hdrsiz; struct ip *ip; #ifdef INET6 struct ip6_hdr *ip6; #endif struct tcphdr *th; if ((tp == NULL) || ((inp = tp->t_inpcb) == NULL)) return (0); m = m_gethdr(M_NOWAIT, MT_DATA); if (!m) return (0); #ifdef INET6 if ((inp->inp_vflag & INP_IPV6) != 0) { ip6 = mtod(m, struct ip6_hdr *); th = (struct tcphdr *)(ip6 + 1); m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + sizeof(struct tcphdr); tcpip_fillheaders(inp, ip6, th); hdrsiz = ipsec_hdrsiz(m, IPSEC_DIR_OUTBOUND, inp); } else #endif /* INET6 */ { ip = mtod(m, struct ip *); th = (struct tcphdr *)(ip + 1); m->m_pkthdr.len = m->m_len = sizeof(struct tcpiphdr); tcpip_fillheaders(inp, ip, th); hdrsiz = ipsec_hdrsiz(m, IPSEC_DIR_OUTBOUND, inp); } m_free(m); return (hdrsiz); } #endif /* IPSEC */ #ifdef TCP_SIGNATURE /* * Callback function invoked by m_apply() to digest TCP segment data * contained within an mbuf chain. */ static int tcp_signature_apply(void *fstate, void *data, u_int len) { MD5Update(fstate, (u_char *)data, len); return (0); } /* * XXX The key is retrieved from the system's PF_KEY SADB, by keying a * search with the destination IP address, and a 'magic SPI' to be * determined by the application. This is hardcoded elsewhere to 1179 */ struct secasvar * tcp_get_sav(struct mbuf *m, u_int direction) { union sockaddr_union dst; struct secasvar *sav; struct ip *ip; #ifdef INET6 struct ip6_hdr *ip6; char ip6buf[INET6_ADDRSTRLEN]; #endif /* Extract the destination from the IP header in the mbuf. */ bzero(&dst, sizeof(union sockaddr_union)); ip = mtod(m, struct ip *); #ifdef INET6 ip6 = NULL; /* Make the compiler happy. */ #endif switch (ip->ip_v) { #ifdef INET case IPVERSION: dst.sa.sa_len = sizeof(struct sockaddr_in); dst.sa.sa_family = AF_INET; dst.sin.sin_addr = (direction == IPSEC_DIR_INBOUND) ? ip->ip_src : ip->ip_dst; break; #endif #ifdef INET6 case (IPV6_VERSION >> 4): ip6 = mtod(m, struct ip6_hdr *); dst.sa.sa_len = sizeof(struct sockaddr_in6); dst.sa.sa_family = AF_INET6; dst.sin6.sin6_addr = (direction == IPSEC_DIR_INBOUND) ? ip6->ip6_src : ip6->ip6_dst; break; #endif default: return (NULL); /* NOTREACHED */ break; } /* Look up an SADB entry which matches the address of the peer. */ sav = KEY_ALLOCSA(&dst, IPPROTO_TCP, htonl(TCP_SIG_SPI)); if (sav == NULL) { ipseclog((LOG_ERR, "%s: SADB lookup failed for %s\n", __func__, (ip->ip_v == IPVERSION) ? inet_ntoa(dst.sin.sin_addr) : #ifdef INET6 (ip->ip_v == (IPV6_VERSION >> 4)) ? ip6_sprintf(ip6buf, &dst.sin6.sin6_addr) : #endif "(unsupported)")); } return (sav); } /* * Compute TCP-MD5 hash of a TCP segment. (RFC2385) * * Parameters: * m pointer to head of mbuf chain * len length of TCP segment data, excluding options * optlen length of TCP segment options * buf pointer to storage for computed MD5 digest * sav pointer to security assosiation * * We do this over ip, tcphdr, segment data, and the key in the SADB. * When called from tcp_input(), we can be sure that th_sum has been * zeroed out and verified already. * * Releases reference to SADB key before return. * * Return 0 if successful, otherwise return -1. * */ int tcp_signature_do_compute(struct mbuf *m, int len, int optlen, u_char *buf, struct secasvar *sav) { #ifdef INET struct ippseudo ippseudo; #endif MD5_CTX ctx; int doff; struct ip *ip; #ifdef INET struct ipovly *ipovly; #endif struct tcphdr *th; #ifdef INET6 struct ip6_hdr *ip6; struct in6_addr in6; uint32_t plen; uint16_t nhdr; #endif u_short savecsum; KASSERT(m != NULL, ("NULL mbuf chain")); KASSERT(buf != NULL, ("NULL signature pointer")); /* Extract the destination from the IP header in the mbuf. */ ip = mtod(m, struct ip *); #ifdef INET6 ip6 = NULL; /* Make the compiler happy. */ #endif MD5Init(&ctx); /* * Step 1: Update MD5 hash with IP(v6) pseudo-header. * * XXX The ippseudo header MUST be digested in network byte order, * or else we'll fail the regression test. Assume all fields we've * been doing arithmetic on have been in host byte order. * XXX One cannot depend on ipovly->ih_len here. When called from * tcp_output(), the underlying ip_len member has not yet been set. */ switch (ip->ip_v) { #ifdef INET case IPVERSION: ipovly = (struct ipovly *)ip; ippseudo.ippseudo_src = ipovly->ih_src; ippseudo.ippseudo_dst = ipovly->ih_dst; ippseudo.ippseudo_pad = 0; ippseudo.ippseudo_p = IPPROTO_TCP; ippseudo.ippseudo_len = htons(len + sizeof(struct tcphdr) + optlen); MD5Update(&ctx, (char *)&ippseudo, sizeof(struct ippseudo)); th = (struct tcphdr *)((u_char *)ip + sizeof(struct ip)); doff = sizeof(struct ip) + sizeof(struct tcphdr) + optlen; break; #endif #ifdef INET6 /* * RFC 2385, 2.0 Proposal * For IPv6, the pseudo-header is as described in RFC 2460, namely the * 128-bit source IPv6 address, 128-bit destination IPv6 address, zero- * extended next header value (to form 32 bits), and 32-bit segment * length. * Note: Upper-Layer Packet Length comes before Next Header. */ case (IPV6_VERSION >> 4): in6 = ip6->ip6_src; in6_clearscope(&in6); MD5Update(&ctx, (char *)&in6, sizeof(struct in6_addr)); in6 = ip6->ip6_dst; in6_clearscope(&in6); MD5Update(&ctx, (char *)&in6, sizeof(struct in6_addr)); plen = htonl(len + sizeof(struct tcphdr) + optlen); MD5Update(&ctx, (char *)&plen, sizeof(uint32_t)); nhdr = 0; MD5Update(&ctx, (char *)&nhdr, sizeof(uint8_t)); MD5Update(&ctx, (char *)&nhdr, sizeof(uint8_t)); MD5Update(&ctx, (char *)&nhdr, sizeof(uint8_t)); nhdr = IPPROTO_TCP; MD5Update(&ctx, (char *)&nhdr, sizeof(uint8_t)); th = (struct tcphdr *)((u_char *)ip6 + sizeof(struct ip6_hdr)); doff = sizeof(struct ip6_hdr) + sizeof(struct tcphdr) + optlen; break; #endif default: KEY_FREESAV(&sav); return (-1); /* NOTREACHED */ break; } /* * Step 2: Update MD5 hash with TCP header, excluding options. * The TCP checksum must be set to zero. */ savecsum = th->th_sum; th->th_sum = 0; MD5Update(&ctx, (char *)th, sizeof(struct tcphdr)); th->th_sum = savecsum; /* * Step 3: Update MD5 hash with TCP segment data. * Use m_apply() to avoid an early m_pullup(). */ if (len > 0) m_apply(m, doff, len, tcp_signature_apply, &ctx); /* * Step 4: Update MD5 hash with shared secret. */ MD5Update(&ctx, sav->key_auth->key_data, _KEYLEN(sav->key_auth)); MD5Final(buf, &ctx); key_sa_recordxfer(sav, m); KEY_FREESAV(&sav); return (0); } /* * Compute TCP-MD5 hash of a TCP segment. (RFC2385) * * Return 0 if successful, otherwise return -1. */ int tcp_signature_compute(struct mbuf *m, int _unused, int len, int optlen, u_char *buf, u_int direction) { struct secasvar *sav; if ((sav = tcp_get_sav(m, direction)) == NULL) return (-1); return (tcp_signature_do_compute(m, len, optlen, buf, sav)); } /* * Verify the TCP-MD5 hash of a TCP segment. (RFC2385) * * Parameters: * m pointer to head of mbuf chain * len length of TCP segment data, excluding options * optlen length of TCP segment options * buf pointer to storage for computed MD5 digest * direction direction of flow (IPSEC_DIR_INBOUND or OUTBOUND) * * Return 1 if successful, otherwise return 0. */ int tcp_signature_verify(struct mbuf *m, int off0, int tlen, int optlen, struct tcpopt *to, struct tcphdr *th, u_int tcpbflag) { char tmpdigest[TCP_SIGLEN]; if (tcp_sig_checksigs == 0) return (1); if ((tcpbflag & TF_SIGNATURE) == 0) { if ((to->to_flags & TOF_SIGNATURE) != 0) { /* * If this socket is not expecting signature but * the segment contains signature just fail. */ TCPSTAT_INC(tcps_sig_err_sigopt); TCPSTAT_INC(tcps_sig_rcvbadsig); return (0); } /* Signature is not expected, and not present in segment. */ return (1); } /* * If this socket is expecting signature but the segment does not * contain any just fail. */ if ((to->to_flags & TOF_SIGNATURE) == 0) { TCPSTAT_INC(tcps_sig_err_nosigopt); TCPSTAT_INC(tcps_sig_rcvbadsig); return (0); } if (tcp_signature_compute(m, off0, tlen, optlen, &tmpdigest[0], IPSEC_DIR_INBOUND) == -1) { TCPSTAT_INC(tcps_sig_err_buildsig); TCPSTAT_INC(tcps_sig_rcvbadsig); return (0); } if (bcmp(to->to_signature, &tmpdigest[0], TCP_SIGLEN) != 0) { TCPSTAT_INC(tcps_sig_rcvbadsig); return (0); } TCPSTAT_INC(tcps_sig_rcvgoodsig); return (1); } #endif /* TCP_SIGNATURE */ static int sysctl_drop(SYSCTL_HANDLER_ARGS) { /* addrs[0] is a foreign socket, addrs[1] is a local one. */ struct sockaddr_storage addrs[2]; struct inpcb *inp; struct tcpcb *tp; struct tcptw *tw; struct sockaddr_in *fin, *lin; #ifdef INET6 struct sockaddr_in6 *fin6, *lin6; #endif int error; inp = NULL; fin = lin = NULL; #ifdef INET6 fin6 = lin6 = NULL; #endif error = 0; if (req->oldptr != NULL || req->oldlen != 0) return (EINVAL); if (req->newptr == NULL) return (EPERM); if (req->newlen < sizeof(addrs)) return (ENOMEM); error = SYSCTL_IN(req, &addrs, sizeof(addrs)); if (error) return (error); switch (addrs[0].ss_family) { #ifdef INET6 case AF_INET6: fin6 = (struct sockaddr_in6 *)&addrs[0]; lin6 = (struct sockaddr_in6 *)&addrs[1]; if (fin6->sin6_len != sizeof(struct sockaddr_in6) || lin6->sin6_len != sizeof(struct sockaddr_in6)) return (EINVAL); if (IN6_IS_ADDR_V4MAPPED(&fin6->sin6_addr)) { if (!IN6_IS_ADDR_V4MAPPED(&lin6->sin6_addr)) return (EINVAL); in6_sin6_2_sin_in_sock((struct sockaddr *)&addrs[0]); in6_sin6_2_sin_in_sock((struct sockaddr *)&addrs[1]); fin = (struct sockaddr_in *)&addrs[0]; lin = (struct sockaddr_in *)&addrs[1]; break; } error = sa6_embedscope(fin6, V_ip6_use_defzone); if (error) return (error); error = sa6_embedscope(lin6, V_ip6_use_defzone); if (error) return (error); break; #endif #ifdef INET case AF_INET: fin = (struct sockaddr_in *)&addrs[0]; lin = (struct sockaddr_in *)&addrs[1]; if (fin->sin_len != sizeof(struct sockaddr_in) || lin->sin_len != sizeof(struct sockaddr_in)) return (EINVAL); break; #endif default: return (EINVAL); } INP_INFO_WLOCK(&V_tcbinfo); switch (addrs[0].ss_family) { #ifdef INET6 case AF_INET6: inp = in6_pcblookup(&V_tcbinfo, &fin6->sin6_addr, fin6->sin6_port, &lin6->sin6_addr, lin6->sin6_port, INPLOOKUP_WLOCKPCB, NULL); break; #endif #ifdef INET case AF_INET: inp = in_pcblookup(&V_tcbinfo, fin->sin_addr, fin->sin_port, lin->sin_addr, lin->sin_port, INPLOOKUP_WLOCKPCB, NULL); break; #endif } if (inp != NULL) { if (inp->inp_flags & INP_TIMEWAIT) { /* * XXXRW: There currently exists a state where an * inpcb is present, but its timewait state has been * discarded. For now, don't allow dropping of this * type of inpcb. */ tw = intotw(inp); if (tw != NULL) tcp_twclose(tw, 0); else INP_WUNLOCK(inp); } else if (!