Index: head/sys/netinet/ip_fastfwd.c =================================================================== --- head/sys/netinet/ip_fastfwd.c (revision 274358) +++ head/sys/netinet/ip_fastfwd.c (revision 274359) @@ -1,599 +1,598 @@ /*- * 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 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"); static struct sockaddr_in * ip_findroute(struct route *ro, struct in_addr dest, struct mbuf *m) { struct sockaddr_in *dst; struct rtentry *rt; /* * Find route to destination. */ bzero(ro, sizeof(*ro)); dst = (struct sockaddr_in *)&ro->ro_dst; dst->sin_family = AF_INET; dst->sin_len = sizeof(*dst); dst->sin_addr.s_addr = dest.s_addr; in_rtalloc_ign(ro, 0, M_GETFIB(m)); /* * Route there and interface still up? */ rt = ro->ro_rt; if (rt && (rt->rt_flags & RTF_UP) && (rt->rt_ifp->if_flags & IFF_UP) && (rt->rt_ifp->if_drv_flags & IFF_DRV_RUNNING)) { if (rt->rt_flags & RTF_GATEWAY) dst = (struct sockaddr_in *)rt->rt_gateway; } else { IPSTAT_INC(ips_noroute); IPSTAT_INC(ips_cantforward); if (rt) RTFREE(rt); icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); return NULL; } return dst; } /* * 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 route ro; struct sockaddr_in *dst = NULL; struct ifnet *ifp; struct in_addr odest, dest; uint16_t sum, ip_len, ip_off; 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); bzero(&ro, sizeof(ro)); /* * 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. */ if ((dst = ip_findroute(&ro, dest, m)) == NULL) return NULL; /* icmp unreach already sent */ ifp = ro.ro_rt->rt_ifp; /* * Immediately drop blackholed traffic, and directed broadcasts * for either the all-ones or all-zero subnet addresses on * locally attached networks. */ if ((ro.ro_rt->rt_flags & (RTF_BLACKHOLE|RTF_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 (ro.ro_rt) RTFREE(ro.ro_rt); 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; } RTFREE(ro.ro_rt); if ((dst = ip_findroute(&ro, dest, m)) == NULL) return NULL; /* icmp unreach already sent */ ifp = ro.ro_rt->rt_ifp; } passout: /* * Step 6: send off the packet */ ip_len = ntohs(ip->ip_len); ip_off = ntohs(ip->ip_off); /* * 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; } #ifndef ALTQ /* * Check if there is enough space in the interface queue */ if ((ifp->if_snd.ifq_len + ip_len / ifp->if_mtu + 1) >= ifp->if_snd.ifq_maxlen) { IPSTAT_INC(ips_odropped); - /* would send source quench here but that is depreciated */ goto drop; } #endif /* * 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 */ if (ro.ro_rt->rt_mtu) mtu = min(ro.ro_rt->rt_mtu, ifp->if_mtu); else mtu = ifp->if_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 = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst, &ro); } 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 = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst, &ro); 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 { counter_u64_add(ro.ro_rt->rt_pksent, 1); IPSTAT_INC(ips_forward); IPSTAT_INC(ips_fastforward); } consumed: RTFREE(ro.ro_rt); return NULL; drop: if (m) m_freem(m); if (ro.ro_rt) RTFREE(ro.ro_rt); return NULL; } Index: head/sys/netinet/ip_icmp.c =================================================================== --- head/sys/netinet/ip_icmp.c (revision 274358) +++ head/sys/netinet/ip_icmp.c (revision 274359) @@ -1,1009 +1,1004 @@ /*- * 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 "opt_ipsec.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 #ifdef INET #ifdef IPSEC #include #include #endif #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; 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[]; static int sysctl_net_icmp_drop_redir(SYSCTL_HANDLER_ARGS) { int error, new; int i; struct radix_node_head *rnh; new = V_drop_redirect; error = sysctl_handle_int(oidp, &new, 0, req); if (error == 0 && req->newptr) { new = (new != 0) ? 1 : 0; if (new == V_drop_redirect) return (0); for (i = 0; i < rt_numfibs; i++) { if ((rnh = rt_tables_get_rnh(i, AF_INET)) == NULL) continue; RADIX_NODE_HEAD_LOCK(rnh); in_setmatchfunc(rnh, new); RADIX_NODE_HEAD_UNLOCK(rnh); } V_drop_redirect = new; } return (error); } SYSCTL_PROC(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW, 0, 0, sysctl_net_icmp_drop_redir, "I", "Ignore ICMP redirects"); /* * 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; - goto deliver; - - case ICMP_SOURCEQUENCH: - if (code) - goto badcode; - code = PRC_QUENCH; 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); #ifdef IPSEC key_sa_routechange((struct sockaddr *)&icmpsrc); #endif 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 ip *ip = mtod(m, struct ip *); struct ifaddr *ifa; struct ifnet *ifp; struct in_ifaddr *ia; struct in_addr t; struct mbuf *opts = 0; 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(); 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(); goto match; } } IN_IFADDR_RUNLOCK(); /* * 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. */ ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m)); if (ia == NULL) { m_freem(m); ICMPSTAT_INC(icps_noroute); goto done; } t = IA_SIN(ia)->sin_addr; ifa_free(&ia->ia_ifa); 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 GMT 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: head/sys/netinet/ip_input.c =================================================================== --- head/sys/netinet/ip_input.c (revision 274358) +++ head/sys/netinet/ip_input.c (revision 274359) @@ -1,1898 +1,1873 @@ /*- * 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 #ifdef IPSEC #include #endif /* IPSEC */ #include #include #include #ifdef CTASSERT CTASSERT(sizeof(struct ip) == 20); #endif struct rwlock in_ifaddr_lock; RW_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"); -static VNET_DEFINE(int, ip_sendsourcequench); -#define V_ip_sendsourcequench VNET(ip_sendsourcequench) -SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_VNET | CTLFLAG_RW, - &VNET_NAME(ip_sendsourcequench), 0, - "Enable the transmission of source quench packets"); - VNET_DEFINE(int, ip_do_randomid); SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_do_randomid), 0, "Assign random ip_id values"); /* * 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 */ static VNET_DEFINE(uma_zone_t, ipq_zone); static VNET_DEFINE(TAILQ_HEAD(ipqhead, ipq), ipq[IPREASS_NHASH]); static struct mtx ipqlock; #define V_ipq_zone VNET(ipq_zone) #define V_ipq VNET(ipq) #define IPQ_LOCK() mtx_lock(&ipqlock) #define IPQ_UNLOCK() mtx_unlock(&ipqlock) #define IPQ_LOCK_INIT() mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF) #define IPQ_LOCK_ASSERT() mtx_assert(&ipqlock, MA_OWNED) static void maxnipq_update(void); static void ipq_zone_change(void *); static void ip_drain_locked(void); static VNET_DEFINE(int, maxnipq); /* Administrative limit on # reass queues. */ static VNET_DEFINE(int, nipq); /* Total # of reass queues */ #define V_maxnipq VNET(maxnipq) #define V_nipq VNET(nipq) SYSCTL_INT(_net_inet_ip, OID_AUTO, fragpackets, CTLFLAG_VNET | CTLFLAG_RD, &VNET_NAME(nipq), 0, "Current number of IPv4 fragment reassembly queue entries"); static VNET_DEFINE(int, maxfragsperpacket); #define V_maxfragsperpacket VNET(maxfragsperpacket) SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(maxfragsperpacket), 0, "Maximum number of IPv4 fragments allowed per packet"); #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 static void ip_freef(struct ipqhead *, struct ipq *); /* * 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; V_ip_id = time_second & 0xffff; TAILQ_INIT(&V_in_ifaddrhead); V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask); /* Initialize IP reassembly queue. */ for (i = 0; i < IPREASS_NHASH; i++) TAILQ_INIT(&V_ipq[i]); V_maxnipq = nmbclusters / 32; V_maxfragsperpacket = 16; V_ipq_zone = uma_zcreate("ipq", sizeof(struct ipq), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0); maxnipq_update(); /* 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; } EVENTHANDLER_REGISTER(nmbclusters_change, ipq_zone_change, NULL, EVENTHANDLER_PRI_ANY); /* Initialize various other remaining things. */ IPQ_LOCK_INIT(); 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); IPQ_LOCK(); ip_drain_locked(); IPQ_UNLOCK(); uma_zdestroy(V_ipq_zone); } #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 { #ifdef IPSEC if (ip_ipsec_fwd(m)) goto bad; #endif /* IPSEC */ 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)) 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); } /* * After maxnipq has been updated, propagate the change to UMA. The UMA zone * max has slightly different semantics than the sysctl, for historical * reasons. */ static void maxnipq_update(void) { /* * -1 for unlimited allocation. */ if (V_maxnipq < 0) uma_zone_set_max(V_ipq_zone, 0); /* * Positive number for specific bound. */ if (V_maxnipq > 0) uma_zone_set_max(V_ipq_zone, V_maxnipq); /* * Zero specifies no further fragment queue allocation -- set the * bound very low, but rely on implementation elsewhere to actually * prevent allocation and reclaim current queues. */ if (V_maxnipq == 0) uma_zone_set_max(V_ipq_zone, 1); } static void ipq_zone_change(void *tag) { if (V_maxnipq > 0 && V_maxnipq < (nmbclusters / 32)) { V_maxnipq = nmbclusters / 32; maxnipq_update(); } } static int sysctl_maxnipq(SYSCTL_HANDLER_ARGS) { int error, i; i = V_maxnipq; error = sysctl_handle_int(oidp, &i, 0, req); if (error || !req->newptr) return (error); /* * XXXRW: Might be a good idea to sanity check the argument and place * an extreme upper bound. */ if (i < -1) return (EINVAL); V_maxnipq = i; maxnipq_update(); return (0); } SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW, NULL, 0, sysctl_maxnipq, "I", "Maximum number of IPv4 fragment reassembly queue entries"); #define M_IP_FRAG M_PROTO9 /* * Take incoming datagram fragment and try to reassemble it into * whole datagram. If the argument is the first fragment or one * in between the function will return NULL and store the mbuf * in the fragment chain. If the argument is the last fragment * the packet will be reassembled and the pointer to the new * mbuf returned for further processing. Only m_tags attached * to the first packet/fragment are preserved. * The IP header is *NOT* adjusted out of iplen. */ struct mbuf * ip_reass(struct mbuf *m) { struct ip *ip; struct mbuf *p, *q, *nq, *t; struct ipq *fp = NULL; struct ipqhead *head; int i, hlen, next; u_int8_t ecn, ecn0; u_short hash; #ifdef RSS uint32_t rss_hash, rss_type; #endif /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */ if (V_maxnipq == 0 || V_maxfragsperpacket == 0) { IPSTAT_INC(ips_fragments); IPSTAT_INC(ips_fragdropped); m_freem(m); return (NULL); } ip = mtod(m, struct ip *); hlen = ip->ip_hl << 2; hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); head = &V_ipq[hash]; IPQ_LOCK(); /* * Look for queue of fragments * of this datagram. */ TAILQ_FOREACH(fp, head, ipq_list) if (ip->ip_id == fp->ipq_id && ip->ip_src.s_addr == fp->ipq_src.s_addr && ip->ip_dst.s_addr == fp->ipq_dst.s_addr && #ifdef MAC mac_ipq_match(m, fp) && #endif ip->ip_p == fp->ipq_p) goto found; fp = NULL; /* * Attempt to trim the number of allocated fragment queues if it * exceeds the administrative limit. */ if ((V_nipq > V_maxnipq) && (V_maxnipq > 0)) { /* * drop something from the tail of the current queue * before proceeding further */ struct ipq *q = TAILQ_LAST(head, ipqhead); if (q == NULL) { /* gak */ for (i = 0; i < IPREASS_NHASH; i++) { struct ipq *r = TAILQ_LAST(&V_ipq[i], ipqhead); if (r) { IPSTAT_ADD(ips_fragtimeout, r->ipq_nfrags); ip_freef(&V_ipq[i], r); break; } } } else { IPSTAT_ADD(ips_fragtimeout, q->ipq_nfrags); ip_freef(head, q); } } found: /* * Adjust ip_len to not reflect header, * convert offset of this to bytes. */ ip->ip_len = htons(ntohs(ip->ip_len) - hlen); if (ip->ip_off & htons(IP_MF)) { /* * Make sure that fragments have a data length * that's a non-zero multiple of 8 bytes. */ if (ip->ip_len == htons(0) || (ntohs(ip->ip_len) & 0x7) != 0) { IPSTAT_INC(ips_toosmall); /* XXX */ goto dropfrag; } m->m_flags |= M_IP_FRAG; } else m->m_flags &= ~M_IP_FRAG; ip->ip_off = htons(ntohs(ip->ip_off) << 3); /* * Attempt reassembly; if it succeeds, proceed. * ip_reass() will return a different mbuf. */ IPSTAT_INC(ips_fragments); m->m_pkthdr.PH_loc.ptr = ip; /* Previous ip_reass() started here. */ /* * Presence of header sizes in mbufs * would confuse code below. */ m->m_data += hlen; m->m_len -= hlen; /* * If first fragment to arrive, create a reassembly queue. */ if (fp == NULL) { fp = uma_zalloc(V_ipq_zone, M_NOWAIT); if (fp == NULL) goto dropfrag; #ifdef MAC if (mac_ipq_init(fp, M_NOWAIT) != 0) { uma_zfree(V_ipq_zone, fp); fp = NULL; goto dropfrag; } mac_ipq_create(m, fp); #endif TAILQ_INSERT_HEAD(head, fp, ipq_list); V_nipq++; fp->ipq_nfrags = 1; fp->ipq_ttl = IPFRAGTTL; fp->ipq_p = ip->ip_p; fp->ipq_id = ip->ip_id; fp->ipq_src = ip->ip_src; fp->ipq_dst = ip->ip_dst; fp->ipq_frags = m; m->m_nextpkt = NULL; goto done; } else { fp->ipq_nfrags++; #ifdef MAC mac_ipq_update(m, fp); #endif } #define GETIP(m) ((struct ip*)((m)->m_pkthdr.PH_loc.ptr)) /* * Handle ECN by comparing this segment with the first one; * if CE is set, do not lose CE. * drop if CE and not-ECT are mixed for the same packet. */ ecn = ip->ip_tos & IPTOS_ECN_MASK; ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK; if (ecn == IPTOS_ECN_CE) { if (ecn0 == IPTOS_ECN_NOTECT) goto dropfrag; if (ecn0 != IPTOS_ECN_CE) GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE; } if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT) goto dropfrag; /* * Find a segment which begins after this one does. */ for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) if (ntohs(GETIP(q)->ip_off) > ntohs(ip->ip_off)) break; /* * If there is a preceding segment, it may provide some of * our data already. If so, drop the data from the incoming * segment. If it provides all of our data, drop us, otherwise * stick new segment in the proper place. * * If some of the data is dropped from the preceding * segment, then it's checksum is invalidated. */ if (p) { i = ntohs(GETIP(p)->ip_off) + ntohs(GETIP(p)->ip_len) - ntohs(ip->ip_off); if (i > 0) { if (i >= ntohs(ip->ip_len)) goto dropfrag; m_adj(m, i); m->m_pkthdr.csum_flags = 0; ip->ip_off = htons(ntohs(ip->ip_off) + i); ip->ip_len = htons(ntohs(ip->ip_len) - i); } m->m_nextpkt = p->m_nextpkt; p->m_nextpkt = m; } else { m->m_nextpkt = fp->ipq_frags; fp->ipq_frags = m; } /* * While we overlap succeeding segments trim them or, * if they are completely covered, dequeue them. */ for (; q != NULL && ntohs(ip->ip_off) + ntohs(ip->ip_len) > ntohs(GETIP(q)->ip_off); q = nq) { i = (ntohs(ip->ip_off) + ntohs(ip->ip_len)) - ntohs(GETIP(q)->ip_off); if (i < ntohs(GETIP(q)->ip_len)) { GETIP(q)->ip_len = htons(ntohs(GETIP(q)->ip_len) - i); GETIP(q)->ip_off = htons(ntohs(GETIP(q)->ip_off) + i); m_adj(q, i); q->m_pkthdr.csum_flags = 0; break; } nq = q->m_nextpkt; m->m_nextpkt = nq; IPSTAT_INC(ips_fragdropped); fp->ipq_nfrags--; m_freem(q); } /* * Check for complete reassembly and perform frag per packet * limiting. * * Frag limiting is performed here so that the nth frag has * a chance to complete the packet before we drop the packet. * As a result, n+1 frags are actually allowed per packet, but * only n will ever be stored. (n = maxfragsperpacket.) * */ next = 0; for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { if (ntohs(GETIP(q)->ip_off) != next) { if (fp->ipq_nfrags > V_maxfragsperpacket) { IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); ip_freef(head, fp); } goto done; } next += ntohs(GETIP(q)->ip_len); } /* Make sure the last packet didn't have the IP_MF flag */ if (p->m_flags & M_IP_FRAG) { if (fp->ipq_nfrags > V_maxfragsperpacket) { IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); ip_freef(head, fp); } goto done; } /* * Reassembly is complete. Make sure the packet is a sane size. */ q = fp->ipq_frags; ip = GETIP(q); if (next + (ip->ip_hl << 2) > IP_MAXPACKET) { IPSTAT_INC(ips_toolong); IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); ip_freef(head, fp); goto done; } /* * Concatenate fragments. */ m = q; t = m->m_next; m->m_next = NULL; m_cat(m, t); nq = q->m_nextpkt; q->m_nextpkt = NULL; for (q = nq; q != NULL; q = nq) { nq = q->m_nextpkt; q->m_nextpkt = NULL; m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; m_cat(m, q); } /* * In order to do checksumming faster we do 'end-around carry' here * (and not in for{} loop), though it implies we are not going to * reassemble more than 64k fragments. */ while (m->m_pkthdr.csum_data & 0xffff0000) m->m_pkthdr.csum_data = (m->m_pkthdr.csum_data & 0xffff) + (m->m_pkthdr.csum_data >> 16); #ifdef MAC mac_ipq_reassemble(fp, m); mac_ipq_destroy(fp); #endif /* * Create header for new ip packet by modifying header of first * packet; dequeue and discard fragment reassembly header. * Make header visible. */ ip->ip_len = htons((ip->ip_hl << 2) + next); ip->ip_src = fp->ipq_src; ip->ip_dst = fp->ipq_dst; TAILQ_REMOVE(head, fp, ipq_list); V_nipq--; uma_zfree(V_ipq_zone, fp); m->m_len += (ip->ip_hl << 2); m->m_data -= (ip->ip_hl << 2); /* some debugging cruft by sklower, below, will go away soon */ if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */ m_fixhdr(m); IPSTAT_INC(ips_reassembled); IPQ_UNLOCK(); #ifdef RSS /* * Query the RSS layer for the flowid / flowtype for the * mbuf payload. * * For now, just assume we have to calculate a new one. * Later on we should check to see if the assigned flowid matches * what RSS wants for the given IP protocol and if so, just keep it. * * We then queue into the relevant netisr so it can be dispatched * to the correct CPU. * * Note - this may return 1, which means the flowid in the mbuf * is correct for the configured RSS hash types and can be used. */ if (rss_mbuf_software_hash_v4(m, 0, &rss_hash, &rss_type) == 0) { m->m_pkthdr.flowid = rss_hash; M_HASHTYPE_SET(m, rss_type); m->m_flags |= M_FLOWID; } /* * Queue/dispatch for reprocessing. * * Note: this is much slower than just handling the frame in the * current receive context. It's likely worth investigating * why this is. */ netisr_dispatch(NETISR_IP_DIRECT, m); return (NULL); #endif /* Handle in-line */ return (m); dropfrag: IPSTAT_INC(ips_fragdropped); if (fp != NULL) fp->ipq_nfrags--; m_freem(m); done: IPQ_UNLOCK(); return (NULL); #undef GETIP } /* * Free a fragment reassembly header and all * associated datagrams. */ static void ip_freef(struct ipqhead *fhp, struct ipq *fp) { struct mbuf *q; IPQ_LOCK_ASSERT(); while (fp->ipq_frags) { q = fp->ipq_frags; fp->ipq_frags = q->m_nextpkt; m_freem(q); } TAILQ_REMOVE(fhp, fp, ipq_list); uma_zfree(V_ipq_zone, fp); V_nipq--; } /* * IP timer processing; * if a timer expires on a reassembly * queue, discard it. */ void ip_slowtimo(void) { VNET_ITERATOR_DECL(vnet_iter); struct ipq *fp; int i; VNET_LIST_RLOCK_NOSLEEP(); IPQ_LOCK(); VNET_FOREACH(vnet_iter) { CURVNET_SET(vnet_iter); for (i = 0; i < IPREASS_NHASH; i++) { for(fp = TAILQ_FIRST(&V_ipq[i]); fp;) { struct ipq *fpp; fpp = fp; fp = TAILQ_NEXT(fp, ipq_list); if(--fpp->ipq_ttl == 0) { IPSTAT_ADD(ips_fragtimeout, fpp->ipq_nfrags); ip_freef(&V_ipq[i], fpp); } } } /* * If we are over the maximum number of fragments * (due to the limit being lowered), drain off * enough to get down to the new limit. */ if (V_maxnipq >= 0 && V_nipq > V_maxnipq) { for (i = 0; i < IPREASS_NHASH; i++) { while (V_nipq > V_maxnipq && !TAILQ_EMPTY(&V_ipq[i])) { IPSTAT_ADD(ips_fragdropped, TAILQ_FIRST(&V_ipq[i])->ipq_nfrags); ip_freef(&V_ipq[i], TAILQ_FIRST(&V_ipq[i])); } } } CURVNET_RESTORE(); } IPQ_UNLOCK(); VNET_LIST_RUNLOCK_NOSLEEP(); } /* * Drain off all datagram fragments. */ static void ip_drain_locked(void) { int i; IPQ_LOCK_ASSERT(); for (i = 0; i < IPREASS_NHASH; i++) { while(!TAILQ_EMPTY(&V_ipq[i])) { IPSTAT_ADD(ips_fragdropped, TAILQ_FIRST(&V_ipq[i])->ipq_nfrags); ip_freef(&V_ipq[i], TAILQ_FIRST(&V_ipq[i])); } } } void ip_drain(void) { VNET_ITERATOR_DECL(vnet_iter); VNET_LIST_RLOCK_NOSLEEP(); IPQ_LOCK(); VNET_FOREACH(vnet_iter) { CURVNET_SET(vnet_iter); ip_drain_locked(); CURVNET_RESTORE(); } IPQ_UNLOCK(); VNET_LIST_RUNLOCK_NOSLEEP(); in_rtqdrain(); } /* * 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); } /* * Given address of next destination (final or next hop), return (referenced) * internet address info of interface to be used to get there. */ struct in_ifaddr * ip_rtaddr(struct in_addr dst, u_int fibnum) { struct route sro; struct sockaddr_in *sin; struct in_ifaddr *ia; bzero(&sro, sizeof(sro)); sin = (struct sockaddr_in *)&sro.ro_dst; sin->sin_family = AF_INET; sin->sin_len = sizeof(*sin); sin->sin_addr = dst; in_rtalloc_ign(&sro, 0, fibnum); if (sro.ro_rt == NULL) return (NULL); ia = ifatoia(sro.ro_rt->rt_ifa); ifa_ref(&ia->ia_ifa); RTFREE(sro.ro_rt); return (ia); } 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 in_ifaddr *ia; struct mbuf *mcopy; struct in_addr dest; struct route ro; 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 IPSTEALTH if (!V_ipstealth) { #endif if (ip->ip_ttl <= IPTTLDEC) { icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0); return; } #ifdef IPSTEALTH } #endif ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m)); #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 && ia == NULL) { icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 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 && ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) { struct sockaddr_in *sin; struct rtentry *rt; bzero(&ro, sizeof(ro)); sin = (struct sockaddr_in *)&ro.ro_dst; sin->sin_family = AF_INET; sin->sin_len = sizeof(*sin); sin->sin_addr = ip->ip_dst; in_rtalloc_ign(&ro, 0, M_GETFIB(m)); rt = ro.ro_rt; if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && satosin(rt_key(rt))->sin_addr.s_addr != 0) { #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) u_long src = ntohl(ip->ip_src.s_addr); if (RTA(rt) && (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { if (rt->rt_flags & RTF_GATEWAY) dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr; else dest.s_addr = ip->ip_dst.s_addr; /* Router requirements says to only send host redirects */ type = ICMP_REDIRECT; code = ICMP_REDIRECT_HOST; } } if (rt) RTFREE(rt); } /* * Try to cache the route MTU from ip_output so we can consider it for * the ICMP_UNREACH_NEEDFRAG "Next-Hop MTU" field described in RFC1191. */ bzero(&ro, sizeof(ro)); error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL); if (error == EMSGSIZE && ro.ro_rt) mtu = ro.ro_rt->rt_mtu; RO_RTFREE(&ro); if (error) IPSTAT_INC(ips_cantforward); else { IPSTAT_INC(ips_forward); if (type) IPSTAT_INC(ips_redirectsent); else { if (mcopy) m_freem(mcopy); if (ia != NULL) ifa_free(&ia->ia_ifa); return; } } if (mcopy == NULL) { if (ia != NULL) ifa_free(&ia->ia_ifa); 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 = ip_ipsec_mtu(mcopy, mtu); #endif /* IPSEC */ /* * If the MTU was set before make sure we are below the * interface MTU. * If the MTU wasn't set before use the interface mtu or * fall back to the next smaller mtu step compared to the * current packet size. */ if (mtu != 0) { if (ia != NULL) mtu = min(mtu, ia->ia_ifp->if_mtu); } else { if (ia != NULL) mtu = ia->ia_ifp->if_mtu; else mtu = ip_next_mtu(ntohs(ip->ip_len), 0); } IPSTAT_INC(ips_cantfrag); break; case ENOBUFS: - /* - * A router should not generate ICMP_SOURCEQUENCH as - * required in RFC1812 Requirements for IP Version 4 Routers. - * Source quench could be a big problem under DoS attacks, - * or if the underlying interface is rate-limited. - * Those who need source quench packets may re-enable them - * via the net.inet.ip.sendsourcequench sysctl. - */ - if (V_ip_sendsourcequench == 0) { - m_freem(mcopy); - if (ia != NULL) - ifa_free(&ia->ia_ifa); - return; - } else { - type = ICMP_SOURCEQUENCH; - code = 0; - } - break; - case EACCES: /* ipfw denied packet */ m_freem(mcopy); if (ia != NULL) ifa_free(&ia->ia_ifa); return; } if (ia != NULL) ifa_free(&ia->ia_ifa); 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: head/sys/netinet/tcp_subr.c =================================================================== --- head/sys/netinet/tcp_subr.c (revision 274358) +++ head/sys/netinet/tcp_subr.c (revision 274359) @@ -1,2453 +1,2445 @@ /*- * 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 #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); /* * 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; /* 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, CALLOUT_MPSAFE); callout_init(&tp->t_timers->tt_persist, CALLOUT_MPSAFE); callout_init(&tp->t_timers->tt_keep, CALLOUT_MPSAFE); callout_init(&tp->t_timers->tt_2msl, CALLOUT_MPSAFE); callout_init(&tp->t_timers->tt_delack, CALLOUT_MPSAFE); 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); tp->t_inpcb = inp; /* XXX */ /* * 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 */ INP_WLOCK_ASSERT(inp); /* * Make sure that all of our timers are stopped before we delete the * PCB. * * XXXRW: Really, we would like to use callout_drain() here in order * to avoid races experienced in tcp_timer.c where a timer is already * executing at this point. However, we can't, both because we're * running in a context where we can't sleep, and also because we * hold locks required by the timers. What we instead need to do is * test to see if callout_drain() is required, and if so, defer some * portion of the remainder of tcp_discardcb() to an asynchronous * context that can callout_drain() and then continue. Some care * will be required to ensure that no further processing takes place * on the tcpcb, even though it hasn't been freed (a flag?). */ callout_stop(&tp->t_timers->tt_rexmt); callout_stop(&tp->t_timers->tt_persist); callout_stop(&tp->t_timers->tt_keep); callout_stop(&tp->t_timers->tt_2msl); callout_stop(&tp->t_timers->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; tp->t_inpcb = NULL; uma_zfree(V_tcpcb_zone, tp); } /* * 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; /* - * Source quench is depreciated. - */ - else if (cmd == PRC_QUENCH) - 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; - /* Source quench is depreciated. */ - else if (cmd == PRC_QUENCH) 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 route sro; struct sockaddr_in *dst; struct ifnet *ifp; u_long maxmtu = 0; KASSERT(inc != NULL, ("tcp_maxmtu with NULL in_conninfo pointer")); bzero(&sro, sizeof(sro)); if (inc->inc_faddr.s_addr != INADDR_ANY) { dst = (struct sockaddr_in *)&sro.ro_dst; dst->sin_family = AF_INET; dst->sin_len = sizeof(*dst); dst->sin_addr = inc->inc_faddr; in_rtalloc_ign(&sro, 0, inc->inc_fibnum); } if (sro.ro_rt != NULL) { ifp = sro.ro_rt->rt_ifp; if (sro.ro_rt->rt_mtu == 0) maxmtu = ifp->if_mtu; else maxmtu = min(sro.ro_rt->rt_mtu, ifp->if_mtu); /* Report additional interface capabilities. */ if (cap != NULL) { 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; } } RTFREE(sro.ro_rt); } 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: 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: head/sys/sys/protosw.h =================================================================== --- head/sys/sys/protosw.h (revision 274358) +++ head/sys/sys/protosw.h (revision 274359) @@ -1,347 +1,347 @@ /*- * Copyright (c) 1982, 1986, 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. * * @(#)protosw.h 8.1 (Berkeley) 6/2/93 * $FreeBSD$ */ #ifndef _SYS_PROTOSW_H_ #define _SYS_PROTOSW_H_ /* Forward declare these structures referenced from prototypes below. */ struct mbuf; struct