Index: projects/ipsec/sys/netinet/ip_input.c =================================================================== --- projects/ipsec/sys/netinet/ip_input.c (revision 309109) +++ projects/ipsec/sys/netinet/ip_input.c (revision 309110) @@ -1,1345 +1,1347 @@ /*- * 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_ipstealth.h" #include "opt_ipsec.h" #include "opt_route.h" #include "opt_rss.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef IPSEC +#include +#include #include #endif /* IPSEC */ #include #include #include #ifdef CTASSERT CTASSERT(sizeof(struct ip) == 20); #endif /* IP reassembly functions are defined in ip_reass.c. */ extern void ipreass_init(void); extern void ipreass_drain(void); extern void ipreass_slowtimo(void); #ifdef VIMAGE extern void ipreass_destroy(void); #endif struct rmlock in_ifaddr_lock; RM_SYSINIT(in_ifaddr_lock, &in_ifaddr_lock, "in_ifaddr_lock"); VNET_DEFINE(int, rsvp_on); VNET_DEFINE(int, ipforwarding); SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipforwarding), 0, "Enable IP forwarding between interfaces"); static VNET_DEFINE(int, ipsendredirects) = 1; /* XXX */ #define V_ipsendredirects VNET(ipsendredirects) SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsendredirects), 0, "Enable sending IP redirects"); /* * XXX - Setting ip_checkinterface mostly implements the receive side of * the Strong ES model described in RFC 1122, but since the routing table * and transmit implementation do not implement the Strong ES model, * setting this to 1 results in an odd hybrid. * * XXX - ip_checkinterface currently must be disabled if you use ipnat * to translate the destination address to another local interface. * * XXX - ip_checkinterface must be disabled if you add IP aliases * to the loopback interface instead of the interface where the * packets for those addresses are received. */ static VNET_DEFINE(int, ip_checkinterface); #define V_ip_checkinterface VNET(ip_checkinterface) SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_checkinterface), 0, "Verify packet arrives on correct interface"); VNET_DEFINE(struct pfil_head, inet_pfil_hook); /* Packet filter hooks */ static struct netisr_handler ip_nh = { .nh_name = "ip", .nh_handler = ip_input, .nh_proto = NETISR_IP, #ifdef RSS .nh_m2cpuid = rss_soft_m2cpuid_v4, .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_soft_m2cpuid_v4, .nh_policy = NETISR_POLICY_CPU, .nh_dispatch = NETISR_DISPATCH_HYBRID, }; #endif extern struct domain inetdomain; extern struct protosw inetsw[]; u_char ip_protox[IPPROTO_MAX]; VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead); /* first inet address */ VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table */ VNET_DEFINE(u_long, in_ifaddrhmask); /* mask for hash table */ #ifdef IPCTL_DEFMTU SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, &ip_mtu, 0, "Default MTU"); #endif #ifdef IPSTEALTH VNET_DEFINE(int, ipstealth); SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipstealth), 0, "IP stealth mode, no TTL decrementation on forwarding"); #endif /* * IP statistics are stored in the "array" of counter(9)s. */ VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat); VNET_PCPUSTAT_SYSINIT(ipstat); SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)"); #ifdef VIMAGE VNET_PCPUSTAT_SYSUNINIT(ipstat); #endif /* VIMAGE */ /* * Kernel module interface for updating ipstat. The argument is an index * into ipstat treated as an array. */ void kmod_ipstat_inc(int statnum) { counter_u64_add(VNET(ipstat)[statnum], 1); } void kmod_ipstat_dec(int statnum) { counter_u64_add(VNET(ipstat)[statnum], -1); } static int sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS) { int error, qlimit; netisr_getqlimit(&ip_nh, &qlimit); error = sysctl_handle_int(oidp, &qlimit, 0, req); if (error || !req->newptr) return (error); if (qlimit < 1) return (EINVAL); return (netisr_setqlimit(&ip_nh, qlimit)); } SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_queue_maxlen, "I", "Maximum size of the IP input queue"); static int sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS) { u_int64_t qdrops_long; int error, qdrops; netisr_getqdrops(&ip_nh, &qdrops_long); qdrops = qdrops_long; error = sysctl_handle_int(oidp, &qdrops, 0, req); if (error || !req->newptr) return (error); if (qdrops != 0) return (EINVAL); netisr_clearqdrops(&ip_nh); return (0); } SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_queue_drops, "I", "Number of packets dropped from the IP input queue"); #ifdef RSS static int sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS) { int error, qlimit; netisr_getqlimit(&ip_direct_nh, &qlimit); error = sysctl_handle_int(oidp, &qlimit, 0, req); if (error || !req->newptr) return (error); if (qlimit < 1) return (EINVAL); return (netisr_setqlimit(&ip_direct_nh, qlimit)); } SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_direct_queue_maxlen, CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_direct_queue_maxlen, "I", "Maximum size of the IP direct input queue"); static int sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS) { u_int64_t qdrops_long; int error, qdrops; netisr_getqdrops(&ip_direct_nh, &qdrops_long); qdrops = qdrops_long; error = sysctl_handle_int(oidp, &qdrops, 0, req); if (error || !req->newptr) return (error); if (qdrops != 0) return (EINVAL); netisr_clearqdrops(&ip_direct_nh); return (0); } SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_direct_queue_drops, CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_direct_queue_drops, "I", "Number of packets dropped from the IP direct input queue"); #endif /* RSS */ /* * IP initialization: fill in IP protocol switch table. * All protocols not implemented in kernel go to raw IP protocol handler. */ void ip_init(void) { struct protosw *pr; int i; TAILQ_INIT(&V_in_ifaddrhead); V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask); /* Initialize IP reassembly queue. */ ipreass_init(); /* Initialize packet filter hooks. */ V_inet_pfil_hook.ph_type = PFIL_TYPE_AF; V_inet_pfil_hook.ph_af = AF_INET; if ((i = pfil_head_register(&V_inet_pfil_hook)) != 0) printf("%s: WARNING: unable to register pfil hook, " "error %d\n", __func__, i); if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET, &V_ipsec_hhh_in[HHOOK_IPSEC_INET], HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0) printf("%s: WARNING: unable to register input helper hook\n", __func__); if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET, &V_ipsec_hhh_out[HHOOK_IPSEC_INET], HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0) printf("%s: WARNING: unable to register output helper hook\n", __func__); /* Skip initialization of globals for non-default instances. */ #ifdef VIMAGE if (!IS_DEFAULT_VNET(curvnet)) { netisr_register_vnet(&ip_nh); #ifdef RSS netisr_register_vnet(&ip_direct_nh); #endif return; } #endif pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); if (pr == NULL) panic("ip_init: PF_INET not found"); /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */ for (i = 0; i < IPPROTO_MAX; i++) ip_protox[i] = pr - inetsw; /* * Cycle through IP protocols and put them into the appropriate place * in ip_protox[]. */ for (pr = inetdomain.dom_protosw; pr < inetdomain.dom_protoswNPROTOSW; pr++) if (pr->pr_domain->dom_family == PF_INET && pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) { /* Be careful to only index valid IP protocols. */ if (pr->pr_protocol < IPPROTO_MAX) ip_protox[pr->pr_protocol] = pr - inetsw; } netisr_register(&ip_nh); #ifdef RSS netisr_register(&ip_direct_nh); #endif } #ifdef VIMAGE static void ip_destroy(void *unused __unused) { struct ifnet *ifp; int error; #ifdef RSS netisr_unregister_vnet(&ip_direct_nh); #endif netisr_unregister_vnet(&ip_nh); if ((error = pfil_head_unregister(&V_inet_pfil_hook)) != 0) printf("%s: WARNING: unable to unregister pfil hook, " "error %d\n", __func__, error); error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET]); if (error != 0) { printf("%s: WARNING: unable to deregister input helper hook " "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET: " "error %d returned\n", __func__, error); } error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET]); if (error != 0) { printf("%s: WARNING: unable to deregister output helper hook " "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET: " "error %d returned\n", __func__, error); } /* Remove the IPv4 addresses from all interfaces. */ in_ifscrub_all(); /* Make sure the IPv4 routes are gone as well. */ IFNET_RLOCK(); TAILQ_FOREACH(ifp, &V_ifnet, if_link) rt_flushifroutes_af(ifp, AF_INET); IFNET_RUNLOCK(); /* Destroy IP reassembly queue. */ ipreass_destroy(); /* Cleanup in_ifaddr hash table; should be empty. */ hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask); } VNET_SYSUNINIT(ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_destroy, NULL); #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); } /* Try to forward the packet, but if we fail continue */ #ifdef IPSEC /* For now we do not handle IPSEC in tryforward. */ if (!key_havesp(IPSEC_DIR_INBOUND) && !key_havesp(IPSEC_DIR_OUTBOUND) && (V_ipforwarding == 1)) if (ip_tryforward(m) == NULL) return; /* * Bypass packet filtering for packets previously handled by IPsec. */ if (ip_ipsec_filtertunnel(m)) goto passin; #else if (V_ipforwarding == 1) if (ip_tryforward(m) == NULL) return; #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) { if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) { /* * Directly ship the packet on. This allows * forwarding packets originally destined to us * to some other directly connected host. */ ip_forward(m, 1); return; } } passin: /* * Process options and, if not destined for us, * ship it on. ip_dooptions returns 1 when an * error was detected (causing an icmp message * to be sent and the original packet to be freed). */ if (hlen > sizeof (struct ip) && ip_dooptions(m, 0)) return; /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no * matter if it is destined to another node, or whether it is * a multicast one, RSVP wants it! and prevents it from being forwarded * anywhere else. Also checks if the rsvp daemon is running before * grabbing the packet. */ if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP) goto ours; /* * Check our list of addresses, to see if the packet is for us. * If we don't have any addresses, assume any unicast packet * we receive might be for us (and let the upper layers deal * with it). */ if (TAILQ_EMPTY(&V_in_ifaddrhead) && (m->m_flags & (M_MCAST|M_BCAST)) == 0) goto ours; /* * Enable a consistency check between the destination address * and the arrival interface for a unicast packet (the RFC 1122 * strong ES model) if IP forwarding is disabled and the packet * is not locally generated and the packet is not subject to * 'ipfw fwd'. * * XXX - Checking also should be disabled if the destination * address is ipnat'ed to a different interface. * * XXX - Checking is incompatible with IP aliases added * to the loopback interface instead of the interface where * the packets are received. * * XXX - This is the case for carp vhost IPs as well so we * insert a workaround. If the packet got here, we already * checked with carp_iamatch() and carp_forus(). */ checkif = V_ip_checkinterface && (V_ipforwarding == 0) && ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) && ifp->if_carp == NULL && (dchg == 0); /* * Check for exact addresses in the hash bucket. */ /* IN_IFADDR_RLOCK(); */ LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) { /* * If the address matches, verify that the packet * arrived via the correct interface if checking is * enabled. */ if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr && (!checkif || ia->ia_ifp == ifp)) { counter_u64_add(ia->ia_ifa.ifa_ipackets, 1); counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len); /* IN_IFADDR_RUNLOCK(); */ goto ours; } } /* IN_IFADDR_RUNLOCK(); */ /* * Check for broadcast addresses. * * Only accept broadcast packets that arrive via the matching * interface. Reception of forwarded directed broadcasts would * be handled via ip_forward() and ether_output() with the loopback * into the stack for SIMPLEX interfaces handled by ether_output(). */ if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) { IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { if (ifa->ifa_addr->sa_family != AF_INET) continue; ia = ifatoia(ifa); if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == ip->ip_dst.s_addr) { counter_u64_add(ia->ia_ifa.ifa_ipackets, 1); counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len); IF_ADDR_RUNLOCK(ifp); goto ours; } #ifdef BOOTP_COMPAT if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) { counter_u64_add(ia->ia_ifa.ifa_ipackets, 1); counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len); IF_ADDR_RUNLOCK(ifp); goto ours; } #endif } IF_ADDR_RUNLOCK(ifp); ia = NULL; } /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */ if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) { IPSTAT_INC(ips_cantforward); m_freem(m); return; } if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { if (V_ip_mrouter) { /* * If we are acting as a multicast router, all * incoming multicast packets are passed to the * kernel-level multicast forwarding function. * The packet is returned (relatively) intact; if * ip_mforward() returns a non-zero value, the packet * must be discarded, else it may be accepted below. */ if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) { IPSTAT_INC(ips_cantforward); m_freem(m); return; } /* * The process-level routing daemon needs to receive * all multicast IGMP packets, whether or not this * host belongs to their destination groups. */ if (ip->ip_p == IPPROTO_IGMP) goto ours; IPSTAT_INC(ips_forward); } /* * Assume the packet is for us, to avoid prematurely taking * a lock on the in_multi hash. Protocols must perform * their own filtering and update statistics accordingly. */ goto ours; } if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) goto ours; if (ip->ip_dst.s_addr == INADDR_ANY) goto ours; /* * Not for us; forward if possible and desirable. */ if (V_ipforwarding == 0) { IPSTAT_INC(ips_cantforward); m_freem(m); } else { ip_forward(m, dchg); } return; ours: #ifdef IPSTEALTH /* * IPSTEALTH: Process non-routing options only * if the packet is destined for us. */ if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1)) return; #endif /* IPSTEALTH */ /* * Attempt reassembly; if it succeeds, proceed. * ip_reass() will return a different mbuf. */ if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) { /* XXXGL: shouldn't we save & set m_flags? */ m = ip_reass(m); if (m == NULL) return; ip = mtod(m, struct ip *); /* Get the header length of the reassembled packet */ hlen = ip->ip_hl << 2; } #ifdef IPSEC /* * enforce IPsec policy checking if we are seeing last header. * note that we do not visit this with protocols with pcb layer * code - like udp/tcp/raw ip. */ if (IPSEC_INPUT(ipv4, m, ip->ip_p) != 0) goto bad; #endif /* IPSEC */ /* * Switch out to protocol's input routine. */ IPSTAT_INC(ips_delivered); (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p); return; bad: m_freem(m); } /* * IP