(inp->inp_flags & INP_DROPPED) && !(inp->inp_socket->so_options & SO_ACCEPTCONN)) { tp = intotcpcb(inp); tp = tcp_drop(tp, ECONNABORTED); if (tp != NULL) INP_WUNLOCK(inp); } else INP_WUNLOCK(inp); } else error = ESRCH; INP_INFO_WUNLOCK(&V_tcbinfo); return (error); } SYSCTL_PROC(_net_inet_tcp, TCPCTL_DROP, drop, CTLFLAG_VNET | CTLTYPE_STRUCT | CTLFLAG_WR | CTLFLAG_SKIP, NULL, 0, sysctl_drop, "", "Drop TCP connection"); /* * Generate a standardized TCP log line for use throughout the * tcp subsystem. Memory allocation is done with M_NOWAIT to * allow use in the interrupt context. * * NB: The caller MUST free(s, M_TCPLOG) the returned string. * NB: The function may return NULL if memory allocation failed. * * Due to header inclusion and ordering limitations the struct ip * and ip6_hdr pointers have to be passed as void pointers. */ char * tcp_log_vain(struct in_conninfo *inc, struct tcphdr *th, void *ip4hdr, const void *ip6hdr) { /* Is logging enabled? */ if (tcp_log_in_vain == 0) return (NULL); return (tcp_log_addr(inc, th, ip4hdr, ip6hdr)); } char * tcp_log_addrs(struct in_conninfo *inc, struct tcphdr *th, void *ip4hdr, const void *ip6hdr) { /* Is logging enabled? */ if (tcp_log_debug == 0) return (NULL); return (tcp_log_addr(inc, th, ip4hdr, ip6hdr)); } static char * tcp_log_addr(struct in_conninfo *inc, struct tcphdr *th, void *ip4hdr, const void *ip6hdr) { char *s, *sp; size_t size; struct ip *ip; #ifdef INET6 const struct ip6_hdr *ip6; ip6 = (const struct ip6_hdr *)ip6hdr; #endif /* INET6 */ ip = (struct ip *)ip4hdr; /* * The log line looks like this: * "TCP: [1.2.3.4]:50332 to [1.2.3.4]:80 tcpflags 0x2" */ size = sizeof("TCP: []:12345 to []:12345 tcpflags 0x2<>") + sizeof(PRINT_TH_FLAGS) + 1 + #ifdef INET6 2 * INET6_ADDRSTRLEN; #else 2 * INET_ADDRSTRLEN; #endif /* INET6 */ s = malloc(size, M_TCPLOG, M_ZERO|M_NOWAIT); if (s == NULL) return (NULL); strcat(s, "TCP: ["); sp = s + strlen(s); if (inc && ((inc->inc_flags & INC_ISIPV6) == 0)) { inet_ntoa_r(inc->inc_faddr, sp); sp = s + strlen(s); sprintf(sp, "]:%i to [", ntohs(inc->inc_fport)); sp = s + strlen(s); inet_ntoa_r(inc->inc_laddr, sp); sp = s + strlen(s); sprintf(sp, "]:%i", ntohs(inc->inc_lport)); #ifdef INET6 } else if (inc) { ip6_sprintf(sp, &inc->inc6_faddr); sp = s + strlen(s); sprintf(sp, "]:%i to [", ntohs(inc->inc_fport)); sp = s + strlen(s); ip6_sprintf(sp, &inc->inc6_laddr); sp = s + strlen(s); sprintf(sp, "]:%i", ntohs(inc->inc_lport)); } else if (ip6 && th) { ip6_sprintf(sp, &ip6->ip6_src); sp = s + strlen(s); sprintf(sp, "]:%i to [", ntohs(th->th_sport)); sp = s + strlen(s); ip6_sprintf(sp, &ip6->ip6_dst); sp = s + strlen(s); sprintf(sp, "]:%i", ntohs(th->th_dport)); #endif /* INET6 */ #ifdef INET } else if (ip && th) { inet_ntoa_r(ip->ip_src, sp); sp = s + strlen(s); sprintf(sp, "]:%i to [", ntohs(th->th_sport)); sp = s + strlen(s); inet_ntoa_r(ip->ip_dst, sp); sp = s + strlen(s); sprintf(sp, "]:%i", ntohs(th->th_dport)); #endif /* INET */ } else { free(s, M_TCPLOG); return (NULL); } sp = s + strlen(s); if (th) sprintf(sp, " tcpflags 0x%b", th->th_flags, PRINT_TH_FLAGS); if (*(s + size - 1) != '\0') panic("%s: string too long", __func__); return (s); } /* * A subroutine which makes it easy to track TCP state changes with DTrace. * This function shouldn't be called for t_state initializations that don't * correspond to actual TCP state transitions. */ void tcp_state_change(struct tcpcb *tp, int newstate) { #if defined(KDTRACE_HOOKS) int pstate = tp->t_state; #endif tp->t_state = newstate; TCP_PROBE6(state__change, NULL, tp, NULL, tp, NULL, pstate); } Index: projects/routing/sys/netpfil/ipfw/ip_fw2.c =================================================================== --- projects/routing/sys/netpfil/ipfw/ip_fw2.c (revision 287053) +++ projects/routing/sys/netpfil/ipfw/ip_fw2.c (revision 287054) @@ -1,2910 +1,2910 @@ /*- * Copyright (c) 2002-2009 Luigi Rizzo, Universita` di Pisa * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); /* * The FreeBSD IP packet firewall, main file */ #include "opt_ipfw.h" #include "opt_ipdivert.h" #include "opt_inet.h" #ifndef INET #error "IPFIREWALL requires INET" #endif /* INET */ #include "opt_inet6.h" #include "opt_ipsec.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include /* for ETHERTYPE_IP */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef INET6 #include #include #include #endif #include #include #include /* XXX for in_cksum */ #ifdef MAC #include #endif /* * static variables followed by global ones. * All ipfw global variables are here. */ static VNET_DEFINE(int, fw_deny_unknown_exthdrs); #define V_fw_deny_unknown_exthdrs VNET(fw_deny_unknown_exthdrs) static VNET_DEFINE(int, fw_permit_single_frag6) = 1; #define V_fw_permit_single_frag6 VNET(fw_permit_single_frag6) #ifdef IPFIREWALL_DEFAULT_TO_ACCEPT static int default_to_accept = 1; #else static int default_to_accept; #endif VNET_DEFINE(int, autoinc_step); VNET_DEFINE(int, fw_one_pass) = 1; VNET_DEFINE(unsigned int, fw_tables_max); VNET_DEFINE(unsigned int, fw_tables_sets) = 0; /* Don't use set-aware tables */ /* Use 128 tables by default */ static unsigned int default_fw_tables = IPFW_TABLES_DEFAULT; #ifndef LINEAR_SKIPTO static int jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num, int tablearg, int jump_backwards); #define JUMP(ch, f, num, targ, back) jump_fast(ch, f, num, targ, back) #else static int jump_linear(struct ip_fw_chain *chain, struct ip_fw *f, int num, int tablearg, int jump_backwards); #define JUMP(ch, f, num, targ, back) jump_linear(ch, f, num, targ, back) #endif /* * Each rule belongs to one of 32 different sets (0..31). * The variable set_disable contains one bit per set. * If the bit is set, all rules in the corresponding set * are disabled. Set RESVD_SET(31) is reserved for the default rule * and rules that are not deleted by the flush command, * and CANNOT be disabled. * Rules in set RESVD_SET can only be deleted individually. */ VNET_DEFINE(u_int32_t, set_disable); #define V_set_disable VNET(set_disable) VNET_DEFINE(int, fw_verbose); /* counter for ipfw_log(NULL...) */ VNET_DEFINE(u_int64_t, norule_counter); VNET_DEFINE(int, verbose_limit); /* layer3_chain contains the list of rules for layer 3 */ VNET_DEFINE(struct ip_fw_chain, layer3_chain); /* ipfw_vnet_ready controls when we are open for business */ VNET_DEFINE(int, ipfw_vnet_ready) = 0; VNET_DEFINE(int, ipfw_nat_ready) = 0; ipfw_nat_t *ipfw_nat_ptr = NULL; struct cfg_nat *(*lookup_nat_ptr)(struct nat_list *, int); ipfw_nat_cfg_t *ipfw_nat_cfg_ptr; ipfw_nat_cfg_t *ipfw_nat_del_ptr; ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr; ipfw_nat_cfg_t *ipfw_nat_get_log_ptr; #ifdef SYSCTL_NODE uint32_t dummy_def = IPFW_DEFAULT_RULE; static int sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS); static int sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS); SYSBEGIN(f3) SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass, CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_one_pass), 0, "Only do a single pass through ipfw when using dummynet(4)"); SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(autoinc_step), 0, "Rule number auto-increment step"); SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose, CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_verbose), 0, "Log matches to ipfw rules"); SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(verbose_limit), 0, "Set upper limit of matches of ipfw rules logged"); SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD, &dummy_def, 0, "The default/max possible rule number."); SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, tables_max, CTLFLAG_VNET | CTLTYPE_UINT | CTLFLAG_RW, 0, 0, sysctl_ipfw_table_num, "IU", "Maximum number of concurrently used tables"); SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, tables_sets, CTLFLAG_VNET | CTLTYPE_UINT | CTLFLAG_RW, 0, 0, sysctl_ipfw_tables_sets, "IU", "Use per-set namespace for tables"); SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, default_to_accept, CTLFLAG_RDTUN, &default_to_accept, 0, "Make the default rule accept all packets."); TUNABLE_INT("net.inet.ip.fw.tables_max", (int *)&default_fw_tables); SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, CTLFLAG_VNET | CTLFLAG_RD, &VNET_NAME(layer3_chain.n_rules), 0, "Number of static rules"); #ifdef INET6 SYSCTL_DECL(_net_inet6_ip6); SYSCTL_NODE(_net_inet6_ip6, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); SYSCTL_INT(_net_inet6_ip6_fw, OID_AUTO, deny_unknown_exthdrs, CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE, &VNET_NAME(fw_deny_unknown_exthdrs), 0, "Deny packets with unknown IPv6 Extension Headers"); SYSCTL_INT(_net_inet6_ip6_fw, OID_AUTO, permit_single_frag6, CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE, &VNET_NAME(fw_permit_single_frag6), 0, "Permit single packet IPv6 fragments"); #endif /* INET6 */ SYSEND #endif /* SYSCTL_NODE */ /* * Some macros used in the various matching options. * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T * Other macros just cast void * into the appropriate type */ #define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) #define TCP(p) ((struct tcphdr *)(p)) #define SCTP(p) ((struct sctphdr *)(p)) #define UDP(p) ((struct udphdr *)(p)) #define ICMP(p) ((struct icmphdr *)(p)) #define ICMP6(p) ((struct icmp6_hdr *)(p)) static __inline int icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd) { int type = icmp->icmp_type; return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<icmp_type; return (type <= ICMP_MAXTYPE && (TT & (1<arg1 or cmd->d[0]. * * We scan options and store the bits we find set. We succeed if * * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear * * The code is sometimes optimized not to store additional variables. */ static int flags_match(ipfw_insn *cmd, u_int8_t bits) { u_char want_clear; bits = ~bits; if ( ((cmd->arg1 & 0xff) & bits) != 0) return 0; /* some bits we want set were clear */ want_clear = (cmd->arg1 >> 8) & 0xff; if ( (want_clear & bits) != want_clear) return 0; /* some bits we want clear were set */ return 1; } static int ipopts_match(struct ip *ip, ipfw_insn *cmd) { int optlen, bits = 0; u_char *cp = (u_char *)(ip + 1); int x = (ip->ip_hl << 2) - sizeof (struct ip); for (; x > 0; x -= optlen, cp += optlen) { int opt = cp[IPOPT_OPTVAL]; if (opt == IPOPT_EOL) break; if (opt == IPOPT_NOP) optlen = 1; else { optlen = cp[IPOPT_OLEN]; if (optlen <= 0 || optlen > x) return 0; /* invalid or truncated */ } switch (opt) { default: break; case IPOPT_LSRR: bits |= IP_FW_IPOPT_LSRR; break; case IPOPT_SSRR: bits |= IP_FW_IPOPT_SSRR; break; case IPOPT_RR: bits |= IP_FW_IPOPT_RR; break; case IPOPT_TS: bits |= IP_FW_IPOPT_TS; break; } } return (flags_match(cmd, bits)); } static int tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd) { int optlen, bits = 0; u_char *cp = (u_char *)(tcp + 1); int x = (tcp->th_off << 2) - sizeof(struct tcphdr); for (; x > 0; x -= optlen, cp += optlen) { int opt = cp[0]; if (opt == TCPOPT_EOL) break; if (opt == TCPOPT_NOP) optlen = 1; else { optlen = cp[1]; if (optlen <= 0) break; } switch (opt) { default: break; case TCPOPT_MAXSEG: bits |= IP_FW_TCPOPT_MSS; break; case TCPOPT_WINDOW: bits |= IP_FW_TCPOPT_WINDOW; break; case TCPOPT_SACK_PERMITTED: case TCPOPT_SACK: bits |= IP_FW_TCPOPT_SACK; break; case TCPOPT_TIMESTAMP: bits |= IP_FW_TCPOPT_TS; break; } } return (flags_match(cmd, bits)); } static int iface_match(struct ifnet *ifp, ipfw_insn_if *cmd, struct ip_fw_chain *chain, uint32_t *tablearg) { if (ifp == NULL) /* no iface with this packet, match fails */ return (0); /* Check by name or by IP address */ if (cmd->name[0] != '\0') { /* match by name */ if (cmd->name[0] == '\1') /* use tablearg to match */ return ipfw_lookup_table_extended(chain, cmd->p.kidx, 0, &ifp->if_index, tablearg); /* Check name */ if (cmd->p.glob) { if (fnmatch(cmd->name, ifp->if_xname, 0) == 0) return(1); } else { if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0) return(1); } } else { #if !defined(USERSPACE) && defined(__FreeBSD__) /* and OSX too ? */ struct ifaddr *ia; if_addr_rlock(ifp); TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { if (ia->ifa_addr->sa_family != AF_INET) continue; if (cmd->p.ip.s_addr == ((struct sockaddr_in *) (ia->ifa_addr))->sin_addr.s_addr) { if_addr_runlock(ifp); return(1); /* match */ } } if_addr_runlock(ifp); #endif /* __FreeBSD__ */ } return(0); /* no match, fail ... */ } /* * The verify_path function checks if a route to the src exists and * if it is reachable via ifp (when provided). * * The 'verrevpath' option checks that the interface that an IP packet * arrives on is the same interface that traffic destined for the * packet's source address would be routed out of. * The 'versrcreach' option just checks that the source address is * reachable via any route (except default) in the routing table. * These two are a measure to block forged packets. This is also * commonly known as "anti-spoofing" or Unicast Reverse Path * Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs * is purposely reminiscent of the Cisco IOS command, * * ip verify unicast reverse-path * ip verify unicast source reachable-via any * * which implements the same functionality. But note that the syntax * is misleading, and the check may be performed on all IP packets * whether unicast, multicast, or broadcast. */ static int verify_path(struct in_addr src, struct ifnet *ifp, u_int fib) { #if defined(USERSPACE) || !defined(__FreeBSD__) return 0; #else struct nhop4_basic nh4; if (fib4_lookup_nh_basic(fib, src, 0, &nh4) != 0) return (0); /* * If ifp is provided, check for equality with rtentry. * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, * in order to pass packets injected back by if_simloop(): * routing entry (via lo0) for our own address * may exist, so we need to handle routing assymetry. */ if (ifp != NULL && ifp != nh4.nh_ifp) return (0); /* if no ifp provided, check if rtentry is not default route */ - if (ifp == NULL && (nh4.nh_flags & NHOP_DEFAULT) != 0) + if (ifp == NULL && (nh4.nh_flags & NHF_DEFAULT) != 0) return (0); /* or if this is a blackhole/reject route */ - if (ifp == NULL && (nh4.nh_flags & (NHOP_REJECT|NHOP_BLACKHOLE)) != 0) + if (ifp == NULL && (nh4.nh_flags & (NHF_REJECT|NHF_BLACKHOLE)) != 0) return (0); return (1); #endif /* __FreeBSD__ */ } #ifdef INET6 /* * ipv6 specific rules here... */ static __inline int icmp6type_match (int type, ipfw_insn_u32 *cmd) { return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) ); } static int flow6id_match( int curr_flow, ipfw_insn_u32 *cmd ) { int i; for (i=0; i <= cmd->o.arg1; ++i ) if (curr_flow == cmd->d[i] ) return 1; return 0; } /* support for IP6_*_ME opcodes */ static const struct in6_addr lla_mask = {{{ 0xff, 0xff, 0x00, 0x00, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }}}; static int ipfw_localip6(struct in6_addr *in6) { struct rm_priotracker in6_ifa_tracker; struct in6_ifaddr *ia; if (IN6_IS_ADDR_MULTICAST(in6)) return (0); if (!IN6_IS_ADDR_LINKLOCAL(in6)) return (in6_localip(in6)); IN6_IFADDR_RLOCK(&in6_ifa_tracker); TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) { if (!IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) continue; if (IN6_ARE_MASKED_ADDR_EQUAL(&ia->ia_addr.sin6_addr, in6, &lla_mask)) { IN6_IFADDR_RUNLOCK(&in6_ifa_tracker); return (1); } } IN6_IFADDR_RUNLOCK(&in6_ifa_tracker); return (0); } static int verify_path6(struct in6_addr *src, struct ifnet *ifp, u_int fib) { struct nhop6_basic nh6; if (fib6_lookup_nh_basic(fib, *src, 0, &nh6) != 0) return (0); /* If ifp is provided, check for equality with route table. */ if (ifp != NULL && ifp != nh6.nh_ifp) return (0); /* if no ifp provided, check if rtentry is not default route */ - if (ifp == NULL && (nh6.nh_flags & NHOP_DEFAULT) != 0) + if (ifp == NULL && (nh6.nh_flags & NHF_DEFAULT) != 0) return (0); /* or if this is a blackhole/reject route */ - if (ifp == NULL && (nh6.nh_flags & (NHOP_REJECT|NHOP_BLACKHOLE)) != 0) + if (ifp == NULL && (nh6.nh_flags & (NHF_REJECT|NHF_BLACKHOLE)) != 0) return (0); return (1); } static int is_icmp6_query(int icmp6_type) { if ((icmp6_type <= ICMP6_MAXTYPE) && (icmp6_type == ICMP6_ECHO_REQUEST || icmp6_type == ICMP6_MEMBERSHIP_QUERY || icmp6_type == ICMP6_WRUREQUEST || icmp6_type == ICMP6_FQDN_QUERY || icmp6_type == ICMP6_NI_QUERY)) return (1); return (0); } static void send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6) { struct mbuf *m; m = args->m; if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) { struct tcphdr *tcp; tcp = (struct tcphdr *)((char *)ip6 + hlen); if ((tcp->th_flags & TH_RST) == 0) { struct mbuf *m0; m0 = ipfw_send_pkt(args->m, &(args->f_id), ntohl(tcp->th_seq), ntohl(tcp->th_ack), tcp->th_flags | TH_RST); if (m0 != NULL) ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL); } FREE_PKT(m); } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */ #if 0 /* * Unlike above, the mbufs need to line up with the ip6 hdr, * as the contents are read. We need to m_adj() the * needed amount. * The mbuf will however be thrown away so we can adjust it. * Remember we did an m_pullup on it already so we * can make some assumptions about contiguousness. */ if (args->L3offset) m_adj(m, args->L3offset); #endif icmp6_error(m, ICMP6_DST_UNREACH, code, 0); } else FREE_PKT(m); args->m = NULL; } #endif /* INET6 */ /* * sends a reject message, consuming the mbuf passed as an argument. */ static void send_reject(struct ip_fw_args *args, int code, int iplen, struct ip *ip) { #if 0 /* XXX When ip is not guaranteed to be at mtod() we will * need to account for this */ * The mbuf will however be thrown away so we can adjust it. * Remember we did an m_pullup on it already so we * can make some assumptions about contiguousness. */ if (args->L3offset) m_adj(m, args->L3offset); #endif if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */ icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); } else if (args->f_id.proto == IPPROTO_TCP) { struct tcphdr *const tcp = L3HDR(struct tcphdr, mtod(args->m, struct ip *)); if ( (tcp->th_flags & TH_RST) == 0) { struct mbuf *m; m = ipfw_send_pkt(args->m, &(args->f_id), ntohl(tcp->th_seq), ntohl(tcp->th_ack), tcp->th_flags | TH_RST); if (m != NULL) ip_output(m, NULL, NULL, 0, NULL, NULL); } FREE_PKT(args->m); } else FREE_PKT(args->m); args->m = NULL; } /* * Support for uid/gid/jail lookup. These tests are expensive * (because we may need to look into the list of active sockets) * so we cache the results. ugid_lookupp is 0 if we have not * yet done a lookup, 1 if we succeeded, and -1 if we tried * and failed. The function always returns the match value. * We could actually spare the variable and use *uc, setting * it to '(void *)check_uidgid if we have no info, NULL if * we tried and failed, or any other value if successful. */ static int check_uidgid(ipfw_insn_u32 *insn, struct ip_fw_args *args, int *ugid_lookupp, struct ucred **uc) { #if defined(USERSPACE) return 0; // not supported in userspace #else #ifndef __FreeBSD__ /* XXX */ return cred_check(insn, proto, oif, dst_ip, dst_port, src_ip, src_port, (struct bsd_ucred *)uc, ugid_lookupp, ((struct mbuf *)inp)->m_skb); #else /* FreeBSD */ struct in_addr src_ip, dst_ip; struct inpcbinfo *pi; struct ipfw_flow_id *id; struct inpcb *pcb, *inp; struct ifnet *oif; int lookupflags; int match; id = &args->f_id; inp = args->inp; oif = args->oif; /* * Check to see if the UDP or TCP stack supplied us with * the PCB. If so, rather then holding a lock and looking * up the PCB, we can use the one that was supplied. */ if (inp && *ugid_lookupp == 0) { INP_LOCK_ASSERT(inp); if (inp->inp_socket != NULL) { *uc = crhold(inp->inp_cred); *ugid_lookupp = 1; } else *ugid_lookupp = -1; } /* * If we have already been here and the packet has no * PCB entry associated with it, then we can safely * assume that this is a no match. */ if (*ugid_lookupp == -1) return (0); if (id->proto == IPPROTO_TCP) { lookupflags = 0; pi = &V_tcbinfo; } else if (id->proto == IPPROTO_UDP) { lookupflags = INPLOOKUP_WILDCARD; pi = &V_udbinfo; } else return 0; lookupflags |= INPLOOKUP_RLOCKPCB; match = 0; if (*ugid_lookupp == 0) { if (id->addr_type == 6) { #ifdef INET6 if (oif == NULL) pcb = in6_pcblookup_mbuf(pi, &id->src_ip6, htons(id->src_port), &id->dst_ip6, htons(id->dst_port), lookupflags, oif, args->m); else pcb = in6_pcblookup_mbuf(pi, &id->dst_ip6, htons(id->dst_port), &id->src_ip6, htons(id->src_port), lookupflags, oif, args->m); #else *ugid_lookupp = -1; return (0); #endif } else { src_ip.s_addr = htonl(id->src_ip); dst_ip.s_addr = htonl(id->dst_ip); if (oif == NULL) pcb = in_pcblookup_mbuf(pi, src_ip, htons(id->src_port), dst_ip, htons(id->dst_port), lookupflags, oif, args->m); else pcb = in_pcblookup_mbuf(pi, dst_ip, htons(id->dst_port), src_ip, htons(id->src_port), lookupflags, oif, args->m); } if (pcb != NULL) { INP_RLOCK_ASSERT(pcb); *uc = crhold(pcb->inp_cred); *ugid_lookupp = 1; INP_RUNLOCK(pcb); } if (*ugid_lookupp == 0) { /* * We tried and failed, set the variable to -1 * so we will not try again on this packet. */ *ugid_lookupp = -1; return (0); } } if (insn->o.opcode == O_UID) match = ((*uc)->cr_uid == (uid_t)insn->d[0]); else if (insn->o.opcode == O_GID) match = groupmember((gid_t)insn->d[0], *uc); else if (insn->o.opcode == O_JAIL) match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]); return (match); #endif /* __FreeBSD__ */ #endif /* not supported in userspace */ } /* * Helper function to set args with info on the rule after the matching * one. slot is precise, whereas we guess rule_id as they are * assigned sequentially. */ static inline void set_match(struct ip_fw_args *args, int slot, struct ip_fw_chain *chain) { args->rule.chain_id = chain->id; args->rule.slot = slot + 1; /* we use 0 as a marker */ args->rule.rule_id = 1 + chain->map[slot]->id; args->rule.rulenum = chain->map[slot]->rulenum; } #ifndef LINEAR_SKIPTO /* * Helper function to enable cached rule lookups using * cached_id and cached_pos fields in ipfw rule. */ static int jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num, int tablearg, int jump_backwards) { int f_pos; /* If possible use cached f_pos (in f->cached_pos), * whose version is written in f->cached_id * (horrible hacks to avoid changing the ABI). */ if (num != IP_FW_TARG && f->cached_id == chain->id) f_pos = f->cached_pos; else { int i = IP_FW_ARG_TABLEARG(chain, num, skipto); /* make sure we do not jump backward */ if (jump_backwards == 0 && i <= f->rulenum) i = f->rulenum + 1; if (chain->idxmap != NULL) f_pos = chain->idxmap[i]; else f_pos = ipfw_find_rule(chain, i, 0); /* update the cache */ if (num != IP_FW_TARG) { f->cached_id = chain->id; f->cached_pos = f_pos; } } return (f_pos); } #else /* * Helper function to enable real fast rule lookups. */ static int jump_linear(struct ip_fw_chain *chain, struct ip_fw *f, int num, int tablearg, int jump_backwards) { int f_pos; num = IP_FW_ARG_TABLEARG(chain, num, skipto); /* make sure we do not jump backward */ if (jump_backwards == 0 && num <= f->rulenum) num = f->rulenum + 1; f_pos = chain->idxmap[num]; return (f_pos); } #endif #define TARG(k, f) IP_FW_ARG_TABLEARG(chain, k, f) /* * The main check routine for the firewall. * * All arguments are in args so we can modify them and return them * back to the caller. * * Parameters: * * args->m (in/out) The packet; we set to NULL when/if we nuke it. * Starts with the IP header. * args->eh (in) Mac header if present, NULL for layer3 packet. * args->L3offset Number of bytes bypassed if we came from L2. * e.g. often sizeof(eh) ** NOTYET ** * args->oif Outgoing interface, NULL if packet is incoming. * The incoming interface is in the mbuf. (in) * args->divert_rule (in/out) * Skip up to the first rule past this rule number; * upon return, non-zero port number for divert or tee. * * args->rule Pointer to the last matching rule (in/out) * args->next_hop Socket we are forwarding to (out). * args->next_hop6 IPv6 next hop we are forwarding to (out). * args->f_id Addresses grabbed from the packet (out) * args->rule.info a cookie depending on rule action * * Return value: * * IP_FW_PASS the packet must be accepted * IP_FW_DENY the packet must be dropped * IP_FW_DIVERT divert packet, port in m_tag * IP_FW_TEE tee packet, port in m_tag * IP_FW_DUMMYNET to dummynet, pipe in args->cookie * IP_FW_NETGRAPH into netgraph, cookie args->cookie * args->rule contains the matching rule, * args->rule.info has additional information. * */ int ipfw_chk(struct ip_fw_args *args) { /* * Local variables holding state while processing a packet: * * IMPORTANT NOTE: to speed up the processing of rules, there * are some assumption on the values of the variables, which * are documented here. Should you change them, please check * the implementation of the various instructions to make sure * that they still work. * * args->eh The MAC header. It is non-null for a layer2 * packet, it is NULL for a layer-3 packet. * **notyet** * args->L3offset Offset in the packet to the L3 (IP or equiv.) header. * * m | args->m Pointer to the mbuf, as received from the caller. * It may change if ipfw_chk() does an m_pullup, or if it * consumes the packet because it calls send_reject(). * XXX This has to change, so that ipfw_chk() never modifies * or consumes the buffer. * ip is the beginning of the ip(4 or 6) header. * Calculated by adding the L3offset to the start of data. * (Until we start using L3offset, the packet is * supposed to start with the ip header). */ struct mbuf *m = args->m; struct ip *ip = mtod(m, struct ip *); /* * For rules which contain uid/gid or jail constraints, cache * a copy of the users credentials after the pcb lookup has been * executed. This will speed up the processing of rules with * these types of constraints, as well as decrease contention * on pcb related locks. */ #ifndef __FreeBSD__ struct bsd_ucred ucred_cache; #else struct ucred *ucred_cache = NULL; #endif int ucred_lookup = 0; /* * oif | args->oif If NULL, ipfw_chk has been called on the * inbound path (ether_input, ip_input). * If non-NULL, ipfw_chk has been called on the outbound path * (ether_output, ip_output). */ struct ifnet *oif = args->oif; int f_pos = 0; /* index of current rule in the array */ int retval = 0; /* * hlen The length of the IP header. */ u_int hlen = 0; /* hlen >0 means we have an IP pkt */ /* * offset The offset of a fragment. offset != 0 means that * we have a fragment at this offset of an IPv4 packet. * offset == 0 means that (if this is an IPv4 packet) * this is the first or only fragment. * For IPv6 offset|ip6f_mf == 0 means there is no Fragment Header * or there is a single packet fragement (fragement header added * without needed). We will treat a single packet fragment as if * there was no fragment header (or log/block depending on the * V_fw_permit_single_frag6 sysctl setting). */ u_short offset = 0; u_short ip6f_mf = 0; /* * Local copies of addresses. They are only valid if we have * an IP packet. * * proto The protocol. Set to 0 for non-ip packets, * or to the protocol read from the packet otherwise. * proto != 0 means that we have an IPv4 packet. * * src_port, dst_port port numbers, in HOST format. Only * valid for TCP and UDP packets. * * src_ip, dst_ip ip addresses, in NETWORK format. * Only valid for IPv4 packets. */ uint8_t proto; uint16_t src_port = 0, dst_port = 0; /* NOTE: host format */ struct in_addr src_ip, dst_ip; /* NOTE: network format */ uint16_t iplen=0; int pktlen; uint16_t etype = 0; /* Host order stored ether type */ /* * dyn_dir = MATCH_UNKNOWN when rules unchecked, * MATCH_NONE when checked and not matched (q = NULL), * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) */ int dyn_dir = MATCH_UNKNOWN; ipfw_dyn_rule *q = NULL; struct ip_fw_chain *chain = &V_layer3_chain; /* * We store in ulp a pointer to the upper layer protocol header. * In the ipv4 case this is easy to determine from the header, * but for ipv6 we might have some additional headers in the middle. * ulp is NULL if not found. */ void *ulp = NULL; /* upper layer protocol pointer. */ /* XXX ipv6 variables */ int is_ipv6 = 0; uint8_t icmp6_type = 0; uint16_t ext_hd = 0; /* bits vector for extension header filtering */ /* end of ipv6 variables */ int is_ipv4 = 0; int done = 0; /* flag to exit the outer loop */ IPFW_RLOCK_TRACKER; if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready)) return (IP_FW_PASS); /* accept */ dst_ip.s_addr = 0; /* make sure it is initialized */ src_ip.s_addr = 0; /* make sure it is initialized */ pktlen = m->m_pkthdr.len; args->f_id.fib = M_GETFIB(m); /* note mbuf not altered) */ proto = args->f_id.proto = 0; /* mark f_id invalid */ /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */ /* * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous, * then it sets p to point at the offset "len" in the mbuf. WARNING: the * pointer might become stale after other pullups (but we never use it * this way). */ #define PULLUP_TO(_len, p, T) PULLUP_LEN(_len, p, sizeof(T)) #define PULLUP_LEN(_len, p, T) \ do { \ int x = (_len) + T; \ if ((m)->m_len < x) { \ args->m = m = m_pullup(m, x); \ if (m == NULL) \ goto pullup_failed; \ } \ p = (mtod(m, char *) + (_len)); \ } while (0) /* * if we have an ether header, */ if (args->eh) etype = ntohs(args->eh->ether_type); /* Identify IP packets and fill up variables. */ if (pktlen >= sizeof(struct ip6_hdr) && (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) { struct ip6_hdr *ip6 = (struct ip6_hdr *)ip; is_ipv6 = 1; args->f_id.addr_type = 6; hlen = sizeof(struct ip6_hdr); proto = ip6->ip6_nxt; /* Search extension headers to find upper layer protocols */ while (ulp == NULL && offset == 0) { switch (proto) { case IPPROTO_ICMPV6: PULLUP_TO(hlen, ulp, struct