thread; struct sockaddr; struct socket; struct sockopt; /*#ifdef _KERNEL*/ /* * Protocol switch table. * * Each protocol has a handle initializing one of these structures, * which is used for protocol-protocol and system-protocol communication. * * A protocol is called through the pr_init entry before any other. * Thereafter it is called every 200ms through the pr_fasttimo entry and * every 500ms through the pr_slowtimo for timer based actions. * The system will call the pr_drain entry if it is low on space and * this should throw away any non-critical data. * * Protocols pass data between themselves as chains of mbufs using * the pr_input and pr_output hooks. Pr_input passes data up (towards * the users) and pr_output passes it down (towards the interfaces); control * information passes up and down on pr_ctlinput and pr_ctloutput. * The protocol is responsible for the space occupied by any the * arguments to these entries and must dispose it. * * In retrospect, it would be a lot nicer to use an interface * similar to the vnode VOP interface. */ /* USE THESE FOR YOUR PROTOTYPES ! */ typedef int pr_input_t (struct mbuf **, int*, int); typedef int pr_output_t (struct mbuf *, struct socket *, ...); typedef void pr_ctlinput_t (int, struct sockaddr *, void *); typedef int pr_ctloutput_t (struct socket *, struct sockopt *); typedef void pr_init_t (void); typedef void pr_destroy_t (void); typedef void pr_fasttimo_t (void); typedef void pr_slowtimo_t (void); typedef void pr_drain_t (void); struct protosw { short pr_type; /* socket type used for */ struct domain *pr_domain; /* domain protocol a member of */ short pr_protocol; /* protocol number */ short pr_flags; /* see below */ /* protocol-protocol hooks */ pr_input_t *pr_input; /* input to protocol (from below) */ pr_output_t *pr_output; /* output to protocol (from above) */ pr_ctlinput_t *pr_ctlinput; /* control input (from below) */ pr_ctloutput_t *pr_ctloutput; /* control output (from above) */ /* utility hooks */ pr_init_t *pr_init; pr_destroy_t *pr_destroy; pr_fasttimo_t *pr_fasttimo; /* fast timeout (200ms) */ pr_slowtimo_t *pr_slowtimo; /* slow timeout (500ms) */ pr_drain_t *pr_drain; /* flush any excess space possible */ struct pr_usrreqs *pr_usrreqs; /* user-protocol hook */ }; /*#endif*/ #define PR_SLOWHZ 2 /* 2 slow timeouts per second */ #define PR_FASTHZ 5 /* 5 fast timeouts per second */ /* * This number should be defined again within each protocol family to avoid * confusion. */ #define PROTO_SPACER 32767 /* spacer for loadable protocols */ /* * Values for pr_flags. * PR_ADDR requires PR_ATOMIC; * PR_ADDR and PR_CONNREQUIRED are mutually exclusive. * PR_IMPLOPCL means that the protocol allows sendto without prior connect, * and the protocol understands the MSG_EOF flag. The first property is * is only relevant if PR_CONNREQUIRED is set (otherwise sendto is allowed * anyhow). */ #define PR_ATOMIC 0x01 /* exchange atomic messages only */ #define PR_ADDR 0x02 /* addresses given with messages */ #define PR_CONNREQUIRED 0x04 /* connection required by protocol */ #define PR_WANTRCVD 0x08 /* want PRU_RCVD calls */ #define PR_RIGHTS 0x10 /* passes capabilities */ #define PR_IMPLOPCL 0x20 /* implied open/close */ #define PR_LASTHDR 0x40 /* enforce ipsec policy; last header */ /* * In earlier BSD network stacks, a single pr_usrreq() function pointer was * invoked with an operation number indicating what operation was desired. * We now provide individual function pointers which protocols can implement, * which offers a number of benefits (such as type checking for arguments). * These older constants are still present in order to support TCP debugging. */ #define PRU_ATTACH 0 /* attach protocol to up */ #define PRU_DETACH 1 /* detach protocol from up */ #define PRU_BIND 2 /* bind socket to address */ #define PRU_LISTEN 3 /* listen for connection */ #define PRU_CONNECT 4 /* establish connection to peer */ #define PRU_ACCEPT 5 /* accept connection from peer */ #define PRU_DISCONNECT 6 /* disconnect from peer */ #define PRU_SHUTDOWN 7 /* won't send any more data */ #define PRU_RCVD 8 /* have taken data; more room now */ #define PRU_SEND 9 /* send this data */ #define PRU_ABORT 10 /* abort (fast DISCONNECT, DETATCH) */ #define PRU_CONTROL 11 /* control operations on protocol */ #define PRU_SENSE 12 /* return status into m */ #define PRU_RCVOOB 13 /* retrieve out of band data */ #define PRU_SENDOOB 14 /* send out of band data */ #define PRU_SOCKADDR 15 /* fetch socket's address */ #define PRU_PEERADDR 16 /* fetch peer's address */ #define PRU_CONNECT2 17 /* connect two sockets */ /* begin for protocols internal use */ #define PRU_FASTTIMO 18 /* 200ms timeout */ #define PRU_SLOWTIMO 19 /* 500ms timeout */ #define PRU_PROTORCV 20 /* receive from below */ #define PRU_PROTOSEND 21 /* send to below */ /* end for protocol's internal use */ #define PRU_SEND_EOF 22 /* send and close */ #define PRU_SOSETLABEL 23 /* MAC label change */ #define PRU_CLOSE 24 /* socket close */ #define PRU_FLUSH 25 /* flush the socket */ #define PRU_NREQ 25 #ifdef PRUREQUESTS const char *prurequests[] = { "ATTACH", "DETACH", "BIND", "LISTEN", "CONNECT", "ACCEPT", "DISCONNECT", "SHUTDOWN", "RCVD", "SEND", "ABORT", "CONTROL", "SENSE", "RCVOOB", "SENDOOB", "SOCKADDR", "PEERADDR", "CONNECT2", "FASTTIMO", "SLOWTIMO", "PROTORCV", "PROTOSEND", "SEND_EOF", "SOSETLABEL", "CLOSE", "FLUSH", }; #endif #ifdef _KERNEL /* users shouldn't see this decl */ struct ifnet; struct stat; struct ucred; struct uio; /* * If the ordering here looks odd, that's because it's alphabetical. These * should eventually be merged back into struct protosw. * * Some fields initialized to defaults if they are NULL. * See uipc_domain.c:net_init_domain() */ struct pr_usrreqs { void (*pru_abort)(struct socket *so); int (*pru_accept)(struct socket *so, struct sockaddr **nam); int (*pru_attach)(struct socket *so, int proto, struct thread *td); int (*pru_bind)(struct socket *so, struct sockaddr *nam, struct thread *td); int (*pru_connect)(struct socket *so, struct sockaddr *nam, struct thread *td); int (*pru_connect2)(struct socket *so1, struct socket *so2); int (*pru_control)(struct socket *so, u_long cmd, caddr_t data, struct ifnet *ifp, struct thread *td); void (*pru_detach)(struct socket *so); int (*pru_disconnect)(struct socket *so); int (*pru_listen)(struct socket *so, int backlog, struct thread *td); int (*pru_peeraddr)(struct socket *so, struct sockaddr **nam); int (*pru_rcvd)(struct socket *so, int flags); int (*pru_rcvoob)(struct socket *so, struct mbuf *m, int flags); int (*pru_send)(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr, struct mbuf *control, struct thread *td); #define PRUS_OOB 0x1 #define PRUS_EOF 0x2 #define PRUS_MORETOCOME 0x4 int (*pru_sense)(struct socket *so, struct stat *sb); int (*pru_shutdown)(struct socket *so); int (*pru_flush)(struct socket *so, int direction); int (*pru_sockaddr)(struct socket *so, struct sockaddr **nam); int (*pru_sosend)(struct socket *so, struct sockaddr *addr, struct uio *uio, struct mbuf *top, struct mbuf *control, int flags, struct thread *td); int (*pru_soreceive)(struct socket *so, struct sockaddr **paddr, struct uio *uio, struct mbuf **mp0, struct mbuf **controlp, int *flagsp); int (*pru_sopoll)(struct socket *so, int events, struct ucred *cred, struct thread *td); void (*pru_sosetlabel)(struct socket *so); void (*pru_close)(struct socket *so); int (*pru_bindat)(int fd, struct socket *so, struct sockaddr *nam, struct thread *td); int (*pru_connectat)(int fd, struct socket *so, struct sockaddr *nam, struct thread *td); }; /* * All nonvoid pru_*() functions below return EOPNOTSUPP. */ int pru_accept_notsupp(struct socket *so, struct sockaddr **nam); int pru_attach_notsupp(struct socket *so, int proto, struct thread *td); int pru_bind_notsupp(struct socket *so, struct sockaddr *nam, struct thread *td); int pru_bindat_notsupp(int fd, struct socket *so, struct sockaddr *nam, struct thread *td); int pru_connect_notsupp(struct socket *so, struct sockaddr *nam, struct thread *td); int pru_connectat_notsupp(int fd, struct socket *so, struct sockaddr *nam, struct thread *td); int pru_connect2_notsupp(struct socket *so1, struct socket *so2); int pru_control_notsupp(struct socket *so, u_long cmd, caddr_t data, struct ifnet *ifp, struct thread *td); int pru_disconnect_notsupp(struct socket *so); int pru_listen_notsupp(struct socket *so, int backlog, struct thread *td); int pru_peeraddr_notsupp(struct socket *so, struct sockaddr **nam); int pru_rcvd_notsupp(struct socket *so, int flags); int pru_rcvoob_notsupp(struct socket *so, struct mbuf *m, int flags); int pru_send_notsupp(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr, struct mbuf *control, struct thread *td); int pru_sense_null(struct socket *so, struct stat *sb); int pru_shutdown_notsupp(struct socket *so); int pru_sockaddr_notsupp(struct socket *so, struct sockaddr **nam); int pru_sosend_notsupp(struct socket *so, struct sockaddr *addr, struct uio *uio, struct mbuf *top, struct mbuf *control, int flags, struct thread *td); int pru_soreceive_notsupp(struct socket *so, struct sockaddr **paddr, struct uio *uio, struct