timer processing; * if a timer expires on a reassembly * queue, discard it. */ void ip_slowtimo(void) { VNET_ITERATOR_DECL(vnet_iter); VNET_LIST_RLOCK_NOSLEEP(); VNET_FOREACH(vnet_iter) { CURVNET_SET(vnet_iter); ipreass_slowtimo(); CURVNET_RESTORE(); } VNET_LIST_RUNLOCK_NOSLEEP(); } void ip_drain(void) { VNET_ITERATOR_DECL(vnet_iter); VNET_LIST_RLOCK_NOSLEEP(); VNET_FOREACH(vnet_iter) { CURVNET_SET(vnet_iter); ipreass_drain(); CURVNET_RESTORE(); } VNET_LIST_RUNLOCK_NOSLEEP(); } /* * The protocol to be inserted into ip_protox[] must be already registered * in inetsw[], either statically or through pf_proto_register(). */ int ipproto_register(short ipproto) { struct protosw *pr; /* Sanity checks. */ if (ipproto <= 0 || ipproto >= IPPROTO_MAX) return (EPROTONOSUPPORT); /* * The protocol slot must not be occupied by another protocol * already. An index pointing to IPPROTO_RAW is unused. */ pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); if (pr == NULL) return (EPFNOSUPPORT); if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */ return (EEXIST); /* Find the protocol position in inetsw[] and set the index. */ for (pr = inetdomain.dom_protosw; pr < inetdomain.dom_protoswNPROTOSW; pr++) { if (pr->pr_domain->dom_family == PF_INET && pr->pr_protocol && pr->pr_protocol == ipproto) { ip_protox[pr->pr_protocol] = pr - inetsw; return (0); } } return (EPROTONOSUPPORT); } int ipproto_unregister(short ipproto) { struct protosw *pr; /* Sanity checks. */ if (ipproto <= 0 || ipproto >= IPPROTO_MAX) return (EPROTONOSUPPORT); /* Check if the protocol was indeed registered. */ pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); if (pr == NULL) return (EPFNOSUPPORT); if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */ return (ENOENT); /* Reset the protocol slot to IPPROTO_RAW. */ ip_protox[ipproto] = pr - inetsw; return (0); } u_char inetctlerrmap[PRC_NCMDS] = { 0, 0, 0, 0, 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, EMSGSIZE, EHOSTUNREACH, 0, 0, 0, 0, EHOSTUNREACH, 0, ENOPROTOOPT, ECONNREFUSED }; /* * Forward a packet. If some error occurs return the sender * an icmp packet. Note we can't always generate a meaningful * icmp message because icmp doesn't have a large enough repertoire * of codes and types. * * If not forwarding, just drop the packet. This could be confusing * if ipforwarding was zero but some routing protocol was advancing * us as a gateway to somewhere. However, we must let the routing * protocol deal with that. * * The srcrt parameter indicates whether the packet is being forwarded * via a source route. */ void ip_forward(struct mbuf *m, int srcrt) { struct ip *ip = mtod(m, struct ip *); struct in_ifaddr *ia; struct mbuf *mcopy; struct sockaddr_in *sin; 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; } if ( #ifdef IPSTEALTH V_ipstealth == 0 && #endif ip->ip_ttl <= IPTTLDEC) { icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0); return; } 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; #ifdef RADIX_MPATH rtalloc_mpath_fib(&ro, ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr), M_GETFIB(m)); #else in_rtalloc_ign(&ro, 0, M_GETFIB(m)); #endif if (ro.ro_rt != NULL) { ia = ifatoia(ro.ro_rt->rt_ifa); ifa_ref(&ia->ia_ifa); } else ia = NULL; /* * 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_copym() 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 == 0) #endif ip->ip_ttl -= IPTTLDEC; #ifdef IPSEC if (IPSEC_FORWARD(ipv4, m, &error) != 0) { /* mbuf consumed by IPsec */ m_freem(mcopy); return; } /* * mbuf wasn't consumed by IPsec, check error code. */ if (error != 0) { IPSTAT_INC(ips_cantforward); m_freem(m); m_freem(mcopy); return; } /* No IPsec processing required */ #endif /* IPSEC */ /* * 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 rtentry *rt; 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; } } } 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; /* * 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: 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: projects/ipsec/sys/netinet/ip_ipsec.c =================================================================== --- projects/ipsec/sys/netinet/ip_ipsec.c (revision 309109) +++ projects/ipsec/sys/netinet/ip_ipsec.c (revision 309110) @@ -1,304 +1,287 @@ /*- * 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. */ #include __FBSDID("$FreeBSD$"); #include "opt_ipsec.h" #include "opt_sctp.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef SCTP #include #endif #include #include #include #include extern struct protosw inetsw[]; static VNET_DEFINE(int, ip4_ipsec_filtertunnel) = 0; #define V_ip4_ipsec_filtertunnel VNET(ip4_ipsec_filtertunnel) SYSCTL_DECL(_net_inet_ipsec); SYSCTL_INT(_net_inet_ipsec, OID_AUTO, filtertunnel, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_ipsec_filtertunnel), 0, "If set filter packets from an IPsec tunnel."); /* * Check if we have to jump over firewall processing for this packet. * Called from ip_input(). * 1 = jump over firewall, 0 = packet goes through firewall. */ int ip_ipsec_filtertunnel(struct mbuf *m) { /* * Bypass packet filtering for packets previously handled by IPsec. */ if (!V_ip4_ipsec_filtertunnel && m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL) return 1; return 0; } /* * Check if protocol type doesn't have a further header and do IPSEC * decryption or reject right now. Protocols with further headers get * their IPSEC treatment within the protocol specific processing. * Called from ip_input(). * 1 = drop packet, 0 = continue processing packet. */ int ip_ipsec_input(struct mbuf *m, int nxt) { /* * 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 ((inetsw[ip_protox[nxt]].pr_flags & PR_LASTHDR) != 0) return (ipsec4_in_reject(m, NULL)); return (0); } /* - * Compute the MTU for a forwarded packet that gets IPSEC encapsulated. - * Called from ip_forward(). - * Returns MTU suggestion for ICMP needfrag reply. - */ -int -ip_ipsec_mtu(struct mbuf *m, int mtu) -{ - /* - * If the packet is routed over IPsec tunnel, tell the - * originator the tunnel MTU. - * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz - * XXX quickhack!!! - */ - return (mtu - ipsec_hdrsiz(m, IPSEC_DIR_OUTBOUND, NULL)); -} - -/* * Called from ip_output(). * 0 = continue processing packet * 1 = packet was consumed, stop processing */ int ip_ipsec_output(struct mbuf *m, struct inpcb *inp, int *error) { struct secpolicy *sp; /* * Check the security policy (SP) for the packet and, if * required, do IPsec-related processing. There are two * cases here; the first time a packet is sent through * it will be untagged and handled by ipsec4_checkpolicy. * If the packet is resubmitted to ip_output (e.g. after * AH, ESP, etc. processing), there will be a tag to bypass * the lookup and related policy checking. */ *error = 0; if (m_tag_find(m, PACKET_TAG_IPSEC_OUT_DONE, NULL) != NULL) return (0); sp = ipsec4_checkpolicy(m, inp, error); /* * There are four return cases: * sp != NULL apply IPsec policy * sp == NULL, error == 0 no IPsec handling needed * sp == NULL, error == -EINVAL discard packet w/o error * sp == NULL, error != 0 discard packet, report error */ if (sp != NULL) { /* * Do delayed checksums now because we send before * this is done in the normal processing path. */ if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { in_delayed_cksum(m); m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; } #ifdef SCTP if (m->m_pkthdr.csum_flags & CSUM_SCTP) { struct ip *ip = mtod(m, struct ip *); sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2)); m->m_pkthdr.csum_flags &= ~CSUM_SCTP; } #endif /* NB: callee frees mbuf and releases reference to SP */ *error = ipsec4_process_packet(m, sp, inp); if (*error == EJUSTRETURN) { /* * We had a SP with a level of 'use' and no SA. We * will just continue to process the packet without * IPsec processing and return without error. */ *error = 0; return (0); } return (1); /* mbuf consumed by IPsec */ } else { /* sp == NULL */ if (*error != 0) { /* * Hack: -EINVAL is used to signal that a packet * should be silently discarded. This is typically * because we have DISCARD policy or asked key * management for an SP and it was delayed (e.g. * kicked up to IKE). * XXX: maybe return EACCES to the caller would * be more useful? */ if (*error == -EINVAL) *error = 0; m_freem(m); return (1); } /* No IPsec processing for this packet. */ } return (0); } /* * Called from ip_forward(). * 1 = drop packet, 0 = forward packet. */ int ip_ipsec_forward(struct mbuf *m, int *error) { struct secpolicy *sp; int idx; /* * Check if this packet has an active inbound SP and needs to be * dropped instead of forwarded. */ if (ipsec4_in_reject(m, NULL) != 0) { *error = EACCES; return (0); } /* * Now check outbound SP. */ sp = ipsec4_checkpolicy(m, NULL, error); /* * There are four return cases: * sp != NULL apply IPsec policy * sp == NULL, error == 0 no IPsec handling needed * sp == NULL, error == -EINVAL discard packet w/o error * sp == NULL, error != 0 discard packet, report error */ if (sp != NULL) { /* * We have SP with IPsec transform, but we should check that * it has tunnel mode request, because we can't use transport * mode when forwarding. */ for (idx = 0; idx < sp->tcount; idx++) { if (sp->req[idx]->saidx.mode == IPSEC_MODE_TUNNEL) break; } if (idx == sp->tcount) { *error = EACCES; IPSECSTAT_INC(ips_out_inval); key_freesp(&sp); return (0); } /* NB: callee frees mbuf and releases reference to SP */ *error = ipsec4_process_packet(m, sp, NULL); if (*error == EJUSTRETURN) { /* * We had a SP with a level of 'use' and no SA. We * will just continue to process the packet without * IPsec processing and return without error. */ *error = 0; return (0); } return (1); /* mbuf consumed by IPsec */ } else { /* sp == NULL */ if (*error != 0) { /* * Hack: -EINVAL is used to signal that a packet * should be silently discarded. This is typically * because we have DISCARD policy or asked key * management for an SP and it was delayed (e.g. * kicked up to IKE). * XXX: maybe return EACCES to the caller would * be more useful? */ if (*error == -EINVAL) *error = 0; m_freem(m); return (1); } /* No IPsec processing for this packet. */ } return (0); } /* * Handle IPsec related socket options. * Called from ip_ctloutput(). */ int ip_ipsec_pcbctl(struct inpcb *inp, struct sockopt *sopt) { if (sopt->sopt_name != IP_IPSEC_POLICY) return (ENOPROTOOPT); return (ipsec_control_pcbpolicy(inp, sopt)); } Index: projects/ipsec/sys/netinet/ip_ipsec.h =================================================================== --- projects/ipsec/sys/netinet/ip_ipsec.h (revision 309109) +++ projects/ipsec/sys/netinet/ip_ipsec.h (revision 309110) @@ -1,45 +1,44 @@ /*- * 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. * * $FreeBSD$ */ #ifndef _NETINET_IP_IPSEC_H_ #define _NETINET_IP_IPSEC_H_ #define IPSEC_INPUT(sc, m, arg) ip_ipsec_input((m), (arg)) #define IPSEC_FORWARD(sc, m, perr) ip_ipsec_forward((m), (perr)) #define IPSEC_OUTPUT(sc, m, inp, perr) ip_ipsec_output((m), (inp), (perr)) int ip_ipsec_filtertunnel(struct mbuf *); int ip_ipsec_input(struct mbuf *, int); -int ip_ipsec_mtu(struct mbuf *, int); int ip_ipsec_forward(struct mbuf *, int *); int ip_ipsec_output(struct mbuf *, struct inpcb *, int *); int ip_ipsec_pcbctl(struct inpcb *, struct sockopt *); #endif Index: projects/ipsec/sys/netinet/tcp_subr.c =================================================================== --- projects/ipsec/sys/netinet/tcp_subr.c (revision 309109) +++ projects/ipsec/sys/netinet/tcp_subr.c (revision 309110) @@ -1,3070 +1,3070 @@ /*- * 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 #ifdef TCP_HHOOK #include #endif #include #ifdef TCP_HHOOK #include #endif #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 #include #include #endif #ifdef TCP_RFC7413 #include #endif #include #include #include #include #include #include #include #ifdef INET6 #include #endif #include #ifdef TCPPCAP #include #endif #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 struct rwlock tcp_function_lock; 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) #ifdef TCP_HHOOK VNET_DEFINE(struct hhook_head *, tcp_hhh[HHOOK_TCP_LAST+1]); #endif static struct inpcb *tcp_notify(struct inpcb *, int); static struct inpcb *tcp_mtudisc_notify(struct inpcb *, int); static void tcp_mtudisc(struct inpcb *, int); static char * tcp_log_addr(struct in_conninfo *inc, struct tcphdr *th, void *ip4hdr, const void *ip6hdr); static struct tcp_function_block tcp_def_funcblk = { "default", tcp_output, tcp_do_segment, tcp_default_ctloutput, NULL, NULL, NULL, NULL, NULL, NULL, 0, 0 }; int t_functions_inited = 0; struct tcp_funchead t_functions; static struct tcp_function_block *tcp_func_set_ptr = &tcp_def_funcblk; static void init_tcp_functions(void) { if (t_functions_inited == 0) { TAILQ_INIT(&t_functions); rw_init_flags(&tcp_function_lock, "tcp_func_lock" , 0); t_functions_inited = 1; } } static struct tcp_function_block * find_tcp_functions_locked(struct tcp_function_set *fs) { struct tcp_function *f; struct tcp_function_block *blk=NULL; TAILQ_FOREACH(f, &t_functions, tf_next) { if (strcmp(f->tf_fb->tfb_tcp_block_name, fs->function_set_name) == 0) { blk = f->tf_fb; break; } } return(blk); } static struct tcp_function_block * find_tcp_fb_locked(struct tcp_function_block *blk, struct tcp_function **s) { struct tcp_function_block *rblk=NULL; struct tcp_function *f; TAILQ_FOREACH(f, &t_functions, tf_next) { if (f->tf_fb == blk) { rblk = blk; if (s) { *s = f; } break; } } return (rblk); } struct tcp_function_block * find_and_ref_tcp_functions(struct tcp_function_set *fs) { struct tcp_function_block *blk; rw_rlock(&tcp_function_lock); blk = find_tcp_functions_locked(fs); if (blk) refcount_acquire(&blk->tfb_refcnt); rw_runlock(&tcp_function_lock); return(blk); } struct tcp_function_block * find_and_ref_tcp_fb(struct tcp_function_block *blk) { struct tcp_function_block *rblk; rw_rlock(&tcp_function_lock); rblk = find_tcp_fb_locked(blk, NULL); if (rblk) refcount_acquire(&rblk->tfb_refcnt); rw_runlock(&tcp_function_lock); return(rblk); } static int sysctl_net_inet_default_tcp_functions(SYSCTL_HANDLER_ARGS) { int error=ENOENT; struct tcp_function_set fs; struct tcp_function_block *blk; memset(&fs, 0, sizeof(fs)); rw_rlock(&tcp_function_lock); blk = find_tcp_fb_locked(tcp_func_set_ptr, NULL); if (blk) { /* Found him */ strcpy(fs.function_set_name, blk->tfb_tcp_block_name); fs.pcbcnt = blk->tfb_refcnt; } rw_runlock(&tcp_function_lock); error = sysctl_handle_string(oidp, fs.function_set_name, sizeof(fs.function_set_name), req); /* Check for error or no change */ if (error != 0 || req->newptr == NULL) return(error); rw_wlock(&tcp_function_lock); blk = find_tcp_functions_locked(&fs); if ((blk == NULL) || (blk->tfb_flags & TCP_FUNC_BEING_REMOVED)) { error = ENOENT; goto done; } tcp_func_set_ptr = blk; done: rw_wunlock(&tcp_function_lock); return (error); } SYSCTL_PROC(_net_inet_tcp, OID_AUTO, functions_default, CTLTYPE_STRING | CTLFLAG_RW, NULL, 0, sysctl_net_inet_default_tcp_functions, "A", "Set/get the default TCP functions"); static int sysctl_net_inet_list_available(SYSCTL_HANDLER_ARGS) { int