icmp6_hdr); icmp6_type = ICMP6(ulp)->icmp6_type; break; case IPPROTO_TCP: PULLUP_TO(hlen, ulp, struct tcphdr); dst_port = TCP(ulp)->th_dport; src_port = TCP(ulp)->th_sport; /* save flags for dynamic rules */ args->f_id._flags = TCP(ulp)->th_flags; break; case IPPROTO_SCTP: PULLUP_TO(hlen, ulp, struct sctphdr); src_port = SCTP(ulp)->src_port; dst_port = SCTP(ulp)->dest_port; break; case IPPROTO_UDP: PULLUP_TO(hlen, ulp, struct udphdr); dst_port = UDP(ulp)->uh_dport; src_port = UDP(ulp)->uh_sport; break; case IPPROTO_HOPOPTS: /* RFC 2460 */ PULLUP_TO(hlen, ulp, struct ip6_hbh); ext_hd |= EXT_HOPOPTS; hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; ulp = NULL; break; case IPPROTO_ROUTING: /* RFC 2460 */ PULLUP_TO(hlen, ulp, struct ip6_rthdr); switch (((struct ip6_rthdr *)ulp)->ip6r_type) { case 0: ext_hd |= EXT_RTHDR0; break; case 2: ext_hd |= EXT_RTHDR2; break; default: if (V_fw_verbose) printf("IPFW2: IPV6 - Unknown " "Routing Header type(%d)\n", ((struct ip6_rthdr *) ulp)->ip6r_type); if (V_fw_deny_unknown_exthdrs) return (IP_FW_DENY); break; } ext_hd |= EXT_ROUTING; hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3; proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt; ulp = NULL; break; case IPPROTO_FRAGMENT: /* RFC 2460 */ PULLUP_TO(hlen, ulp, struct ip6_frag); ext_hd |= EXT_FRAGMENT; hlen += sizeof (struct ip6_frag); proto = ((struct ip6_frag *)ulp)->ip6f_nxt; offset = ((struct ip6_frag *)ulp)->ip6f_offlg & IP6F_OFF_MASK; ip6f_mf = ((struct ip6_frag *)ulp)->ip6f_offlg & IP6F_MORE_FRAG; if (V_fw_permit_single_frag6 == 0 && offset == 0 && ip6f_mf == 0) { if (V_fw_verbose) printf("IPFW2: IPV6 - Invalid " "Fragment Header\n"); if (V_fw_deny_unknown_exthdrs) return (IP_FW_DENY); break; } args->f_id.extra = ntohl(((struct ip6_frag *)ulp)->ip6f_ident); ulp = NULL; break; case IPPROTO_DSTOPTS: /* RFC 2460 */ PULLUP_TO(hlen, ulp, struct ip6_hbh); ext_hd |= EXT_DSTOPTS; hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; ulp = NULL; break; case IPPROTO_AH: /* RFC 2402 */ PULLUP_TO(hlen, ulp, struct ip6_ext); ext_hd |= EXT_AH; hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2; proto = ((struct ip6_ext *)ulp)->ip6e_nxt; ulp = NULL; break; case IPPROTO_ESP: /* RFC 2406 */ PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */ /* Anything past Seq# is variable length and * data past this ext. header is encrypted. */ ext_hd |= EXT_ESP; break; case IPPROTO_NONE: /* RFC 2460 */ /* * Packet ends here, and IPv6 header has * already been pulled up. If ip6e_len!=0 * then octets must be ignored. */ ulp = ip; /* non-NULL to get out of loop. */ break; case IPPROTO_OSPFIGP: /* XXX OSPF header check? */ PULLUP_TO(hlen, ulp, struct ip6_ext); break; case IPPROTO_PIM: /* XXX PIM header check? */ PULLUP_TO(hlen, ulp, struct pim); break; case IPPROTO_CARP: PULLUP_TO(hlen, ulp, struct carp_header); if (((struct carp_header *)ulp)->carp_version != CARP_VERSION) return (IP_FW_DENY); if (((struct carp_header *)ulp)->carp_type != CARP_ADVERTISEMENT) return (IP_FW_DENY); break; case IPPROTO_IPV6: /* RFC 2893 */ PULLUP_TO(hlen, ulp, struct ip6_hdr); break; case IPPROTO_IPV4: /* RFC 2893 */ PULLUP_TO(hlen, ulp, struct ip); break; default: if (V_fw_verbose) printf("IPFW2: IPV6 - Unknown " "Extension Header(%d), ext_hd=%x\n", proto, ext_hd); if (V_fw_deny_unknown_exthdrs) return (IP_FW_DENY); PULLUP_TO(hlen, ulp, struct ip6_ext); break; } /*switch */ } ip = mtod(m, struct ip *); ip6 = (struct ip6_hdr *)ip; args->f_id.src_ip6 = ip6->ip6_src; args->f_id.dst_ip6 = ip6->ip6_dst; args->f_id.src_ip = 0; args->f_id.dst_ip = 0; args->f_id.flow_id6 = ntohl(ip6->ip6_flow); } else if (pktlen >= sizeof(struct ip) && (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) { is_ipv4 = 1; hlen = ip->ip_hl << 2; args->f_id.addr_type = 4; /* * Collect parameters into local variables for faster matching. */ proto = ip->ip_p; src_ip = ip->ip_src; dst_ip = ip->ip_dst; offset = ntohs(ip->ip_off) & IP_OFFMASK; iplen = ntohs(ip->ip_len); pktlen = iplen < pktlen ? iplen : pktlen; if (offset == 0) { switch (proto) { case IPPROTO_TCP: PULLUP_TO(hlen, ulp, struct tcphdr); dst_port = TCP(ulp)->th_dport; src_port = TCP(ulp)->th_sport; /* save flags for dynamic rules */ args->f_id._flags = TCP(ulp)->th_flags; break; case IPPROTO_SCTP: PULLUP_TO(hlen, ulp, struct sctphdr); src_port = SCTP(ulp)->src_port; dst_port = SCTP(ulp)->dest_port; break; case IPPROTO_UDP: PULLUP_TO(hlen, ulp, struct udphdr); dst_port = UDP(ulp)->uh_dport; src_port = UDP(ulp)->uh_sport; break; case IPPROTO_ICMP: PULLUP_TO(hlen, ulp, struct icmphdr); //args->f_id.flags = ICMP(ulp)->icmp_type; break; default: break; } } ip = mtod(m, struct ip *); args->f_id.src_ip = ntohl(src_ip.s_addr); args->f_id.dst_ip = ntohl(dst_ip.s_addr); } #undef PULLUP_TO if (proto) { /* we may have port numbers, store them */ args->f_id.proto = proto; args->f_id.src_port = src_port = ntohs(src_port); args->f_id.dst_port = dst_port = ntohs(dst_port); } IPFW_PF_RLOCK(chain); if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */ IPFW_PF_RUNLOCK(chain); return (IP_FW_PASS); /* accept */ } if (args->rule.slot) { /* * Packet has already been tagged as a result of a previous * match on rule args->rule aka args->rule_id (PIPE, QUEUE, * REASS, NETGRAPH, DIVERT/TEE...) * Validate the slot and continue from the next one * if still present, otherwise do a lookup. */ f_pos = (args->rule.chain_id == chain->id) ? args->rule.slot : ipfw_find_rule(chain, args->rule.rulenum, args->rule.rule_id); } else { f_pos = 0; } /* * Now scan the rules, and parse microinstructions for each rule. * We have two nested loops and an inner switch. Sometimes we * need to break out of one or both loops, or re-enter one of * the loops with updated variables. Loop variables are: * * f_pos (outer loop) points to the current rule. * On output it points to the matching rule. * done (outer loop) is used as a flag to break the loop. * l (inner loop) residual length of current rule. * cmd points to the current microinstruction. * * We break the inner loop by setting l=0 and possibly * cmdlen=0 if we don't want to advance cmd. * We break the outer loop by setting done=1 * We can restart the inner loop by setting l>0 and f_pos, f, cmd * as needed. */ for (; f_pos < chain->n_rules; f_pos++) { ipfw_insn *cmd; uint32_t tablearg = 0; int l, cmdlen, skip_or; /* skip rest of OR block */ struct ip_fw *f; f = chain->map[f_pos]; if (V_set_disable & (1 << f->set) ) continue; skip_or = 0; for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; l -= cmdlen, cmd += cmdlen) { int match; /* * check_body is a jump target used when we find a * CHECK_STATE, and need to jump to the body of * the target rule. */ /* check_body: */ cmdlen = F_LEN(cmd); /* * An OR block (insn_1 || .. || insn_n) has the * F_OR bit set in all but the last instruction. * The first match will set "skip_or", and cause * the following instructions to be skipped until * past the one with the F_OR bit clear. */ if (skip_or) { /* skip this instruction */ if ((cmd->len & F_OR) == 0) skip_or = 0; /* next one is good */ continue; } match = 0; /* set to 1 if we succeed */ switch (cmd->opcode) { /* * The first set of opcodes compares the packet's * fields with some pattern, setting 'match' if a * match is found. At the end of the loop there is * logic to deal with F_NOT and F_OR flags associated * with the opcode. */ case O_NOP: match = 1; break; case O_FORWARD_MAC: printf("ipfw: opcode %d unimplemented\n", cmd->opcode); break; case O_GID: case O_UID: case O_JAIL: /* * We only check offset == 0 && proto != 0, * as this ensures that we have a * packet with the ports info. */ if (offset != 0) break; if (proto == IPPROTO_TCP || proto == IPPROTO_UDP) match = check_uidgid( (ipfw_insn_u32 *)cmd, args, &ucred_lookup, #ifdef __FreeBSD__ &ucred_cache); #else (void *)&ucred_cache); #endif break; case O_RECV: match = iface_match(m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd, chain, &tablearg); break; case O_XMIT: match = iface_match(oif, (ipfw_insn_if *)cmd, chain, &tablearg); break; case O_VIA: match = iface_match(oif ? oif : m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd, chain, &tablearg); break; case O_MACADDR2: if (args->eh != NULL) { /* have MAC header */ u_int32_t *want = (u_int32_t *) ((ipfw_insn_mac *)cmd)->addr; u_int32_t *mask = (u_int32_t *) ((ipfw_insn_mac *)cmd)->mask; u_int32_t *hdr = (u_int32_t *)args->eh; match = ( want[0] == (hdr[0] & mask[0]) && want[1] == (hdr[1] & mask[1]) && want[2] == (hdr[2] & mask[2]) ); } break; case O_MAC_TYPE: if (args->eh != NULL) { u_int16_t *p = ((ipfw_insn_u16 *)cmd)->ports; int i; for (i = cmdlen - 1; !match && i>0; i--, p += 2) match = (etype >= p[0] && etype <= p[1]); } break; case O_FRAG: match = (offset != 0); break; case O_IN: /* "out" is "not in" */ match = (oif == NULL); break; case O_LAYER2: match = (args->eh != NULL); break; case O_DIVERTED: { /* For diverted packets, args->rule.info * contains the divert port (in host format) * reason and direction. */ uint32_t i = args->rule.info; match = (i&IPFW_IS_MASK) == IPFW_IS_DIVERT && cmd->arg1 & ((i & IPFW_INFO_IN) ? 1 : 2); } break; case O_PROTO: /* * We do not allow an arg of 0 so the * check of "proto" only suffices. */ match = (proto == cmd->arg1); break; case O_IP_SRC: match = is_ipv4 && (((ipfw_insn_ip *)cmd)->addr.s_addr == src_ip.s_addr); break; case O_IP_SRC_LOOKUP: case O_IP_DST_LOOKUP: if (is_ipv4) { uint32_t key = (cmd->opcode == O_IP_DST_LOOKUP) ? dst_ip.s_addr : src_ip.s_addr; uint32_t v = 0; if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) { /* generic lookup. The key must be * in 32bit big-endian format. */ v = ((ipfw_insn_u32 *)cmd)->d[1]; if (v == 0) key = dst_ip.s_addr; else if (v == 1) key = src_ip.s_addr; else if (v == 6) /* dscp */ key = (ip->ip_tos >> 2) & 0x3f; else if (offset != 0) break; else if (proto != IPPROTO_TCP && proto != IPPROTO_UDP) break; else if (v == 2) key = dst_port; else if (v == 3) key = src_port; #ifndef USERSPACE else if (v == 4 || v == 5) { check_uidgid( (ipfw_insn_u32 *)cmd, args, &ucred_lookup, #ifdef __FreeBSD__ &ucred_cache); if (v == 4 /* O_UID */) key = ucred_cache->cr_uid; else if (v == 5 /* O_JAIL */) key = ucred_cache->cr_prison->pr_id; #else /* !