mbuf **mp0, struct mbuf **controlp, int *flagsp); int pru_sopoll_notsupp(struct socket *so, int events, struct ucred *cred, struct thread *td); #endif /* _KERNEL */ /* * The arguments to the ctlinput routine are * (*protosw[].pr_ctlinput)(cmd, sa, arg); * where cmd is one of the commands below, sa is a pointer to a sockaddr, * and arg is a `void *' argument used within a protocol family. */ #define PRC_IFDOWN 0 /* interface transition */ #define PRC_ROUTEDEAD 1 /* select new route if possible ??? */ #define PRC_IFUP 2 /* interface has come back up */ -#define PRC_QUENCH2 3 /* DEC congestion bit says slow down */ -#define PRC_QUENCH 4 /* some one said to slow down */ +/* was PRC_QUENCH2 3 DEC congestion bit says slow down */ +/* was PRC_QUENCH 4 Deprecated by RFC 6633 */ #define PRC_MSGSIZE 5 /* message size forced drop */ #define PRC_HOSTDEAD 6 /* host appears to be down */ #define PRC_HOSTUNREACH 7 /* deprecated (use PRC_UNREACH_HOST) */ #define PRC_UNREACH_NET 8 /* no route to network */ #define PRC_UNREACH_HOST 9 /* no route to host */ #define PRC_UNREACH_PROTOCOL 10 /* dst says bad protocol */ #define PRC_UNREACH_PORT 11 /* bad port # */ /* was PRC_UNREACH_NEEDFRAG 12 (use PRC_MSGSIZE) */ #define PRC_UNREACH_SRCFAIL 13 /* source route failed */ #define PRC_REDIRECT_NET 14 /* net routing redirect */ #define PRC_REDIRECT_HOST 15 /* host routing redirect */ #define PRC_REDIRECT_TOSNET 16 /* redirect for type of service & net */ #define PRC_REDIRECT_TOSHOST 17 /* redirect for tos & host */ #define PRC_TIMXCEED_INTRANS 18 /* packet lifetime expired in transit */ #define PRC_TIMXCEED_REASS 19 /* lifetime expired on reass q */ #define PRC_PARAMPROB 20 /* header incorrect */ #define PRC_UNREACH_ADMIN_PROHIB 21 /* packet administrativly prohibited */ #define PRC_NCMDS 22 #define PRC_IS_REDIRECT(cmd) \ ((cmd) >= PRC_REDIRECT_NET && (cmd) <= PRC_REDIRECT_TOSHOST) #ifdef PRCREQUESTS char *prcrequests[] = { "IFDOWN", "ROUTEDEAD", "IFUP", "DEC-BIT-QUENCH2", "QUENCH", "MSGSIZE", "HOSTDEAD", "#7", "NET-UNREACH", "HOST-UNREACH", "PROTO-UNREACH", "PORT-UNREACH", "#12", "SRCFAIL-UNREACH", "NET-REDIRECT", "HOST-REDIRECT", "TOSNET-REDIRECT", "TOSHOST-REDIRECT", "TX-INTRANS", "TX-REASS", "PARAMPROB", "ADMIN-UNREACH" }; #endif /* * The arguments to ctloutput are: * (*protosw[].pr_ctloutput)(req, so, level, optname, optval, p); * req is one of the actions listed below, so is a (struct socket *), * level is an indication of which protocol layer the option is intended. * optname is a protocol dependent socket option request, * optval is a pointer to a mbuf-chain pointer, for value-return results. * The protocol is responsible for disposal of the mbuf chain *optval * if supplied, * the caller is responsible for any space held by *optval, when returned. * A non-zero return from ctloutput gives an * UNIX error number which should be passed to higher level software. */ #define PRCO_GETOPT 0 #define PRCO_SETOPT 1 #define PRCO_NCMDS 2 #ifdef PRCOREQUESTS char *prcorequests[] = { "GETOPT", "SETOPT", }; #endif #ifdef _KERNEL void pfctlinput(int, struct sockaddr *); void pfctlinput2(int, struct sockaddr *, void *); struct domain *pffinddomain(int family); struct protosw *pffindproto(int family, int protocol, int type); struct protosw *pffindtype(int family, int type); int pf_proto_register(int family, struct protosw *npr); int pf_proto_unregister(int family, int protocol, int type); #endif #endif Index: head/tools/tools/sysdoc/tunables.mdoc =================================================================== --- head/tools/tools/sysdoc/tunables.mdoc (revision 274358) +++ head/tools/tools/sysdoc/tunables.mdoc (revision 274359) @@ -1,2396 +1,2387 @@ # $FreeBSD$ --- debug.disablecwd bool Determines whether or not the .Xr getwcd 3 system call should be allowed. --- debug.disablefullpath bool Determines whether or not the .Fn vn_fullpath function may be used. --- debug.dobkgrdwrite bool Determines if background writes should be performed. --- debug.hashstat.nchash struct Displays nchash chain lengths. This is a read-only variable. --- debug.hashstat.rawnchash --- debug.ieee80211 bool This .Nm allows you to enable or disable debugging for 802.11 devices. --- debug.kdb.available variable Used to retrieve a list of currently available debugger backends. --- debug.kdb.current variable Allows for the selection of the debugger backend which is used to handle debugger requests. --- debug.kdb.enter variable When written to, the system should break to the debugger. --- debug.malloc.failure_count bool Number of times a coerced malloc failure has occurred as a result of .Va debug.malloc.failure_rate . Useful for tracking what might have happened and whether failures are being generated. --- debug.malloc.failure_rate bool Debugging feature causing .Dv M_NOWAIT allocations to fail at a specified rate. How often to generate a failure: if set to 0 (default), this feature is disabled. In other words if set to 10 (one in ten .Xr malloc 3 calls will fail). --- debug.rman_debug bool This .Nm allows you to enable or disable debugging for .Xr rman 9 , the .Fx resource manager. --- debug.sizeof.bio --- debug.sizeof.buf --- debug.sizeof.cdev --- debug.sizeof.devstat --- debug.sizeof.kinfo_proc --- debug.sizeof.proc --- debug.sizeof.vnode --- debug.vnlru_nowhere --- hw.acpi.cpu.current_speed bool Display the current CPU speed. This is adjustable, but doing so is not recommended. --- hw.acpi.cpu.max_speed int Allows you to change the stepping for processor speed on machines which support .Xr acpi 4 . --- hw.acpi.disable_on_poweroff bool Some systems using .Xr acpi 4 have problems powering off when shutting down with .Xr acpi 4 enabled. This .Nm disables .Xr acpi 4 when rebooting and shutting down. --- hw.acpi.s4bios bool This .Nm determines whether or not the S4BIOS sleep implementation should be used. --- hw.acpi.sleep_delay int Set the sleep delay for .Xr acpi 4 . --- hw.acpi.supported_sleep_state bool List supported .Tn ACPI sleep states --- hw.acpi.thermal.min_runtime --- hw.acpi.thermal.polling_rate int The interval in seconds that should be used to check the current system temperature. --- hw.acpi.thermal.tz0.temperature str Displays the current temperature. This is a read-only variable. --- hw.acpi.thermal.tz0.thermal_flags --- hw.acpi.verbose bool Determines whether or not .Xr acpi 4 should be verbose. --- hw.ata.ata_dma bool Allows the enabling and disabling of DMA for ATA devices. --- hw.ata.atapi_dma bool Allows the enabling and disabling of DMA for atapi devices, such as CD-ROM drives. --- hw.ata.tags bool An experimental feature for IDE hard drives which allows write caching to be turned on. Please read the .Xr tuning 7 manual page carefully before using this. --- hw.ata.wc bool Determines whether or not IDE write caching should be turned on or off. See .Xr tuning 7 for more information. --- hw.bus.devices --- hw.bus.info int This is an internally used function that returns the kernel bus interface version. --- hw.bus.rman --- hw.busdmafree_bpages --- hw.busdma.reserved_bpages --- hw.busdma.active_bpages --- hw.busdma.total_bpages --- hw.busdma.total_bounced --- hw.busdma.total_deferred --- hw.byteorder int Returns the system byte order. This is a read-only variable. --- hw.cardbus.cis_debug --- hw.cardbus.debug --- hw.cbb.debug --- hw.cbb.start_16_io --- hw.cbb.start_32_io --- hw.cbb.start_memory --- hw.floatingpoint bool Reports true if the machine has a floating point processor. This is a read-only variable. --- hw.fxp0.bundle_max int Controls the receive interrupt microcode bundle size limit for the .Xr fxp 4 device. --- hw.fxp0.int_delay int Controls the receive interrupt microcode bundling delay for the .Xr fxp 4 device. --- hw.fxp_noflow bool Disables flow control support on .Xr fxp 4 cards. When flow control is enabled, and if the operating system does not acknowledge the packet buffer filling, the card will begin to generate Ethernet quench packets, but appears to get into a feedback loop of some sort, hosing local switches. This is a workaround for this issue. --- hw.fxp_rnr int Set the amount of times that a no-resource condition may occur before the .Xr fxp 4 device may restart. --- hw.instruction_sse bool Returns true if SSE support is enabled in the kernel. This is a read-only variable. --- hw.intrcnt bool Displays a list of interrupt counters. This is a read-only variable. --- hw.intrnames str Displays a list of zero-terminated interrupt names. This is a read-only variable. --- hw.kbd.keymap_restrict_change bool This sysctl acts as a sort of secure-level, allowing control of the console keymap. Giving this a value of 1 means that only the root user can change restricted keys (like boot, panic...). A value of 2 means that only root can change restricted keys and regular keys. Regular users still can change accents and function keys. A value of 3 means only root can change restricted, regular and accent keys, while a value of 4 means that no changes to the keymap are allowed by anyone other than the root user. --- hw.machine str Displays the machine class. This is a read-only variable. --- hw.machine_arch str Displays the current