error, cnt, linesz; struct tcp_function *f; char *buffer, *cp; size_t bufsz, outsz; cnt = 0; rw_rlock(&tcp_function_lock); TAILQ_FOREACH(f, &t_functions, tf_next) { cnt++; } rw_runlock(&tcp_function_lock); bufsz = (cnt+2) * (TCP_FUNCTION_NAME_LEN_MAX + 12) + 1; buffer = malloc(bufsz, M_TEMP, M_WAITOK); error = 0; cp = buffer; linesz = snprintf(cp, bufsz, "\n%-32s%c %s\n", "Stack", 'D', "PCB count"); cp += linesz; bufsz -= linesz; outsz = linesz; rw_rlock(&tcp_function_lock); TAILQ_FOREACH(f, &t_functions, tf_next) { linesz = snprintf(cp, bufsz, "%-32s%c %u\n", f->tf_fb->tfb_tcp_block_name, (f->tf_fb == tcp_func_set_ptr) ? '*' : ' ', f->tf_fb->tfb_refcnt); if (linesz >= bufsz) { error = EOVERFLOW; break; } cp += linesz; bufsz -= linesz; outsz += linesz; } rw_runlock(&tcp_function_lock); if (error == 0) error = sysctl_handle_string(oidp, buffer, outsz + 1, req); free(buffer, M_TEMP); return (error); } SYSCTL_PROC(_net_inet_tcp, OID_AUTO, functions_available, CTLTYPE_STRING|CTLFLAG_RD, NULL, 0, sysctl_net_inet_list_available, "A", "list available TCP Function sets"); /* * 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; #ifdef TCP_HHOOK struct osd osd; #endif }; 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"); MALLOC_DEFINE(M_TCPFUNCTIONS, "tcpfunc", "TCP function set memory"); 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); } int register_tcp_functions(struct tcp_function_block *blk, int wait) { struct tcp_function_block *lblk; struct tcp_function *n; struct tcp_function_set fs; if (t_functions_inited == 0) { init_tcp_functions(); } if ((blk->tfb_tcp_output == NULL) || (blk->tfb_tcp_do_segment == NULL) || (blk->tfb_tcp_ctloutput == NULL) || (strlen(blk->tfb_tcp_block_name) == 0)) { /* * These functions are required and you * need a name. */ return (EINVAL); } if (blk->tfb_tcp_timer_stop_all || blk->tfb_tcp_timer_activate || blk->tfb_tcp_timer_active || blk->tfb_tcp_timer_stop) { /* * If you define one timer function you * must have them all. */ if ((blk->tfb_tcp_timer_stop_all == NULL) || (blk->tfb_tcp_timer_activate == NULL) || (blk->tfb_tcp_timer_active == NULL) || (blk->tfb_tcp_timer_stop == NULL)) { return (EINVAL); } } n = malloc(sizeof(struct tcp_function), M_TCPFUNCTIONS, wait); if (n == NULL) { return (ENOMEM); } n->tf_fb = blk; strcpy(fs.function_set_name, blk->tfb_tcp_block_name); rw_wlock(&tcp_function_lock); lblk = find_tcp_functions_locked(&fs); if (lblk) { /* Duplicate name space not allowed */ rw_wunlock(&tcp_function_lock); free(n, M_TCPFUNCTIONS); return (EALREADY); } refcount_init(&blk->tfb_refcnt, 0); blk->tfb_flags = 0; TAILQ_INSERT_TAIL(&t_functions, n, tf_next); rw_wunlock(&tcp_function_lock); return(0); } int deregister_tcp_functions(struct tcp_function_block *blk) { struct tcp_function_block *lblk; struct tcp_function *f; int error=ENOENT; if (strcmp(blk->tfb_tcp_block_name, "default") == 0) { /* You can't un-register the default */ return (EPERM); } rw_wlock(&tcp_function_lock); if (blk == tcp_func_set_ptr) { /* You can't free the current default */ rw_wunlock(&tcp_function_lock); return (EBUSY); } if (blk->tfb_refcnt) { /* Still tcb attached, mark it. */ blk->tfb_flags |= TCP_FUNC_BEING_REMOVED; rw_wunlock(&tcp_function_lock); return (EBUSY); } lblk = find_tcp_fb_locked(blk, &f); if (lblk) { /* Found */ TAILQ_REMOVE(&t_functions, f, tf_next); f->tf_fb = NULL; free(f, M_TCPFUNCTIONS); error = 0; } rw_wunlock(&tcp_function_lock); return (error); } void tcp_init(void) { const char *tcbhash_tuneable; int hashsize; tcbhash_tuneable = "net.inet.tcp.tcbhashsize"; #ifdef TCP_HHOOK 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__); #endif 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 && IS_DEFAULT_VNET(curvnet)) 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, 0, 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, 0); 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, 0); /* Skip initialization of globals for non-default instances. */ if (!IS_DEFAULT_VNET(curvnet)) return; tcp_reass_global_init(); /* XXX virtualize those bellow? */ tcp_delacktime = TCPTV_DELACK; tcp_keepinit = TCPTV_KEEP_INIT; tcp_keepidle = TCPTV_KEEP_IDLE; tcp_keepintvl = TCPTV_KEEPINTVL; tcp_maxpersistidle = TCPTV_KEEP_IDLE; tcp_msl = TCPTV_MSL; tcp_rexmit_min = TCPTV_MIN; if (tcp_rexmit_min < 1) tcp_rexmit_min = 1; tcp_persmin = TCPTV_PERSMIN; tcp_persmax = TCPTV_PERSMAX; tcp_rexmit_slop = TCPTV_CPU_VAR; tcp_finwait2_timeout = TCPTV_FINWAIT2_TIMEOUT; tcp_tcbhashsize = hashsize; /* Setup the tcp function block list */ init_tcp_functions(); register_tcp_functions(&tcp_def_funcblk, M_WAITOK); 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 TCPPCAP tcp_pcap_init(); #endif #ifdef TCP_RFC7413 tcp_fastopen_init(); #endif } #ifdef VIMAGE static void tcp_destroy(void *unused __unused) { int n; #ifdef TCP_HHOOK int error; #endif /* * All our processes are gone, all our sockets should be cleaned * up, which means, we should be past the tcp_discardcb() calls. * Sleep to let all tcpcb timers really disappear and cleanup. */ for (;;) { INP_LIST_RLOCK(&V_tcbinfo); n = V_tcbinfo.ipi_count; INP_LIST_RUNLOCK(&V_tcbinfo); if (n == 0) break; pause("tcpdes", hz / 10); } tcp_hc_destroy(); syncache_destroy(); tcp_tw_destroy(); in_pcbinfo_destroy(&V_tcbinfo); /* tcp_discardcb() clears the sack_holes up. */ uma_zdestroy(V_sack_hole_zone); uma_zdestroy(V_tcpcb_zone); #ifdef TCP_RFC7413 /* * Cannot free the zone until all tcpcbs are released as we attach * the allocations to them. */ tcp_fastopen_destroy(); #endif #ifdef TCP_HHOOK 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 } VNET_SYSUNINIT(tcp, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH, tcp_destroy, NULL); #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) { struct tcpopt to; struct inpcb *inp; struct ip *ip; struct mbuf *optm; struct tcphdr *nth; u_char *optp; #ifdef INET6 struct ip6_hdr *ip6; int isipv6; #endif /* INET6 */ int optlen, tlen, win; bool incl_opts; 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; incl_opts = false; win = 0; if (tp != NULL) { if (!(flags & TH_RST)) { win = sbspace(&inp->inp_socket->so_rcv); if (win > TCP_MAXWIN << tp->rcv_scale) win = TCP_MAXWIN << tp->rcv_scale; } if ((tp->t_flags & TF_NOOPT) == 0) incl_opts = true; } if (m == NULL) { m = m_gethdr(M_NOWAIT, MT_DATA); if (m == NULL) return; 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 if (!M_WRITABLE(m)) { struct mbuf *n; /* Can't reuse 'm', allocate a new mbuf. */ n = m_gethdr(M_NOWAIT, MT_DATA); if (n == NULL) { m_freem(m); return; } if (!m_dup_pkthdr(n, m, M_NOWAIT)) { m_freem(m); m_freem(n); return; } n->m_data += max_linkhdr; /* m_len is set later */ #define xchg(a,b,type) { type t; t=a; a=b; b=t; } #ifdef INET6 if (isipv6) { bcopy((caddr_t)ip6, mtod(n, caddr_t), sizeof(struct ip6_hdr)); ip6 = mtod(n, struct ip6_hdr *); xchg(ip6->ip6_dst, ip6->ip6_src, struct in6_addr); nth = (struct tcphdr *)(ip6 + 1); } else #endif /* INET6 */ { bcopy((caddr_t)ip, mtod(n, caddr_t), sizeof(struct ip)); ip = mtod(n, struct ip *); xchg(ip->ip_dst.s_addr, ip->ip_src.s_addr, uint32_t); nth = (struct tcphdr *)(ip + 1); } bcopy((caddr_t)th, (caddr_t)nth, sizeof(struct tcphdr)); xchg(nth->th_dport, nth->th_sport, uint16_t); th = nth; m_freem(m); m = n; } else { /* * reuse the mbuf. * XXX MRT We inherit 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 */ #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 } tlen = 0; #ifdef INET6 if (isipv6) tlen = sizeof (struct ip6_hdr) + sizeof (struct tcphdr); #endif #if defined(INET) && defined(INET6) else #endif #ifdef