__FreeBSD__ */ (void *)&ucred_cache); if (v ==4 /* O_UID */) key = ucred_cache.uid; else if (v == 5 /* O_JAIL */) key = ucred_cache.xid; #endif /* !__FreeBSD__ */ } #endif /* !USERSPACE */ else break; } match = ipfw_lookup_table(chain, cmd->arg1, key, &v); if (!match) break; if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) match = ((ipfw_insn_u32 *)cmd)->d[0] == v; else tablearg = v; } else if (is_ipv6) { uint32_t v = 0; void *pkey = (cmd->opcode == O_IP_DST_LOOKUP) ? &args->f_id.dst_ip6: &args->f_id.src_ip6; match = ipfw_lookup_table_extended(chain, cmd->arg1, sizeof(struct in6_addr), pkey, &v); if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) match = ((ipfw_insn_u32 *)cmd)->d[0] == v; if (match) tablearg = v; } break; case O_IP_FLOW_LOOKUP: { uint32_t v = 0; match = ipfw_lookup_table_extended(chain, cmd->arg1, 0, &args->f_id, &v); if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) match = ((ipfw_insn_u32 *)cmd)->d[0] == v; if (match) tablearg = v; } break; case O_IP_SRC_MASK: case O_IP_DST_MASK: if (is_ipv4) { uint32_t a = (cmd->opcode == O_IP_DST_MASK) ? dst_ip.s_addr : src_ip.s_addr; uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; int i = cmdlen-1; for (; !match && i>0; i-= 2, p+= 2) match = (p[0] == (a & p[1])); } break; case O_IP_SRC_ME: if (is_ipv4) { struct ifnet *tif; INADDR_TO_IFP(src_ip, tif); match = (tif != NULL); break; } #ifdef INET6 /* FALLTHROUGH */ case O_IP6_SRC_ME: match= is_ipv6 && ipfw_localip6(&args->f_id.src_ip6); #endif break; case O_IP_DST_SET: case O_IP_SRC_SET: if (is_ipv4) { u_int32_t *d = (u_int32_t *)(cmd+1); u_int32_t addr = cmd->opcode == O_IP_DST_SET ? args->f_id.dst_ip : args->f_id.src_ip; if (addr < d[0]) break; addr -= d[0]; /* subtract base */ match = (addr < cmd->arg1) && ( d[ 1 + (addr>>5)] & (1<<(addr & 0x1f)) ); } break; case O_IP_DST: match = is_ipv4 && (((ipfw_insn_ip *)cmd)->addr.s_addr == dst_ip.s_addr); break; case O_IP_DST_ME: if (is_ipv4) { struct ifnet *tif; INADDR_TO_IFP(dst_ip, tif); match = (tif != NULL); break; } #ifdef INET6 /* FALLTHROUGH */ case O_IP6_DST_ME: match= is_ipv6 && ipfw_localip6(&args->f_id.dst_ip6); #endif break; case O_IP_SRCPORT: case O_IP_DSTPORT: /* * offset == 0 && proto != 0 is enough * to guarantee that we have a * packet with port info. */ if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) && offset == 0) { u_int16_t x = (cmd->opcode == O_IP_SRCPORT) ? src_port : dst_port ; u_int16_t *p = ((ipfw_insn_u16 *)cmd)->ports; int i; for (i = cmdlen - 1; !match && i>0; i--, p += 2) match = (x>=p[0] && x<=p[1]); } break; case O_ICMPTYPE: match = (offset == 0 && proto==IPPROTO_ICMP && icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) ); break; #ifdef INET6 case O_ICMP6TYPE: match = is_ipv6 && offset == 0 && proto==IPPROTO_ICMPV6 && icmp6type_match( ICMP6(ulp)->icmp6_type, (ipfw_insn_u32 *)cmd); break; #endif /* INET6 */ case O_IPOPT: match = (is_ipv4 && ipopts_match(ip, cmd) ); break; case O_IPVER: match = (is_ipv4 && cmd->arg1 == ip->ip_v); break; case O_IPID: case O_IPLEN: case O_IPTTL: if (is_ipv4) { /* only for IP packets */ uint16_t x; uint16_t *p; int i; if (cmd->opcode == O_IPLEN) x = iplen; else if (cmd->opcode == O_IPTTL) x = ip->ip_ttl; else /* must be IPID */ x = ntohs(ip->ip_id); if (cmdlen == 1) { match = (cmd->arg1 == x); break; } /* otherwise we have ranges */ p = ((ipfw_insn_u16 *)cmd)->ports; i = cmdlen - 1; for (; !match && i>0; i--, p += 2) match = (x >= p[0] && x <= p[1]); } break; case O_IPPRECEDENCE: match = (is_ipv4 && (cmd->arg1 == (ip->ip_tos & 0xe0)) ); break; case O_IPTOS: match = (is_ipv4 && flags_match(cmd, ip->ip_tos)); break; case O_DSCP: { uint32_t *p; uint16_t x; p = ((ipfw_insn_u32 *)cmd)->d; if (is_ipv4) x = ip->ip_tos >> 2; else if (is_ipv6) { uint8_t *v; v = &((struct ip6_hdr *)ip)->ip6_vfc; x = (*v & 0x0F) << 2; v++; x |= *v >> 6; } else break; /* DSCP bitmask is stored as low_u32 high_u32 */ if (x > 32) match = *(p + 1) & (1 << (x - 32)); else match = *p & (1 << x); } break; case O_TCPDATALEN: if (proto == IPPROTO_TCP && offset == 0) { struct tcphdr *tcp; uint16_t x; uint16_t *p; int i; tcp = TCP(ulp); x = iplen - ((ip->ip_hl + tcp->th_off) << 2); if (cmdlen == 1) { match = (cmd->arg1 == x); break; } /* otherwise we have ranges */ p = ((ipfw_insn_u16 *)cmd)->ports; i = cmdlen - 1; for (; !match && i>0; i--, p += 2) match = (x >= p[0] && x <= p[1]); } break; case O_TCPFLAGS: match = (proto == IPPROTO_TCP && offset == 0 && flags_match(cmd, TCP(ulp)->th_flags)); break; case O_TCPOPTS: if (proto == IPPROTO_TCP && offset == 0 && ulp){ PULLUP_LEN(hlen, ulp, (TCP(ulp)->th_off << 2)); match = tcpopts_match(TCP(ulp), cmd); } break; case O_TCPSEQ: match = (proto == IPPROTO_TCP && offset == 0 && ((ipfw_insn_u32 *)cmd)->d[0] == TCP(ulp)->th_seq); break; case O_TCPACK: match = (proto == IPPROTO_TCP && offset == 0 && ((ipfw_insn_u32 *)cmd)->d[0] == TCP(ulp)->th_ack); break; case O_TCPWIN: if (proto == IPPROTO_TCP && offset == 0) { uint16_t x; uint16_t *p; int i; x = ntohs(TCP(ulp)->th_win); if (cmdlen == 1) { match = (cmd->arg1 == x); break; } /* Otherwise we have ranges. */ p = ((ipfw_insn_u16 *)cmd)->ports; i = cmdlen - 1; for (; !match && i > 0; i--, p += 2) match = (x >= p[0] && x <= p[1]); } break; case O_ESTAB: /* reject packets which have SYN only */ /* XXX should i also check for TH_ACK ? */ match = (proto == IPPROTO_TCP && offset == 0 && (TCP(ulp)->th_flags & (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); break; case O_ALTQ: { struct pf_mtag *at; struct m_tag *mtag; ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; /* * ALTQ uses mbuf tags from another * packet filtering system - pf(4). * We allocate a tag in its format * and fill it in, pretending to be pf(4). */ match = 1; at = pf_find_mtag(m); if (at != NULL && at->qid != 0) break; mtag = m_tag_get(PACKET_TAG_PF, sizeof(struct pf_mtag), M_NOWAIT | M_ZERO); if (mtag == NULL) { /* * Let the packet fall back to the * default ALTQ. */ break; } m_tag_prepend(m, mtag); at = (struct pf_mtag *)(mtag + 1); at->qid = altq->qid; at->hdr = ip; break; } case O_LOG: ipfw_log(chain, f, hlen, args, m, oif, offset | ip6f_mf, tablearg, ip); match = 1; break; case O_PROB: match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); break; case O_VERREVPATH: /* Outgoing packets automatically pass/match */ match = ((oif != NULL) || (m->m_pkthdr.rcvif == NULL) || ( #ifdef INET6 is_ipv6 ? verify_path6(&(args->f_id.src_ip6), m->m_pkthdr.rcvif, args->f_id.fib) : #endif verify_path(src_ip, m->m_pkthdr.rcvif, args->f_id.fib))); break; case O_VERSRCREACH: /* Outgoing packets automatically pass/match */ match = (hlen > 0 && ((oif != NULL) || #ifdef INET6 is_ipv6 ? verify_path6(&(args->f_id.src_ip6), NULL, args->f_id.fib) : #endif verify_path(src_ip, NULL, args->f_id.fib))); break; case O_ANTISPOOF: /* Outgoing packets automatically pass/match */ if (oif == NULL && hlen > 0 && ( (is_ipv4 && in_localaddr(src_ip)) #ifdef INET6 || (is_ipv6 && in6_localaddr(&(args->f_id.src_ip6))) #endif )) match = #ifdef INET6 is_ipv6 ? verify_path6( &(args->f_id.src_ip6), m->m_pkthdr.rcvif, args->f_id.fib) : #endif verify_path(src_ip, m->m_pkthdr.rcvif, args->f_id.fib); else match = 1; break; case O_IPSEC: #ifdef IPSEC match = (m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); #endif /* otherwise no match */ break; #ifdef INET6 case O_IP6_SRC: match = is_ipv6 && IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6, &((ipfw_insn_ip6 *)cmd)->addr6); break; case O_IP6_DST: match = is_ipv6 && IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6, &((ipfw_insn_ip6 *)cmd)->addr6); break; case O_IP6_SRC_MASK: case O_IP6_DST_MASK: if (is_ipv6) { int i = cmdlen - 1; struct in6_addr p; struct in6_addr *d = &((ipfw_insn_ip6 *)cmd)->addr6; for (; !match && i > 0; d += 2, i -= F_INSN_SIZE(struct in6_addr) * 2) { p = (cmd->opcode == O_IP6_SRC_MASK) ? args->f_id.src_ip6: args->f_id.dst_ip6; APPLY_MASK(&p, &d[1]); match = IN6_ARE_ADDR_EQUAL(&d[0], &p); } } break; case O_FLOW6ID: match = is_ipv6 && flow6id_match(args->f_id.flow_id6, (ipfw_insn_u32 *) cmd); break; case O_EXT_HDR: match = is_ipv6 && (ext_hd & ((ipfw_insn *) cmd)->arg1); break; case O_IP6: match = is_ipv6; break; #endif case O_IP4: match = is_ipv4; break; case O_TAG: { struct m_tag *mtag; uint32_t tag = TARG(cmd->arg1, tag); /* Packet is already tagged with this tag? */ mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL); /* We have `untag' action when F_NOT flag is * present. And we must remove this mtag from * mbuf and reset `match' to zero (`match' will * be inversed later). * Otherwise we should allocate new mtag and * push it into mbuf. */ if (cmd->len & F_NOT) { /* `untag' action */ if (mtag != NULL) m_tag_delete(m, mtag); match = 0; } else { if (mtag == NULL) { mtag = m_tag_alloc( MTAG_IPFW, tag, 0, M_NOWAIT); if (mtag != NULL) m_tag_prepend(m, mtag); } match = 1; } break; } case O_FIB: /* try match the specified fib */ if (args->f_id.fib == cmd->arg1) match = 1; break; case O_SOCKARG: { #ifndef USERSPACE /* not supported in userspace */ struct inpcb *inp = args->inp; struct inpcbinfo *pi; if (is_ipv6) /* XXX can we remove this ? */ break; if (proto == IPPROTO_TCP) pi = &V_tcbinfo; else if (proto == IPPROTO_UDP) pi = &V_udbinfo; else break; /* * XXXRW: so_user_cookie should almost * certainly be inp_user_cookie? */ /* For incomming packet, lookup up the inpcb using the src/dest ip/port tuple */ if (inp == NULL) { inp = in_pcblookup(pi, src_ip, htons(src_port), dst_ip, htons(dst_port), INPLOOKUP_RLOCKPCB, NULL); if (inp != NULL) { tablearg = inp->inp_socket->so_user_cookie; if (tablearg) match = 1; INP_RUNLOCK(inp); } } else { if (inp->inp_socket) { tablearg = inp->inp_socket->so_user_cookie; if (tablearg) match = 1; } } #endif /* !USERSPACE */ break; } case O_TAGGED: { struct m_tag *mtag; uint32_t