architecture. This is a read-only variable. --- hw.model str Displays the model information of the current running hardware. This is a read-only variable. --- hw.ncpu bool Report the number of CPU's in the system. This is a read-only variable. --- hw.pagesize int Displays the current .Xr pagesize 1 . This is a read-only variable. --- hw.pccard.cis_debug int Allows debugging to be turned on or off for CIS. --- hw.pccard.debug bool Determines whether or not to use debugging for the PC Card bus driver. --- hw.pci.allow_unsupported_io_range bool Some machines do not detect their CardBus slots correctly because they use unsupported I/O ranges. This .Nm allows FreeBSD to use those ranges. --- hw.pci.enable_io_modes --- hw.snd.pcm0.ac97rate --- hw.snd.verbose int Control the level of verbosity for the .Pa /dev/sndstat device. See the .Xr pcm 4 man page for more information on debug levels. --- hw.snd.report_soft_formats bool Controls the internal format conversion if it is available transparently to the application software. See .Xr pcm 4 for more information. --- hw.syscons.bell bool Allows you to control whether or not to use the 'bell' while using the console. This is turned on by default. --- hw.syscons.saver.keybonly bool This variable tells the system that the screen saver may only wake up if the keyboard is used. This means that log messages that are pushed to the console will not cause the screen saver to stop, and display the log message will not display. This can be disabled to mimic the behavior of older syscons. --- hw.syscons.sc_no_suspend_vtswitch bool Disables switching between virtual terminals during suspend or resume. See .Xr syscons 4 for more information. --- hw.wi.debug bool Controls the level of debugging for .Xr wi 4 devices. --- hw.wi.txerate int This value allows controls the maximum amount of error messages per second. Giving this .Nm a value of 0 (zero) disables error messages completely. --- kern.acct_chkfreq int Specifies the frequency (in minutes) with which free disk space should be checked. This is used in conjunction with .Va kern.acct_resume and .Va kern.acct_suspend. --- kern.acct_resume int The percentage of free disk space above which process accounting will resume. --- kern.acct_suspend int The percentage of free disk space below which process accounting stops. --- kern.argmax bool The maximum number of bytes that can be used in an argument to .Xr execve 2 . This is basically the maximum number of characters which can be used in a single command line. On some rare occasions, this value needs altering. If so, please check out the .Xr xargs 1 utility. --- kern.bootfile str The kernel which was used to boot the system. --- kern.boottime str The time at which the current kernel became active after the system booted. This is a read-only variable. --- kern.chroot_allow_open_directories bool Depending on the setting of this variable, open file descriptors which reference directories will fail. If set to .Em 0 , .Xr chroot 8 will always fail with .Er EPERM if there are any directories open. If set to .Em 1 (the default), .Xr chroot 8 will fail with .Er EPERM if there are any directories open and the process is already subject to the .Xr chroot 8 system call. Any other value will bypass the check for open directories. Please see the .Xr chroot 2 man page for more information. --- kern.clockrate struct Displays information about the system clock. This is a read-only variable. --- kern.console --- kern.coredump bool Determines where the kernel should dump a core file in the event of a kernel panic. --- kern.corefile str Describes the file name that a core image should be stored to. See the .Xr core 5 man page for more information on this variable. --- kern.cp_time struct Contains CPU time statistics. This is a read-only variable. --- kern.devname struct An internally used .Nm that returns suitable device names for the .Fn devname function. See the .Xr devname 3 manual page for more information. --- kern.devstat.all struct An internally used .Nm that returns current devstat statistics as well as the current devstat generation number. See the .Xr devstat 3 man page for more information. --- kern.devstat.generation --- kern.devstat.numdevs --- kern.devstat.version int Displays the devstat list version number. This is a read-only variable. --- kern.disks str Display disk devices that the kernel is currently aware of. This is a read-only variable. --- kern.domainname str This shows the name of the current YP/NIS domain. --- kern.drainwait int The time to wait after dropping DTR to the given number. The units are measured in hundredths of a second. The default is 300 hundredths, i.e., 3 seconds. This option is needed mainly to set proper recover time after modem resets. --- kern.elf32.fallback_brand --- kern.fallback_elf_brand --- kern.file struct Returns the entire file structure. --- kern.function_list struct Returns all functions names in the kernel. --- kern.geom.confdot --- kern.geom.conftxt --- kern.geom.confxml --- kern.hostid int This .Nm may contain the IP address of the system. --- kern.hostname str Display the system hostname. This can be modified with the .Xr hostname 1 utility. --- kern.init_path string The path to search for the .Xr init 8 process. This is a read-only variable. --- kern.iov_max --- kern.ipc.clust_hiwm --- kern.ipc.clust_lowm --- kern.ipc.maxsockbuf int The maximum buffer size that may be allocated for sockets. See .Xr getsockopt 2 for more information. --- kern.ipc.maxsockets int The maximum number of sockets available. --- kern.ipc.mb_statpcpu --- kern.ipc.mbstat --- kern.ipc.mbuf_hiwm --- kern.ipc.mbuf_lowm --- kern.ipc.mbuf_wait --- kern.ipc.msqids --- kern.ipc.nmbclusters bool Maximum number of mbuf clusters available. The kernel uses a preallocated pool of .Dq mbuf clusters for the .Xr mbuf 9 allocator. The pool size is tuned by the kernel during boot. That size is set to a value which seems appropriate for the current system. --- kern.ipc.nmbcnt --- kern.ipc.nmbufs --- kern.ipc.nsfbufs --- kern.ipc.numopensockets --- kern.ipc.somaxconn int The maximum pending socket connection queue size. --- kern.ipc.zero_copy.receive bool When set to a non-zero value, zero copy is enabled for received packets. This reduces copying of data around for outgoing packets and can significantly improve throughput for network connections. --- kern.ipc.zero_copy.send bool When set to a non-zero value, zero copy is enabled for sent packets. This reduces copying of data around for outgoing packets and can significantly improve throughput for network connections. --- kern.job_control bool Reports whether or not job control is available. This is a read-only variable. --- kern.kq_calloutmax --- kern.lastpid int Displays the last PID used by a process. This is a read-only variable. --- kern.logsigexit bool Tells the kernel whether or not to log fatal signal exits. --- kern.malloc str Displays how memory is currently being allocated. This is a read-only variable. --- kern.maxfiles int The maximum number of files allowed for all the processes of the running kernel. You can override the default value which the kernel calculates by explicitly setting this to a non-zero value. Also see the .Xr tuning 7 man page for more information. --- kern.maxfilesperproc int The maximum number of files any one process can open. See the .Xr ps 1 utility for more information on monitoring processes. --- kern.maxproc int The maximum number of processes that the system can be running at any time. See the .Xr ps 1 utility for more information on monitoring processes. --- kern.maxprocperuid int The maximum number of processes one user ID can run. See the .Xr ps 1 utility for more information on monitoring processes. --- kern.maxusers int Controls the scaling of a number of static system tables, including defaults for the maximum number of open files, sizing of network memory resources, etc. See the .Xr tuning 7 man page for more information. This .Nm cannot be set using .Xr sysctl 8 . Use .Xr loader 8 instead to set this at boot time. --- kern.maxvnodes bool The maximum number of .Em vnodes (virtual file system nodes) the system can have open simultaneously. --- kern.minvnodes bool The minimun number of .Em vnodes (virtual file system nodes) the system can have open simultaneously. --- kern.module_path str This .Nm holds a colon-separated list of directories in which the kernel will search for loadable kernel modules. This path is search when using commands such as .Xr kldload 8 and .Xr kldunload 8 . --- kern.msgbuf string Contains the kernel message buffer. --- kern.msgbuf_clear bool Giving this .Nm a value of 1 (one) will cause the kernel message buffer to be cleared. It should be noted though, that the .Nm will then automatically revert back to it's original value of 0 (zero). --- kern.ngroups int Contains the maximum number of groups that a user may belong to. This is a read-only variable. --- kern.openfiles int Shows the current amount of system-wide open files. This is useful when used in conjunction with .Va kern.maxfiles for tuning your system. This is a read-only variable. --- kern.osreldate string