INET tlen = sizeof (struct tcpiphdr); #endif #ifdef INVARIANTS m->m_len = 0; KASSERT(M_TRAILINGSPACE(m) >= tlen, ("Not enough trailing space for message (m=%p, need=%d, have=%ld)", m, tlen, (long)M_TRAILINGSPACE(m))); #endif m->m_len = tlen; to.to_flags = 0; if (incl_opts) { /* Make sure we have room. */ if (M_TRAILINGSPACE(m) < TCP_MAXOLEN) { m->m_next = m_get(M_NOWAIT, MT_DATA); if (m->m_next) { optp = mtod(m->m_next, u_char *); optm = m->m_next; } else incl_opts = false; } else { optp = (u_char *) (nth + 1); optm = m; } } if (incl_opts) { /* Timestamps. */ if (tp->t_flags & TF_RCVD_TSTMP) { to.to_tsval = tcp_ts_getticks() + tp->ts_offset; to.to_tsecr = tp->ts_recent; to.to_flags |= TOF_TS; } #ifdef TCP_SIGNATURE /* TCP-MD5 (RFC2385). */ if (tp->t_flags & TF_SIGNATURE) to.to_flags |= TOF_SIGNATURE; #endif /* Add the options. */ tlen += optlen = tcp_addoptions(&to, optp); /* Update m_len in the correct mbuf. */ optm->m_len += optlen; } else optlen = 0; #ifdef INET6 if (isipv6) { ip6->ip6_flow = 0; ip6->ip6_vfc = IPV6_VERSION; ip6->ip6_nxt = IPPROTO_TCP; ip6->ip6_plen = htons(tlen - sizeof(*ip6)); } #endif #if defined(INET) && defined(INET6) else #endif #ifdef INET { 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_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) + optlen) >> 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; #ifdef TCP_SIGNATURE if (to.to_flags & TOF_SIGNATURE) { tcp_signature_compute(m, 0, 0, optlen, to.to_signature, IPSEC_DIR_OUTBOUND); } #endif 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 TCP_PROBE3(debug__output, tp, th, mtod(m, const char *)); 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, 0, NULL, NULL, inp); #endif /* INET6 */ #if defined(INET) && defined(INET6) else #endif #ifdef INET (void) ip_output(m, NULL, NULL, 0, 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; rw_rlock(&tcp_function_lock); tp->t_fb = tcp_func_set_ptr; refcount_acquire(&tp->t_fb->tfb_refcnt); rw_runlock(&tcp_function_lock); /* * 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) { if (tp->t_fb->tfb_tcp_fb_fini) (*tp->t_fb->tfb_tcp_fb_fini)(tp, 1); refcount_release(&tp->t_fb->tfb_refcnt); uma_zfree(V_tcpcb_zone, tm); return (NULL); } #ifdef TCP_HHOOK tp->osd = &tm->osd; if (khelp_init_osd(HELPER_CLASS_TCP, tp->osd)) { if (tp->t_fb->tfb_tcp_fb_fini) (*tp->t_fb->tfb_tcp_fb_fini)(tp, 1); refcount_release(&tp->t_fb->tfb_refcnt); uma_zfree(V_tcpcb_zone, tm); return (NULL); } #endif #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 = #ifdef INET6 isipv6 ? V_tcp_v6mssdflt : #endif /* INET6 */ V_tcp_mssdflt; /* Set up our timeouts. */ callout_init(&tp->t_timers->tt_rexmt, 1); callout_init(&tp->t_timers->tt_persist, 1); callout_init(&tp->t_timers->tt_keep, 1); callout_init(&tp->t_timers->tt_2msl, 1); callout_init(&tp->t_timers->tt_delack, 1); if (V_tcp_do_rfc1323) tp->t_flags = (TF_REQ_SCALE|TF_REQ_TSTMP); if (V_tcp_do_sack) tp->t_flags |= TF_SACK_PERMIT; TAILQ_INIT(&tp->snd_holes); /* * The tcpcb will hold a reference on its inpcb until tcp_discardcb() * is called. */ in_pcbref(inp); /* Reference for tcpcb */ tp->t_inpcb = inp; /* * Init srtt to TCPTV_SRTTBASE (0), so we can tell that we have no * rtt estimate. Set rttvar so that srtt + 4 * rttvar gives * reasonable initial retransmit time. */ tp->t_srtt = TCPTV_SRTTBASE; tp->t_rttvar = ((TCPTV_RTOBASE - TCPTV_SRTTBASE) << TCP_RTTVAR_SHIFT) / 4; tp->t_rttmin = tcp_rexmit_min; tp->t_rxtcur = TCPTV_RTOBASE; tp->snd_cwnd = TCP_MAXWIN << TCP_MAX_WINSHIFT; tp->snd_ssthresh = TCP_MAXWIN << TCP_MAX_WINSHIFT; tp->t_rcvtime = ticks; /* * IPv4 TTL initialization is necessary for an IPv6 socket as well, * because the socket may be bound to an IPv6 wildcard address, * which may match an IPv4-mapped IPv6 address. */ inp->inp_ip_ttl = V_ip_defttl; inp->inp_ppcb = tp; #ifdef TCPPCAP /* * Init the TCP PCAP queues. */ tcp_pcap_tcpcb_init(tp); #endif if (tp->t_fb->tfb_tcp_fb_init) { (*tp->t_fb->tfb_tcp_fb_init)(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_WLOCK(&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_WUNLOCK(&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_LOCK_ASSERT(&V_tcbinfo); INP_WLOCK_ASSERT(tp->t_inpcb); if (TCPS_HAVERCVDSYN(tp->t_state)) { tcp_state_change(tp, TCPS_CLOSED); (void) tp->t_fb->tfb_tcp_output(tp); TCPSTAT_INC(tcps_drops); } else TCPSTAT_INC(tcps_conndrops); if (errno == ETIMEDOUT && tp->t_softerror) errno = tp->t_softerror; so->so_error = errno; return (tcp_close(tp)); } void tcp_discardcb(struct tcpcb *tp) { struct inpcb *inp = tp->t_inpcb; struct socket *so = inp->inp_socket; #ifdef INET6 int isipv6 = (inp->inp_vflag & INP_IPV6) != 0; #endif /* INET6 */ int released; INP_WLOCK_ASSERT(inp); /* * Make sure that all of our timers are stopped before we delete the * PCB. * * If stopping a timer fails, we schedule a discard function in same * callout, and the last discard function called will take care of * deleting the tcpcb. */ tp->t_timers->tt_draincnt = 0; tcp_timer_stop(tp, TT_REXMT); tcp_timer_stop(tp, TT_PERSIST); tcp_timer_stop(tp, TT_KEEP); tcp_timer_stop(tp, TT_2MSL); tcp_timer_stop(tp, TT_DELACK); if (tp->t_fb->tfb_tcp_timer_stop_all) { /* * Call the stop-all function of the methods, * this function should call the tcp_timer_stop() * method with each of the function specific timeouts. * That stop will be called via the tfb_tcp_timer_stop() * which should use the async drain function of the * callout system (see tcp_var.h). */ tp->t_fb->tfb_tcp_timer_stop_all(tp); } /* * 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; uint32_t 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 occurred 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 *= (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); #ifdef TCPPCAP /* Free the TCP PCAP queues. */ tcp_pcap_drain(&(tp->t_inpkts)); tcp_pcap_drain(&(tp->t_outpkts)); #endif /* Allow the CC algorithm to clean up after itself. */ if (CC_ALGO(tp)->cb_destroy != NULL) CC_ALGO(tp)->cb_destroy(tp->ccv); #ifdef TCP_HHOOK khelp_destroy_osd(tp->osd); #endif CC_ALGO(tp) = NULL; inp->inp_ppcb = NULL; if (tp->t_timers->tt_draincnt == 0) { /* We own the last reference on tcpcb, let's free it. */ if (tp->t_fb->tfb_tcp_fb_fini) (*tp->t_fb->tfb_tcp_fb_fini)(tp, 1); refcount_release(&tp->t_fb->tfb_refcnt); tp->t_inpcb = NULL; uma_zfree(V_tcpcb_zone, tp); released = in_pcbrele_wlocked(inp); KASSERT(!released, ("%s: inp %p should not have been released " "here", __func__, inp)); } } void tcp_timer_discard(void *ptp) { struct inpcb *inp; struct tcpcb *tp; tp = (struct tcpcb *)ptp; CURVNET_SET(tp->t_vnet); INP_INFO_RLOCK(&V_tcbinfo); inp = tp->t_inpcb; KASSERT(inp != NULL, ("%s: tp %p tp->t_inpcb == NULL", __func__, tp)); INP_WLOCK(inp); KASSERT((tp->t_timers->tt_flags & TT_STOPPED) != 0, ("%s: tcpcb has to be stopped here", __func__)); tp->t_timers->tt_draincnt--; if (tp->t_timers->tt_draincnt == 0) { /* We own the last reference on this tcpcb, let's free it. */ if (tp->t_fb->tfb_tcp_fb_fini) (*tp->t_fb->tfb_tcp_fb_fini)(tp, 1); refcount_release(&tp->t_fb->tfb_refcnt); tp->t_inpcb = NULL; uma_zfree(V_tcpcb_zone, tp); if (in_pcbrele_wlocked(inp)) { INP_INFO_RUNLOCK(&V_tcbinfo); CURVNET_RESTORE(); return; } } INP_WUNLOCK(inp); INP_INFO_RUNLOCK(&V_tcbinfo); CURVNET_RESTORE(); } /* * Attempt to close a TCP control block, marking it as dropped, and freeing * the socket if we hold the only reference. */ struct tcpcb * tcp_close(struct tcpcb *tp) { struct inpcb *inp = tp->t_inpcb; struct socket *so; INP_INFO_LOCK_ASSERT(&V_tcbinfo); INP_WLOCK_ASSERT(inp); #ifdef TCP_OFFLOAD if (tp->t_state == TCPS_LISTEN) tcp_offload_listen_stop(tp); #endif #ifdef TCP_RFC7413 /* * This releases the TFO pending counter resource for TFO listen * sockets as well as passively-created TFO sockets that transition * from SYN_RECEIVED to CLOSED. */ if (tp->t_tfo_pending) { tcp_fastopen_decrement_counter(tp->t_tfo_pending); tp->t_tfo_pending = NULL; } #endif in_pcbdrop(inp); TCPSTAT_INC(tcps_closed); TCPSTATES_DEC(tp->t_state); 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_WLOCK(&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); #ifdef TCPPCAP if (tcp_pcap_aggressive_free) { /* Free the TCP PCAP queues. */ tcp_pcap_drain(&(tcpb->t_inpkts)); tcp_pcap_drain(&(tcpb->t_outpkts)); } #endif } INP_WUNLOCK(inpb); } INP_INFO_WUNLOCK(&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_LOCK_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)) { if (inp->inp_route.ro_rt) { RTFREE(inp->inp_route.ro_rt); inp->inp_route.ro_rt = (struct rtentry *)NULL; } 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 + counter_u64_fetch(V_tcps_states[TCPS_SYN_RECEIVED]); 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_LIST_RLOCK(&V_tcbinfo); gencnt = V_tcbinfo.ipi_gencnt; n = V_tcbinfo.ipi_count; INP_LIST_RUNLOCK(&V_tcbinfo); m = counter_u64_fetch(V_tcps_states[TCPS_SYN_RECEIVED]); 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); INP_INFO_WLOCK(&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_WUNLOCK(&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_RLOCK(&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_RUNLOCK(&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_LIST_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_LIST_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_UNREACH_PROTOCOL || cmd == PRC_TIMXCEED_INTRANS) && ip) notify = tcp_drop_syn_sent; /* * 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) { in_pcbnotifyall(&V_tcbinfo, faddr, inetctlerrmap[cmd], notify); return; } icp = (struct icmp *)((caddr_t)ip - offsetof(struct icmp, icmp_ip)); th = (struct tcphdr *)((caddr_t)ip + (ip->ip_hl << 2)); INP_INFO_RLOCK(&V_tcbinfo); inp = in_pcblookup(&V_tcbinfo, faddr, th->th_dport, ip->ip_src, th->th_sport, INPLOOKUP_WLOCKPCB, NULL); if (inp != NULL && PRC_IS_REDIRECT(cmd)) { /* signal EHOSTDOWN, as it flushes the cached route */ inp = (*notify)(inp, EHOSTDOWN); if (inp != NULL) INP_WUNLOCK(inp); } else if (inp != NULL) { if (!(inp->inp_flags & INP_TIMEWAIT) && !(inp->inp_flags & INP_DROPPED) && !(inp->inp_socket == NULL)) { icmp_tcp_seq = ntohl(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. */ 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 process the offered MTU if it * is smaller than the current one. */ if (mtu < tp->t_maxseg + sizeof(struct tcpiphdr)) { bzero(&inc, sizeof(inc)); inc.inc_faddr = faddr; inc.inc_fibnum = inp->inp_inc.inc_fibnum; 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_RUNLOCK(&V_tcbinfo); } #endif /* INET */ #ifdef INET6 void tcp6_ctlinput(int cmd, struct sockaddr *sa, void *d) { struct in6_addr *dst; struct tcphdr *th; struct inpcb *(*notify)(struct inpcb *, int) = tcp_notify; struct ip6_hdr *ip6; struct mbuf *m; struct inpcb *inp; struct tcpcb *tp; struct icmp6_hdr *icmp6; struct ip6ctlparam *ip6cp = NULL; const struct sockaddr_in6 *sa6_src = NULL; struct in_conninfo inc; tcp_seq icmp_tcp_seq; unsigned int mtu; unsigned int off; if (sa->sa_family != AF_INET6 || sa->sa_len != sizeof(struct sockaddr_in6)) return; /* if the parameter is from icmp6, decode it. */ if (d != NULL) { ip6cp = (struct ip6ctlparam *)d; icmp6 = ip6cp->ip6c_icmp6; m = ip6cp->ip6c_m; ip6 = ip6cp->ip6c_ip6; off = ip6cp->ip6c_off; sa6_src = ip6cp->ip6c_src; dst = ip6cp->ip6c_finaldst; } else { m = NULL; ip6 = NULL; off = 0; /* fool gcc */ sa6_src = &sa6_any; dst = NULL; } 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_UNREACH_PROTOCOL || cmd == PRC_TIMXCEED_INTRANS) && ip6 != NULL) notify = tcp_drop_syn_sent; /* * 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) ip6 = NULL; else if ((unsigned)cmd >= PRC_NCMDS || inet6ctlerrmap[cmd] == 0) return; if (ip6 == NULL) { in6_pcbnotify(&V_tcbinfo, sa, 0, (const struct sockaddr *)sa6_src, 0, cmd, NULL, notify); return; } /* Check if we can safely get the ports from the tcp hdr */ if (m == NULL || (m->m_pkthdr.len < (int32_t) (off + offsetof(struct tcphdr, th_seq)))) { return; } th = (struct tcphdr *) mtodo(ip6cp->ip6c_m, ip6cp->ip6c_off); INP_INFO_RLOCK(&V_tcbinfo); inp = in6_pcblookup(&V_tcbinfo, &ip6->ip6_dst, th->th_dport, &ip6->ip6_src, th->th_sport, INPLOOKUP_WLOCKPCB, NULL); if (inp != NULL && PRC_IS_REDIRECT(cmd)) { /* signal EHOSTDOWN, as it flushes the cached route */ inp = (*notify)(inp, EHOSTDOWN); if (inp != NULL) INP_WUNLOCK(inp); } else if (inp != NULL) { if (!(inp->inp_flags & INP_TIMEWAIT) && !(inp->inp_flags & INP_DROPPED) && !(inp->inp_socket == NULL)) { icmp_tcp_seq = ntohl(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. */ mtu = ntohl(icmp6->icmp6_mtu); /* * If no alternative MTU was * proposed, or the proposed * MTU was too small, set to * the min. */ if (mtu < IPV6_MMTU) mtu = IPV6_MMTU - 8; bzero(&inc, sizeof(inc)); inc.inc_fibnum = M_GETFIB(m); inc.inc_flags |= INC_ISIPV6; inc.inc6_faddr = *dst; if (in6_setscope(&inc.inc6_faddr, m->m_pkthdr.rcvif, NULL)) goto unlock_inp; /* * Only process the offered MTU if it * is smaller than the current one. */ if (mtu < tp->t_maxseg + (sizeof (*th) + sizeof (*ip6))) { tcp_hc_updatemtu(&inc, mtu); tcp_mtudisc(inp, mtu); ICMP6STAT_INC(icp6s_pmtuchg); } } else inp = (*notify)(inp, inet6ctlerrmap[cmd]); } } unlock_inp: if (inp != NULL) INP_WUNLOCK(inp); } else { bzero(&inc, sizeof(inc)); inc.inc_fibnum = M_GETFIB(m); inc.inc_flags |= INC_ISIPV6; inc.inc_fport = th->th_dport; inc.inc_lport = th->th_sport; inc.inc6_faddr = *dst; inc.inc6_laddr = ip6->ip6_src; syncache_unreach(&inc, th); } INP_INFO_RUNLOCK(&V_tcbinfo); } #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_RLOCK_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) { tcp_mtudisc(inp, -1); return (inp); } static void 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; 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); tp->t_fb->tfb_tcp_output(tp); } #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. */ uint32_t tcp_maxmtu(struct in_conninfo *inc, struct tcp_ifcap *cap) { struct nhop4_extended nh4; struct ifnet *ifp; uint32_t maxmtu = 0; KASSERT(inc != NULL, ("tcp_maxmtu with NULL in_conninfo pointer")); if (inc->inc_faddr.s_addr != INADDR_ANY) { if (fib4_lookup_nh_ext(inc->inc_fibnum, inc->inc_faddr, NHR_REF, 0, &nh4) != 0) return (0); ifp = nh4.nh_ifp; maxmtu = nh4.nh_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; } } fib4_free_nh_ext(inc->inc_fibnum, &nh4); } return (maxmtu); } #endif /* INET */ #ifdef INET6 uint32_t tcp_maxmtu6(struct in_conninfo *inc, struct tcp_ifcap *cap) { struct nhop6_extended nh6; struct in6_addr dst6; uint32_t scopeid; struct ifnet *ifp; uint32_t maxmtu = 0; KASSERT(inc != NULL, ("tcp_maxmtu6 with NULL in_conninfo pointer")); if (!IN6_IS_ADDR_UNSPECIFIED(&inc->inc6_faddr)) { in6_splitscope(&inc->inc6_faddr, &dst6, &scopeid); if (fib6_lookup_nh_ext(inc->inc_fibnum, &dst6, scopeid, 0, 0, &nh6) != 0) return (0); ifp = nh6.nh_ifp; maxmtu = nh6.nh_mtu; /* 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; } } fib6_free_nh_ext(inc->inc_fibnum, &nh6); } return (maxmtu); } #endif /* INET6 */ /* * Calculate effective SMSS per RFC5681 definition for a given TCP * connection at its current state, taking into account SACK and etc. */ u_int tcp_maxseg(const struct tcpcb *tp) { u_int optlen; if (tp->t_flags & TF_NOOPT) return (tp->t_maxseg); /* * Here we have a simplified code from tcp_addoptions(), * without a proper loop, and having most of paddings hardcoded. * We might make mistakes with padding here in some edge cases, * but this is harmless, since result of tcp_maxseg() is used * only in cwnd and ssthresh estimations. */ #define PAD(len) ((((len) / 4) + !!