tag = TARG(cmd->arg1, tag); if (cmdlen == 1) { match = m_tag_locate(m, MTAG_IPFW, tag, NULL) != NULL; break; } /* we have ranges */ for (mtag = m_tag_first(m); mtag != NULL && !match; mtag = m_tag_next(m, mtag)) { uint16_t *p; int i; if (mtag->m_tag_cookie != MTAG_IPFW) continue; p = ((ipfw_insn_u16 *)cmd)->ports; i = cmdlen - 1; for(; !match && i > 0; i--, p += 2) match = mtag->m_tag_id >= p[0] && mtag->m_tag_id <= p[1]; } break; } /* * The second set of opcodes represents 'actions', * i.e. the terminal part of a rule once the packet * matches all previous patterns. * Typically there is only one action for each rule, * and the opcode is stored at the end of the rule * (but there are exceptions -- see below). * * In general, here we set retval and terminate the * outer loop (would be a 'break 3' in some language, * but we need to set l=0, done=1) * * Exceptions: * O_COUNT and O_SKIPTO actions: * instead of terminating, we jump to the next rule * (setting l=0), or to the SKIPTO target (setting * f/f_len, cmd and l as needed), respectively. * * O_TAG, O_LOG and O_ALTQ action parameters: * perform some action and set match = 1; * * O_LIMIT and O_KEEP_STATE: these opcodes are * not real 'actions', and are stored right * before the 'action' part of the rule. * These opcodes try to install an entry in the * state tables; if successful, we continue with * the next opcode (match=1; break;), otherwise * the packet must be dropped (set retval, * break loops with l=0, done=1) * * O_PROBE_STATE and O_CHECK_STATE: these opcodes * cause a lookup of the state table, and a jump * to the 'action' part of the parent rule * if an entry is found, or * (CHECK_STATE only) a jump to the next rule if * the entry is not found. * The result of the lookup is cached so that * further instances of these opcodes become NOPs. * The jump to the next rule is done by setting * l=0, cmdlen=0. */ case O_LIMIT: case O_KEEP_STATE: if (ipfw_install_state(chain, f, (ipfw_insn_limit *)cmd, args, tablearg)) { /* error or limit violation */ retval = IP_FW_DENY; l = 0; /* exit inner loop */ done = 1; /* exit outer loop */ } match = 1; break; case O_PROBE_STATE: case O_CHECK_STATE: /* * dynamic rules are checked at the first * keep-state or check-state occurrence, * with the result being stored in dyn_dir. * The compiler introduces a PROBE_STATE * instruction for us when we have a * KEEP_STATE (because PROBE_STATE needs * to be run first). */ if (dyn_dir == MATCH_UNKNOWN && (q = ipfw_lookup_dyn_rule(&args->f_id, &dyn_dir, proto == IPPROTO_TCP ? TCP(ulp) : NULL)) != NULL) { /* * Found dynamic entry, update stats * and jump to the 'action' part of * the parent rule by setting * f, cmd, l and clearing cmdlen. */ IPFW_INC_DYN_COUNTER(q, pktlen); /* XXX we would like to have f_pos * readily accessible in the dynamic * rule, instead of having to * lookup q->rule. */ f = q->rule; f_pos = ipfw_find_rule(chain, f->rulenum, f->id); cmd = ACTION_PTR(f); l = f->cmd_len - f->act_ofs; ipfw_dyn_unlock(q); cmdlen = 0; match = 1; break; } /* * Dynamic entry not found. If CHECK_STATE, * skip to next rule, if PROBE_STATE just * ignore and continue with next opcode. */ if (cmd->opcode == O_CHECK_STATE) l = 0; /* exit inner loop */ match = 1; break; case O_ACCEPT: retval = 0; /* accept */ l = 0; /* exit inner loop */ done = 1; /* exit outer loop */ break; case O_PIPE: case O_QUEUE: set_match(args, f_pos, chain); args->rule.info = TARG(cmd->arg1, pipe); if (cmd->opcode == O_PIPE) args->rule.info |= IPFW_IS_PIPE; if (V_fw_one_pass) args->rule.info |= IPFW_ONEPASS; retval = IP_FW_DUMMYNET; l = 0; /* exit inner loop */ done = 1; /* exit outer loop */ break; case O_DIVERT: case O_TEE: if (args->eh) /* not on layer 2 */ break; /* otherwise this is terminal */ l = 0; /* exit inner loop */ done = 1; /* exit outer loop */ retval = (cmd->opcode == O_DIVERT) ? IP_FW_DIVERT : IP_FW_TEE; set_match(args, f_pos, chain); args->rule.info = TARG(cmd->arg1, divert); break; case O_COUNT: IPFW_INC_RULE_COUNTER(f, pktlen); l = 0; /* exit inner loop */ break; case O_SKIPTO: IPFW_INC_RULE_COUNTER(f, pktlen); f_pos = JUMP(chain, f, cmd->arg1, tablearg, 0); /* * Skip disabled rules, and re-enter * the inner loop with the correct * f_pos, f, l and cmd. * Also clear cmdlen and skip_or */ for (; f_pos < chain->n_rules - 1 && (V_set_disable & (1 << chain->map[f_pos]->set)); f_pos++) ; /* Re-enter the inner loop at the skipto rule. */ f = chain->map[f_pos]; l = f->cmd_len; cmd = f->cmd; match = 1; cmdlen = 0; skip_or = 0; continue; break; /* not reached */ case O_CALLRETURN: { /* * Implementation of `subroutine' call/return, * in the stack carried in an mbuf tag. This * is different from `skipto' in that any call * address is possible (`skipto' must prevent * backward jumps to avoid endless loops). * We have `return' action when F_NOT flag is * present. The `m_tag_id' field is used as * stack pointer. */ struct m_tag *mtag; uint16_t jmpto, *stack; #define IS_CALL ((cmd->len & F_NOT) == 0) #define IS_RETURN ((cmd->len & F_NOT) != 0) /* * Hand-rolled version of m_tag_locate() with * wildcard `type'. * If not already tagged, allocate new tag. */ mtag = m_tag_first(m); while (mtag != NULL) { if (mtag->m_tag_cookie == MTAG_IPFW_CALL) break; mtag = m_tag_next(m, mtag); } if (mtag == NULL && IS_CALL) { mtag = m_tag_alloc(MTAG_IPFW_CALL, 0, IPFW_CALLSTACK_SIZE * sizeof(uint16_t), M_NOWAIT); if (mtag != NULL) m_tag_prepend(m, mtag); } /* * On error both `call' and `return' just * continue with next rule. */ if (IS_RETURN && (mtag == NULL || mtag->m_tag_id == 0)) { l = 0; /* exit inner loop */ break; } if (IS_CALL && (mtag == NULL || mtag->m_tag_id >= IPFW_CALLSTACK_SIZE)) { printf("ipfw: call stack error, " "go to next rule\n"); l = 0; /* exit inner loop */ break; } IPFW_INC_RULE_COUNTER(f, pktlen); stack = (uint16_t *)(mtag + 1); /* * The `call' action may use cached f_pos * (in f->next_rule), whose version is written * in f->next_rule. * The `return' action, however, doesn't have * fixed jump address in cmd->arg1 and can't use * cache. */ if (IS_CALL) { stack[mtag->m_tag_id] = f->rulenum; mtag->m_tag_id++; f_pos = JUMP(chain, f, cmd->arg1, tablearg, 1); } else { /* `return' action */ mtag->m_tag_id--; jmpto = stack[mtag->m_tag_id] + 1; f_pos = ipfw_find_rule(chain, jmpto, 0); } /* * Skip disabled rules, and re-enter * the inner loop with the correct * f_pos, f, l and cmd. * Also clear cmdlen and skip_or */ for (; f_pos < chain->n_rules - 1 && (V_set_disable & (1 << chain->map[f_pos]->set)); f_pos++) ; /* Re-enter the inner loop at the dest rule. */ f = chain->map[f_pos]; l = f->cmd_len; cmd = f->cmd; cmdlen = 0; skip_or = 0; continue; break; /* NOTREACHED */ } #undef IS_CALL #undef IS_RETURN case O_REJECT: /* * Drop the packet and send a reject notice * if the packet is not ICMP (or is an ICMP * query), and it is not multicast/broadcast. */ if (hlen > 0 && is_ipv4 && offset == 0 && (proto != IPPROTO_ICMP || is_icmp_query(ICMP(ulp))) && !(m->m_flags & (M_BCAST|M_MCAST)) && !IN_MULTICAST(ntohl(dst_ip.s_addr))) { send_reject(args, cmd->arg1, iplen, ip); m = args->m; } /* FALLTHROUGH */ #ifdef INET6 case O_UNREACH6: if (hlen > 0 && is_ipv6 && ((offset & IP6F_OFF_MASK) == 0) && (proto != IPPROTO_ICMPV6 || (is_icmp6_query(icmp6_type) == 1)) && !(m->m_flags & (M_BCAST|M_MCAST)) && !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) { send_reject6( args, cmd->arg1, hlen, (struct ip6_hdr *)ip); m = args->m; } /* FALLTHROUGH */ #endif case O_DENY: retval = IP_FW_DENY; l = 0; /* exit inner loop */ done = 1; /* exit outer loop */ break; case O_FORWARD_IP: if (args->eh) /* not valid on layer2 pkts */ break; if (q == NULL || q->rule != f || dyn_dir == MATCH_FORWARD) { struct sockaddr_in *sa; sa = &(((ipfw_insn_sa *)cmd)->sa); if (sa->sin_addr.s_addr == INADDR_ANY) { #ifdef INET6 /* * We use O_FORWARD_IP opcode for * fwd rule with tablearg, but tables * now support IPv6 addresses. And * when we are inspecting IPv6 packet, * we can use nh6 field from * table_value as next_hop6 address. */ if (is_ipv6) { struct sockaddr_in6 *sa6; sa6 = args->next_hop6 = &args->hopstore6; sa6->sin6_family = AF_INET6; sa6->sin6_len = sizeof(*sa6); sa6->sin6_addr = TARG_VAL( chain, tablearg, nh6); /* * Set sin6_scope_id only for * link-local unicast addresses. */ if (IN6_IS_ADDR_LINKLOCAL( &sa6->sin6_addr)) sa6->sin6_scope_id = TARG_VAL(chain, tablearg, zoneid); } else #endif { sa = args->next_hop = &args->hopstore; sa->sin_family = AF_INET; sa->sin_len = sizeof(*sa); sa->sin_addr.s_addr = htonl( TARG_VAL(chain, tablearg, nh4)); } } else { args->next_hop = sa; } } retval = IP_FW_PASS; l = 0; /* exit inner loop */ done = 1; /* exit outer loop */ break; #ifdef INET6 case O_FORWARD_IP6: if (args->eh) /* not valid on layer2 pkts */ break; if (q == NULL || q->rule != f || dyn_dir == MATCH_FORWARD) { struct sockaddr_in6 *sin6; sin6 = &(((ipfw_insn_sa6 *)cmd)->sa); args->next_hop6 = sin6; } retval = IP_FW_PASS; l = 0; /* exit inner loop */ done = 1; /* exit outer loop */ break; #endif case O_NETGRAPH: case O_NGTEE: set_match(args, f_pos, chain); args->rule.info = TARG(cmd->arg1, netgraph); if (V_fw_one_pass) args->rule.info |= IPFW_ONEPASS; retval = (cmd->opcode == O_NETGRAPH) ? IP_FW_NETGRAPH : IP_FW_NGTEE; l = 0; /* exit inner loop */ done = 1; /* exit outer loop */ break; case O_SETFIB: { uint32_t fib; IPFW_INC_RULE_COUNTER(f, pktlen); fib = TARG(cmd->arg1, fib) & 0x7FFFF; if (fib >= rt_numfibs) fib = 0; M_SETFIB(m, fib); args->f_id.fib = fib; l = 0; /* exit inner loop */ break; } case O_SETDSCP: { uint16_t code; code = TARG(cmd->arg1, dscp) & 0x3F; l = 0; /* exit inner loop */ if (is_ipv4) { uint16_t old; old = *(uint16_t *)ip; ip->ip_tos = (code << 2) | (ip->ip_tos & 0x03); ip->ip_sum = cksum_adjust(ip->ip_sum, old, *(uint16_t *)ip); } else if (is_ipv6) { uint8_t *v; v = &((struct ip6_hdr *)ip)->ip6_vfc; *v = (*v & 0xF0) | (code >> 2); v++; *v = (*v & 0x3F) | ((code & 0x03) << 6); } else break; IPFW_INC_RULE_COUNTER(f, pktlen); break; } case O_NAT: l = 0; /* exit inner loop */ done = 1; /* exit outer loop */ if (!IPFW_NAT_LOADED) { retval = IP_FW_DENY; break; } struct cfg_nat *t; int nat_id; set_match(args, f_pos, chain); /* Check if this is 'global' nat rule */ if (cmd->arg1 == 0) { retval = ipfw_nat_ptr(args, NULL, m); break; } t = ((ipfw_insn_nat *)cmd)->nat; if (t == NULL) { nat_id = TARG(cmd->arg1, nat); t = (*lookup_nat_ptr)(&chain->nat, nat_id); if (t == NULL) { retval = IP_FW_DENY; break; } if (cmd->arg1 != IP_FW_TARG) ((ipfw_insn_nat *)cmd)->nat = t; } retval = ipfw_nat_ptr(args, t, m); break; case O_REASS: { int ip_off; IPFW_INC_RULE_COUNTER(f, pktlen); l = 0; /* in any case exit inner loop */ ip_off = ntohs(ip->ip_off); /* if not fragmented, go to next rule */ if ((ip_off & (IP_MF | IP_OFFMASK)) == 0) break; args->m = m = ip_reass(m); /* * do IP header checksum fixup. */ if (m == NULL) { /* fragment got swallowed */ retval = IP_FW_DENY; } else { /* good, packet complete */ int hlen; ip = mtod(m, struct ip *); hlen = ip->ip_hl << 2; ip->ip_sum = 0; if (hlen == sizeof(struct ip)) ip->ip_sum = in_cksum_hdr(ip); else ip->ip_sum = in_cksum(m, hlen); retval = IP_FW_REASS; set_match(args, f_pos, chain); } done = 1; /* exit outer loop */ break; } default: panic("-- unknown opcode %d\n", cmd->opcode); } /* end of switch() on opcodes */ /* * if we get here with l=0, then match is irrelevant. */ if (cmd->len & F_NOT) match = !match; if (match) { if (cmd->len & F_OR) skip_or = 1; } else { if (!