Displays the kernel release date. This is a read-only variable. --- kern.osrelease str Displays the current version of .Fx running. This is a read-only variable. --- kern.osrevision string Displays the operating system revision. This is a read-only variable. --- kern.ostype str Alter the name of the current operating system. Changing this will change the output from the .Xr uname 1 utility. Changing the default is not recommended. --- kern.posix1version string Returns the version of .Tn POSIX that the system is attempting to comply with. This is a read-only variable. --- kern.proc.all --- kern.proc.args int Allows a process to retrieve the argument list or process title for another process without looking in the address space of another program. This is a read-only variable. --- kern.proc.pgrp --- kern.proc.pid struct This internally used .Nm may be used to extract process information. See .Xr sysctl 3 for an example. --- kern.proc.ruid --- kern.proc.tty --- kern.proc.uid --- kern.ps_argsopen bool By setting this to 0, command line arguments are hidden for processes which you are not running. This is useful on multi-user machines where things like passwords might accidentally be added to command line programs. --- kern.quantum --- kern.random.adaptors str Displays registered PRNG adaptors. This is a read-only variable. --- kern.random.sys.burst --- kern.random.sys.harvest.ethernet --- kern.random.sys.harvest.interrupt --- kern.random.sys.harvest.point_to_point --- kern.random.sys.harvest.swi --- kern.random.sys.seeded --- kern.random.yarrow.bins --- kern.random.yarrow.fastthresh --- kern.random.yarrow.gengateinterval --- kern.random.yarrow.slowoverthresh --- kern.random.yarrow.slowthresh --- kern.randompid --- kern.rootdev string Displays the current root file system device. This is a read-only variable. --- kern.saved_ids bool Displays whether or not saved set-group/user ID is available. This is a read-only variable. --- kern.securelevel bool The current kernel security level. See the .Xr init 8 manual page for a good description about what a security level is. --- kern.sugid_coredump bool By default, a process that changes user or group credentials whether real or effective will not create a corefile. This behavior can be changed to generate a core dump by setting this variable to 1. --- kern.sync_on_panic bool In the event of a panic, this variable controls whether or not the system should try and .Xr sync 8 . In some circumstances, this could cause a double panic, and as a result, this may be turned off if needed. --- kern.threads.debug bool Determines whether to use debugging for kernel threads. This is useful for testing. --- kern.threads.max_groups_per_proc --- kern.threads.max_threads_hits --- kern.threads.max_threads_per_proc --- kern.threads.virtual_cpu int The maximum amount of virtual CPU's that be used for threading. --- kern.tty_nin --- kern.tty_nout --- kern.ttys bool Used internally by the .Xr pstat 8 command. This is a read-only variable. --- kern.version str Displays the current kernel version information. This is a read-only variable. --- machdep.acpi_root --- machdep.cpu_idle_hlt bool Halt idle CPUs. This is good for an SMP system. --- machdep.disable_mtrrs --- machdep.guessed_bootdev --- machdep.hyperthreading_allowed bool Setting this tunable to zero disables the use of additional logical processors provided by Intel HTT technology. --- machdep.panic_on_nmi --- machdep.siots --- net.inet.accf.unloadable --- net.inet.icmp.bmcastecho --- net.inet.icmp.drop_redirect --- net.inet.icmp.icmplim --- net.inet.icmp.icmplim_output --- net.inet.icmp.log_redirect --- net.inet.icmp.maskfake --- net.inet.icmp.maskrepl --- net.inet.ip.accept_sourceroute bool Controls forwarding of source-routed IP packets. --- net.inet.ip.check_interface bool This .Nm verifies that packets arrive on the correct interfaces. --- net.inet.ip.fastforwarding bool When fast forwarding is enabled, IP packets are forwarded directly to the appropriate network interface with a minimal validity checking, which greatly improves throughput. Please see the .Xr inet 4 man page for more information. --- net.inet.ip.forwarding bool Act as a gateway machine and forward packets. This can also be configured using the gateway_enable value in .Pa /etc/rc.conf --- net.inet.ip.fw.one_pass int --- net.inet.ip.intr_queue_drops --- net.inet.ip.intr_queue_maxlen --- net.inet.ip.maxfragpackets --- net.inet.ip.maxfragsperpacket --- net.inet.ip.redirect bool Controls the sending of ICMP redirects in response to unforwardable IP packets. --- net.inet.ip.rtexpire int Lifetime in seconds of protocol-cloned IP routes after the last reference drops (default one hour). --- net.inet.ip.rtmaxcache int Trigger level of cached, unreferenced, protocol-cloned routes which initiates dynamic adaptation. --- net.inet.ip.rtminexpire int See .Xr inet 4 for more information. --- -net.inet.ip.sendsourcequench -bool - -This -.Nm -enables or disables the transmission of -source quench packets. - ---- net.inet.ip.sourceroute bool Determines whether or not source routed IP packets should be forwarded. --- net.inet.ip.stats --- net.inet.ip.ttl int The TTL (time-to-live) to use for outgoing packets. --- net.inet.raw.maxdgram --- net.inet.raw.olddiverterror --- net.inet.raw.pcblist --- net.inet.raw.recvspace --- net.inet.tcp.always_keepalive bool Determines whether or not to attempt to detect dead TCP connections by sending 'keepalives' intermittently. This is enabled by default and can also be configured using the tcp_keepalive value in .Pa /etc/rc.conf --- net.inet.tcp.blackhole bool Manipulates system behavior when connection requests are received on a TCP port without a socket listening. See the .Xr blackhole 4 man page for more information. --- net.inet.tcp.delacktime --- net.inet.tcp.delayed_ack bool Historically speaking, this feature was designed to allow the acknowledgment to transmitted data to be returned along with the response. See the .Xr tuning 7 man page for more information. --- net.inet.tcp.do_tcpdrain --- net.inet.tcp.getcred --- net.inet.tcp.icmp_may_rst --- net.inet.tcp.inflight_debug bool Control debugging for the .Va net.inet.tcp.inflight_enable .Nm . Please see the .Xr tuning 7 man page for more information. --- net.inet.tcp.inflight_enable bool Turns on bandwidth delay product limiting for all TCP connections. Please see the .Xr tuning 7 man page for more information. --- net.inet.tcp.inflight_max bool .Em double check The maximum amount of data that may be queued for bandwidth delay product limiting. --- net.inet.tcp.inflight_min bool .Em double check The minimum amount of data that may be queued for bandwidth delay product limiting. --- net.inet.tcp.inflight_stab bool This parameter represents the maximal packets added to the bandwidth delay product window calculation. Changing this is not recommended. --- net.inet.tcp.isn_reseed_interval --- net.inet.tcp.local_slowstart_flightsize --- net.inet.tcp.log_in_vain bool Allows the system to log connections to TCP ports that do not have sockets listening. This variable can also be tuned by changing the value for log_in_vain in .Pa /etc/rc.conf --- net.inet.tcp.minmss bool Enable for network link optimization TCP can adjust its MSS and thus packet size according to the observed path MTU. This is done dynamically based on feedback from the remote host and network components along the packet path. This information can be abused to pretend an extremely low path MTU. --- net.inet.tcp.minmssoverload bool The PSS rate for the .Va net.inet.tcp.minmss sysctl. Setting this will force packets to be reset and dropped, this should hinder the availability of DoS attacks on WWW servers using POST attacks. --- net.inet.tcp.msl --- net.inet.tcp.mssdflt bool This is the default TCP Maximum Segment Size for TCP packets. The default setting is recommended in most cases. --- net.inet.tcp.v6mssdflt bool This is the default TCP Maximum Segment Size for TCP IPv6 packets. The default setting is recommend in most cases. --- net.inet.tcp.newreno --- net.inet.tcp.path_mtu_discovery --- net.inet.tcp.pcbcount --- net.inet.tcp.pcblist --- net.inet.tcp.recvspace bool This variables controls the amount of receive buffer space for any given TCP connection. This can be particularly useful when tuning network applications. See the .Xr tuning 7 man page for more information. --- net.inet.tcp.rexmit_min --- net.inet.tcp.rexmit_slop --- net.inet.tcp.rfc1323 bool Determines whether support for RFC1323 (TCP Extensions for High Performance) should be enabled. This variable can also be tuned by changing the value for tcp_extensions in .Pa /etc/rc.conf --- net.inet.tcp.rfc1644 --- net.inet.tcp.rfc3042 --- net.inet.tcp.rfc3390 --- net.inet.tcp.sendspace bool This variables controls the amount of send buffer space for any given TCP connection. This can be particularly useful when tuning network applications. See the .Xr tuning 7 manual page for more information. --- net.inet.tcp.slowstart_flightsize --- net.inet.tcp.stats --- net.inet.tcp.syncache.bucketlimit --- net.inet.tcp.syncache.cachelimit --- net.inet.tcp.syncache.count --- net.inet.tcp.syncache.hashsize --- net.inet.tcp.syncache.rexmtlimit --- net.inet.tcp.syncookies --- net.inet.tcp.tcbhashsize --- net.inet.tcp.v6mssdflt --- net.inet.udp.blackhole bool Manipulates system behavior when connection requests are received on a UDP port. See the .Xr blackhole 4 man page for more information. --- net.inet.udp.getcred --- net.inet.udp.log_in_vain bool Allows the system to log connections to UDP ports that do not have sockets listening. This variable can also be tuned by changing the value for log_in_vain in .Pa /etc/rc.conf --- net.inet.udp.maxdgram --- net.inet.udp.pcblist --- net.inet.udp.recvspace --- net.inet.udp.stats --- net.inet6.icmp6.errppslimit --- net.inet6.icmp6.nd6_debug --- net.inet6.icmp6.nd6_delay --- net.inet6.icmp6.nd6_maxnudhint --- net.inet6.icmp6.nd6_mmaxtries --- net.inet6.icmp6.nd6_prune --- net.inet6.icmp6.nd6_umaxtries --- net.inet6.icmp6.nd6_useloopback --- net.inet6.icmp6.nodeinfo --- net.inet6.icmp6.rediraccept --- net.inet6.icmp6.redirtimeout --- net.inet6.tcp6.getcred --- net.inet6.udp6.getcred --- net.isr.enable --- net.link.ether.inet.log_arp_movements --- net.link.ether.inet.log_arp_wrong_iface --- net.link.ether.ipfw --- net.link.generic.ifdata --- net.link.generic.system.ifcount --- net.link.gif.max_nesting bool Determines whether to allow recursive tunnels or not. --- net.link.gif.parallel_tunnels bool Determines whether to allow parallel tunnels or not. --- net.local.dgram.pcblist --- net.local.stream.pcblist --- security.bsd.see_other_uids bool Turning this option on will prevent users from viewing information about processes running under other user id numbers (UIDs). --- security.bsd.suser_enabled --- security.bsd.unprivileged_proc_debug --- security.bsd.unprivileged_read_msgbuf --- security.jail.set_hostname_allowed bool Determines whether or not the root user within the jail can set the hostname. --- security.jail.socket_unixiproute_only --- security.jail.sysvipc_allowed --- security.mac.biba.enabled bool Enables enforcement of the Biba integrity policy. --- security.mac.biba.ptys_equal bool Label .Sm off .Xr pty 4 s .Sm on as .Dq biba/equal upon creation. --- security.mac.biba.revocation_enabled bool Revoke access to objects if the label is changed to dominate the subject. --- security.mac.enforce_fs bool Enforce MAC policies for file system accesses. --- security.mac.enforce_kld bool Enforce MAC policies on .Xr kld 4 . --- security.mac.enforce_network bool Enforce MAC policies on network interfaces. --- security.mac.enforce_pipe bool Enforce MAC policies on pipes. --- security.mac.enforce_process bool Enforce MAC policies between system processes (e.g. .Xr ps 1 , .Xr ktrace 2 ). --- security.mac.enforce_socket bool Enforce MAC policies on sockets. --- security.mac.enforce_system bool Enforce MAC policies on system-related items (e.g. .Xr kenv 1 , .Xr acct 2 , .Xr reboot 2 ). --- security.mac.enforce_vm bool Enforce MAC policies on .Xr mmap 2 and .Xr mprotect 2 . --- security.mac.ifoff.lo_enabled bool Use this too disable network traffic over the loopback .Xr lo 4 interface. See .Xr mac_ifoff 4 for more information. --- security.mac.ifoff.other_enabled bool Use this to enable network traffic over other interfaces. See .Xr mac_ifoff 4 for more information. --- security.mac.ifoff.bpfrecv_enabled bool Use this too allow .Xr bpf 4 traffic to be received, even while other traffic is disabled. --- security.mac.mls.enabled bool Enables the enforcement of the MLS confidentiality policy, see .Xr mac_mls 4 for more information. --- security.mac.mls.ptys_equal bool Label .Sm off .Xr pty 4 s .Sm on as .Dq mls/equal upon creation. --- security.mac.mls.revocation_enabled bool Revoke access to objects if the label is changed to a more sensitive level than the subject. --- security.mac.portacl.rules str The port access control list is specified in the following format: .Sy idtype .Li : .Sy id .Li : .Sy protocol .Li : .Sy port .Li [, .Sy idtype .Li : .Sy id .Li : .Sy protocol .Li : .Sy port .Li ,...] .Sy idtype Describes the type of subject match to be performed. Either .Li uid for userid matching, or .Li gid for group ID matching. .Sy id The user or group ID (depending on .Sy idtype ) allowed to bind to the specified port. .Bf -emphasis NOTE: User and group names are not valid; only the actual ID numbers may be used. .Ef .Sy protocol Describes which protocol this entry applies to. Either .Li tcp or .Li udp are supported. .Sy port Describes which port this entry applies to. .Bf -emphasis NOTE: MAC security policies may not override other security system policies by allowing accesses that they may deny, such as .Va net.inet.ip.portrange.reservedlow / .Va net.inet.ip.portrange.reservedhigh . .Ef --- security.mac.seeotheruids.enabled bool Enable/disable .Va security.mac.seeotheruids See .Xr mac_seeotheruids 4 for more information. --- security.mac.seeotheruids.primarygroup_enabled bool Allow users to see processes and sockets owned by the same primary group. --- security.mac.seeotheruids.specificgid_enabled bool Allow processes with a specific group ID to be exempt from the policy, set this to .Li 1 and set .Va security.mac.seeotheruids.specificgid to the gid to be exempted. --- security.mac_test str Used for debugging. See .Xr mac_test 4 for more information. --- user.bc_base_max --- user.bc_dim_max --- user.bc_scale_max --- user.bc_string_max --- user.coll_weights_max --- user.cs_path --- user.line_max --- user.posix2_c_bind --- user.posix2_c_dev --- user.posix2_fort_dev --- user.posix2_fort_run --- user.posix2_localedef --- user.posix2_sw_dev --- user.posix2_upe --- user.posix2_version --- user.re_dup_max --- user.stream_max --- user.tzname_max --- vfs.altbufferflushes --- vfs.bufdefragcnt --- vfs.buffreekvacnt --- vfs.bufmallocspace --- vfs.bufreusecnt --- vfs.bufspace --- vfs.cache.nchstats --- vfs.conflist --- vfs.devfs.generation --- vfs.devfs.inodes --- vfs.devfs.noverflow --- vfs.devfs.topinode --- vfs.dirtybufferflushes --- vfs.dirtybufthresh --- vfs.ffs.adjblkcnt --- vfs.ffs.adjrefcnt --- vfs.ffs.freeblks --- vfs.ffs.freedirs --- vfs.ffs.freefiles --- vfs.ffs.setflags --- vfs.flushwithdeps --- vfs.getnewbufcalls --- vfs.getnewbufrestarts --- vfs.hibufspace --- vfs.hidirtybuffers --- vfs.hifreebuffers --- vfs.hirunningspace --- vfs.lobufspace --- vfs.lodirtybuffers --- vfs.lofreebuffers --- vfs.lorunningspace --- vfs.maxbufspace --- vfs.maxmallocbufspace --- vfs.numdirtybuffers --- vfs.numfreebuffers --- vfs.opv_numops --- vfs.pfs.vncache.entries --- vfs.pfs.vncache.hits --- vfs.pfs.vncache.maxentries --- vfs.pfs.vncache.misses --- vfs.read_max --- vfs.recursiveflushes --- vfs.runningbufspace --- vfs.ufs.dirhash_docheck --- vfs.ufs.dirhash_maxmem --- vfs.ufs.dirhash_mem --- vfs.ufs.dirhash_minsize --- vfs.usermount bool This .Nm allows the root user to grant access to non-root users so that they may mount floppy and CD-ROM drives. --- vfs.vmiodirenable bool Controls how directories are cached by the system. This is turned on by default. See the .Xr tuning 7 man page for a more detailed explanation on this variable. --- vfs.write_behind bool Tells the file system to issue media writes as full clusters are collected, which typically occurs when writing large sequential files. This is turned on by default, but under certain circumstances may stall processes and can therefore be turned off. --- vm.defer_swapspace_pageouts --- vm.disable_swapspace_pageouts --- vm.dmmax --- vm.kvm_free --- vm.kvm_size --- vm.loadavg struct Displays the load average history. This is a read-only variable. --- vm.max_launder --- vm.nswapdev int Displays the number of swap devices available to the system. This is a read-only variable. --- vm.pageout_full_stats_interval --- vm.pageout_lock_miss --- vm.pageout_stats_free_max --- vm.pageout_stats_interval --- vm.pageout_stats_max --- vm.stats.sys.v_intr --- vm.stats.sys.v_soft --- vm.stats.sys.v_swtch --- vm.stats.sys.v_syscall --- vm.stats.sys.v_trap --- vm.stats.vm.v_cow_faults --- vm.stats.vm.v_cow_optim --- vm.stats.vm.v_forkpages --- vm.stats.vm.v_forks --- vm.stats.vm.v_intrans --- vm.stats.vm.v_kthreadpages --- vm.stats.vm.v_kthreads --- vm.stats.vm.v_ozfod --- vm.stats.vm.v_pdpages --- vm.stats.vm.v_pdwakeups --- vm.stats.vm.v_reactivated --- vm.stats.vm.v_rforkpages --- vm.stats.vm.v_rforks --- vm.stats.vm.v_swapin --- vm.stats.vm.v_swapout --- vm.stats.vm.v_swappgsin --- vm.stats.vm.v_swappgsout --- vm.stats.vm.v_vforkpages --- vm.stats.vm.v_vforks --- vm.stats.vm.v_vm_faults --- vm.stats.vm.v_vnodein --- vm.stats.vm.v_vnodeout --- vm.stats.vm.v_vnodepgsin --- vm.stats.vm.v_vnodepgsout --- vm.stats.vm.v_zfod --- vm.swap_async_max int The maximum number of in-progress async operations that may be performed. --- vm.swap_enabled bool Determines whether or not processes may swap. --- vm.swap_idle_enabled See .Xr tuning 7 for a detailed explanation of this .Nm . --- vm.swap_info --- vm.vmtotal string Displays virtual memory statistics which are collected at five second intervals. --- vm.zone string Shows memory used by the kernel zone allocator, by zone. This information can also be found by using the .Xr vmstat 8 command. ---