(cmd->len & F_OR)) /* not an OR block, */ break; /* try next rule */ } } /* end of inner loop, scan opcodes */ #undef PULLUP_LEN if (done) break; /* next_rule:; */ /* try next rule */ } /* end of outer for, scan rules */ if (done) { struct ip_fw *rule = chain->map[f_pos]; /* Update statistics */ IPFW_INC_RULE_COUNTER(rule, pktlen); } else { retval = IP_FW_DENY; printf("ipfw: ouch!, skip past end of rules, denying packet\n"); } IPFW_PF_RUNLOCK(chain); #ifdef __FreeBSD__ if (ucred_cache != NULL) crfree(ucred_cache); #endif return (retval); pullup_failed: if (V_fw_verbose) printf("ipfw: pullup failed\n"); return (IP_FW_DENY); } /* * Set maximum number of tables that can be used in given VNET ipfw instance. */ #ifdef SYSCTL_NODE static int sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS) { int error; unsigned int ntables; ntables = V_fw_tables_max; error = sysctl_handle_int(oidp, &ntables, 0, req); /* Read operation or some error */ if ((error != 0) || (req->newptr == NULL)) return (error); return (ipfw_resize_tables(&V_layer3_chain, ntables)); } /* * Switches table namespace between global and per-set. */ static int sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS) { int error; unsigned int sets; sets = V_fw_tables_sets; error = sysctl_handle_int(oidp, &sets, 0, req); /* Read operation or some error */ if ((error != 0) || (req->newptr == NULL)) return (error); return (ipfw_switch_tables_namespace(&V_layer3_chain, sets)); } #endif /* * Module and VNET glue */ /* * Stuff that must be initialised only on boot or module load */ static int ipfw_init(void) { int error = 0; /* * Only print out this stuff the first time around, * when called from the sysinit code. */ printf("ipfw2 " #ifdef INET6 "(+ipv6) " #endif "initialized, divert %s, nat %s, " "default to %s, logging ", #ifdef IPDIVERT "enabled", #else "loadable", #endif #ifdef IPFIREWALL_NAT "enabled", #else "loadable", #endif default_to_accept ? "accept" : "deny"); /* * Note: V_xxx variables can be accessed here but the vnet specific * initializer may not have been called yet for the VIMAGE case. * Tuneables will have been processed. We will print out values for * the default vnet. * XXX This should all be rationalized AFTER 8.0 */ if (V_fw_verbose == 0) printf("disabled\n"); else if (V_verbose_limit == 0) printf("unlimited\n"); else printf("limited to %d packets/entry by default\n", V_verbose_limit); /* Check user-supplied table count for validness */ if (default_fw_tables > IPFW_TABLES_MAX) default_fw_tables = IPFW_TABLES_MAX; ipfw_init_sopt_handler(); ipfw_log_bpf(1); /* init */ ipfw_iface_init(); return (error); } /* * Called for the removal of the last instance only on module unload. */ static void ipfw_destroy(void) { ipfw_iface_destroy(); ipfw_log_bpf(0); /* uninit */ ipfw_destroy_sopt_handler(); printf("IP firewall unloaded\n"); } /* * Stuff that must be initialized for every instance * (including the first of course). */ static int vnet_ipfw_init(const void *unused) { int error, first; struct ip_fw *rule = NULL; struct ip_fw_chain *chain; chain = &V_layer3_chain; first = IS_DEFAULT_VNET(curvnet) ? 1 : 0; /* First set up some values that are compile time options */ V_autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ V_fw_deny_unknown_exthdrs = 1; #ifdef IPFIREWALL_VERBOSE V_fw_verbose = 1; #endif #ifdef IPFIREWALL_VERBOSE_LIMIT V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT; #endif #ifdef IPFIREWALL_NAT LIST_INIT(&chain->nat); #endif /* Init shared services hash table */ ipfw_init_srv(chain); ipfw_init_obj_rewriter(); ipfw_init_counters(); /* insert the default rule and create the initial map */ chain->n_rules = 1; chain->map = malloc(sizeof(struct ip_fw *), M_IPFW, M_WAITOK | M_ZERO); rule = ipfw_alloc_rule(chain, sizeof(struct ip_fw)); /* Set initial number of tables */ V_fw_tables_max = default_fw_tables; error = ipfw_init_tables(chain, first); if (error) { printf("ipfw2: setting up tables failed\n"); free(chain->map, M_IPFW); free(rule, M_IPFW); return (ENOSPC); } /* fill and insert the default rule */ rule->act_ofs = 0; rule->rulenum = IPFW_DEFAULT_RULE; rule->cmd_len = 1; rule->set = RESVD_SET; rule->cmd[0].len = 1; rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY; chain->default_rule = chain->map[0] = rule; chain->id = rule->id = 1; /* Pre-calculate rules length for legacy dump format */ chain->static_len = sizeof(struct ip_fw_rule0); IPFW_LOCK_INIT(chain); ipfw_dyn_init(chain); #ifdef LINEAR_SKIPTO ipfw_init_skipto_cache(chain); #endif /* First set up some values that are compile time options */ V_ipfw_vnet_ready = 1; /* Open for business */ /* * Hook the sockopt handler and pfil hooks for ipv4 and ipv6. * Even if the latter two fail we still keep the module alive * because the sockopt and layer2 paths are still useful. * ipfw[6]_hook return 0 on success, ENOENT on failure, * so we can ignore the exact return value and just set a flag. * * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so * changes in the underlying (per-vnet) variables trigger * immediate hook()/unhook() calls. * In layer2 we have the same behaviour, except that V_ether_ipfw * is checked on each packet because there are no pfil hooks. */ V_ip_fw_ctl_ptr = ipfw_ctl3; error = ipfw_attach_hooks(1); return (error); } /* * Called for the removal of each instance. */ static int vnet_ipfw_uninit(const void *unused) { struct ip_fw *reap; struct ip_fw_chain *chain = &V_layer3_chain; int i, last; V_ipfw_vnet_ready = 0; /* tell new callers to go away */ /* * disconnect from ipv4, ipv6, layer2 and sockopt. * Then grab, release and grab again the WLOCK so we make * sure the update is propagated and nobody will be in. */ (void)ipfw_attach_hooks(0 /* detach */); V_ip_fw_ctl_ptr = NULL; last = IS_DEFAULT_VNET(curvnet) ? 1 : 0; IPFW_UH_WLOCK(chain); IPFW_UH_WUNLOCK(chain); IPFW_UH_WLOCK(chain); IPFW_WLOCK(chain); ipfw_dyn_uninit(0); /* run the callout_drain */ IPFW_WUNLOCK(chain); reap = NULL; IPFW_WLOCK(chain); for (i = 0; i < chain->n_rules; i++) ipfw_reap_add(chain, &reap, chain->map[i]); free(chain->map, M_IPFW); #ifdef LINEAR_SKIPTO ipfw_destroy_skipto_cache(chain); #endif IPFW_WUNLOCK(chain); IPFW_UH_WUNLOCK(chain); ipfw_destroy_tables(chain, last); if (reap != NULL) ipfw_reap_rules(reap); vnet_ipfw_iface_destroy(chain); ipfw_destroy_srv(chain); IPFW_LOCK_DESTROY(chain); ipfw_dyn_uninit(1); /* free the remaining parts */ ipfw_destroy_counters(); ipfw_destroy_obj_rewriter(); return (0); } /* * Module event handler. * In general we have the choice of handling most of these events by the * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to * use the SYSINIT handlers as they are more capable of expressing the * flow of control during module and vnet operations, so this is just * a skeleton. Note there is no SYSINIT equivalent of the module * SHUTDOWN handler, but we don't have anything to do in that case anyhow. */ static int ipfw_modevent(module_t mod, int type, void *unused) { int err = 0; switch (type) { case MOD_LOAD: /* Called once at module load or * system boot if compiled in. */ break; case MOD_QUIESCE: /* Called before unload. May veto unloading. */ break; case MOD_UNLOAD: /* Called during unload. */ break; case MOD_SHUTDOWN: /* Called during system shutdown. */ break; default: err = EOPNOTSUPP; break; } return err; } static moduledata_t ipfwmod = { "ipfw", ipfw_modevent, 0 }; /* Define startup order. */ #define IPFW_SI_SUB_FIREWALL SI_SUB_PROTO_IFATTACHDOMAIN #define IPFW_MODEVENT_ORDER (SI_ORDER_ANY - 255) /* On boot slot in here. */ #define IPFW_MODULE_ORDER (IPFW_MODEVENT_ORDER + 1) /* A little later. */ #define IPFW_VNET_ORDER (IPFW_MODEVENT_ORDER + 2) /* Later still. */ DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER); FEATURE(ipfw_ctl3, "ipfw new sockopt calls"); MODULE_VERSION(ipfw, 3); /* should declare some dependencies here */ /* * Starting up. Done in order after ipfwmod() has been called. * VNET_SYSINIT is also called for each existing vnet and each new vnet. */ SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER, ipfw_init, NULL); VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER, vnet_ipfw_init, NULL); /* * Closing up shop. These are done in REVERSE ORDER, but still * after ipfwmod() has been called. Not called on reboot. * VNET_SYSUNINIT is also called for each exiting vnet as it exits. * or when the module is unloaded. */ SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER, ipfw_destroy, NULL); VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER, vnet_ipfw_uninit, NULL); /* end of file */