((len) % 4)) * 4) if (TCPS_HAVEESTABLISHED(tp->t_state)) { if (tp->t_flags & TF_RCVD_TSTMP) optlen = TCPOLEN_TSTAMP_APPA; else optlen = 0; #ifdef TCP_SIGNATURE if (tp->t_flags & TF_SIGNATURE) optlen += PAD(TCPOLEN_SIGNATURE); #endif if ((tp->t_flags & TF_SACK_PERMIT) && tp->rcv_numsacks > 0) { optlen += TCPOLEN_SACKHDR; optlen += tp->rcv_numsacks * TCPOLEN_SACK; optlen = PAD(optlen); } } else { if (tp->t_flags & TF_REQ_TSTMP) optlen = TCPOLEN_TSTAMP_APPA; else optlen = PAD(TCPOLEN_MAXSEG); if (tp->t_flags & TF_REQ_SCALE) optlen += PAD(TCPOLEN_WINDOW); #ifdef TCP_SIGNATURE if (tp->t_flags & TF_SIGNATURE) optlen += PAD(TCPOLEN_SIGNATURE); #endif if (tp->t_flags & TF_SACK_PERMIT) optlen += PAD(TCPOLEN_SACK_PERMITTED); } #undef PAD optlen = min(optlen, TCP_MAXOLEN); return (tp->t_maxseg - optlen); } #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)); + 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): ip6 = mtod(m, struct ip6_hdr *); in6 = ip6->ip6_src; in6_clearscope(&in6); MD5Update(&ctx, (char *)&in6, sizeof(struct in6_addr)); in6 = ip6->ip6_dst; in6_clearscope(&in6); MD5Update(&ctx, (char *)&in6, sizeof(struct in6_addr)); plen = htonl(len + sizeof(struct tcphdr) + optlen); MD5Update(&ctx, (char *)&plen, sizeof(uint32_t)); nhdr = 0; MD5Update(&ctx, (char *)&nhdr, sizeof(uint8_t)); MD5Update(&ctx, (char *)&nhdr, sizeof(uint8_t)); MD5Update(&ctx, (char *)&nhdr, sizeof(uint8_t)); nhdr = IPPROTO_TCP; MD5Update(&ctx, (char *)&nhdr, sizeof(uint8_t)); th = (struct tcphdr *)((u_char *)ip6 + sizeof(struct ip6_hdr)); doff = sizeof(struct ip6_hdr) + sizeof(struct tcphdr) + optlen; break; #endif default: - KEY_FREESAV(&sav); + key_freesav(&sav); return (-1); /* NOTREACHED */ break; } /* * Step 2: Update MD5 hash with TCP header, excluding options. * The TCP checksum must be set to zero. */ savecsum = th->th_sum; th->th_sum = 0; MD5Update(&ctx, (char *)th, sizeof(struct tcphdr)); th->th_sum = savecsum; /* * Step 3: Update MD5 hash with TCP segment data. * Use m_apply() to avoid an early m_pullup(). */ if (len > 0) m_apply(m, doff, len, tcp_signature_apply, &ctx); /* * Step 4: Update MD5 hash with shared secret. */ MD5Update(&ctx, sav->key_auth->key_data, _KEYLEN(sav->key_auth)); MD5Final(buf, &ctx); key_sa_recordxfer(sav, m); - KEY_FREESAV(&sav); + 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_RLOCK(&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_RUNLOCK(&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 TCPSTATES_DEC(tp->t_state); TCPSTATES_INC(newstate); tp->t_state = newstate; TCP_PROBE6(state__change, NULL, tp, NULL, tp, NULL, pstate); } Index: projects/ipsec/sys/netipsec/ipsec.c =================================================================== --- projects/ipsec/sys/netipsec/ipsec.c (revision 309109) +++ projects/ipsec/sys/netipsec/ipsec.c (revision 309110) @@ -1,1792 +1,1757 @@ /* $FreeBSD$ */ /* $KAME: ipsec.c,v 1.103 2001/05/24 07:14:18 sakane Exp $ */ /*- * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. * 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. Neither the name of the project 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 PROJECT 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 PROJECT 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. */ /* * IPsec controller part. */ #include "opt_inet.h" #include "opt_inet6.h" #include "opt_ipsec.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef INET6 #include #endif #include #ifdef INET6 #include #endif #include #include #ifdef INET6 #include #endif #include #include #include /*XXX*/ #include #include #include #include #include #include #include #ifdef IPSEC_DEBUG VNET_DEFINE(int, ipsec_debug) = 1; #else VNET_DEFINE(int, ipsec_debug) = 0; #endif /* NB: name changed so netstat doesn't use it. */ VNET_PCPUSTAT_DEFINE(struct ipsecstat, ipsec4stat); VNET_PCPUSTAT_SYSINIT(ipsec4stat); #ifdef VIMAGE VNET_PCPUSTAT_SYSUNINIT(ipsec4stat); #endif /* VIMAGE */ VNET_DEFINE(int, ip4_ah_offsetmask) = 0; /* maybe IP_DF? */ /* DF bit on encap. 0: clear 1: set 2: copy */ VNET_DEFINE(int, ip4_ipsec_dfbit) = 0; VNET_DEFINE(int, ip4_esp_trans_deflev) = IPSEC_LEVEL_USE; VNET_DEFINE(int, ip4_esp_net_deflev) = IPSEC_LEVEL_USE; VNET_DEFINE(int, ip4_ah_trans_deflev) = IPSEC_LEVEL_USE; VNET_DEFINE(int, ip4_ah_net_deflev) = IPSEC_LEVEL_USE; /* ECN ignore(-1)/forbidden(0)/allowed(1) */ VNET_DEFINE(int, ip4_ipsec_ecn) = 0; VNET_DEFINE(int, ip4_esp_randpad) = -1; +static VNET_DEFINE(int, check_policy_history) = 0; +#define V_check_policy_history VNET(check_policy_history) static VNET_DEFINE(struct secpolicy, def_policy); #define V_def_policy VNET(def_policy) /* * Crypto support requirements: * * 1 require hardware support * -1 require software support * 0 take anything */ VNET_DEFINE(int, crypto_support) = CRYPTOCAP_F_HARDWARE | CRYPTOCAP_F_SOFTWARE; FEATURE(ipsec, "Internet Protocol Security (IPsec)"); #ifdef IPSEC_NAT_T FEATURE(ipsec_natt, "UDP Encapsulation of IPsec ESP Packets ('NAT-T')"); #endif SYSCTL_DECL(_net_inet_ipsec); /* net.inet.ipsec */ SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEF_POLICY, def_policy, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(def_policy).policy, 0, "IPsec default policy."); SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEF_ESP_TRANSLEV, esp_trans_deflev, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_esp_trans_deflev), 0, "Default ESP transport mode level"); SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEF_ESP_NETLEV, esp_net_deflev, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_esp_net_deflev), 0, "Default ESP tunnel mode level."); SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEF_AH_TRANSLEV, ah_trans_deflev, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_ah_trans_deflev), 0, "AH transfer mode default level."); SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEF_AH_NETLEV, ah_net_deflev, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_ah_net_deflev), 0, "AH tunnel mode default level."); SYSCTL_INT(_net_inet_ipsec, IPSECCTL_AH_CLEARTOS, ah_cleartos, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ah_cleartos), 0, "If set clear type-of-service field when doing AH computation."); SYSCTL_INT(_net_inet_ipsec, IPSECCTL_AH_OFFSETMASK, ah_offsetmask, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_ah_offsetmask), 0, "If not set clear offset field mask when doing AH computation."); SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DFBIT, dfbit, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_ipsec_dfbit), 0, "Do not fragment bit on encap."); SYSCTL_INT(_net_inet_ipsec, IPSECCTL_ECN, ecn, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_ipsec_ecn), 0, "Explicit Congestion Notification handling."); SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEBUG, debug, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_debug), 0, "Enable IPsec debugging output when set."); SYSCTL_INT(_net_inet_ipsec, OID_AUTO, crypto_support, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(crypto_support), 0, "Crypto driver selection."); SYSCTL_INT(_net_inet_ipsec, OID_AUTO, check_policy_history, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(check_policy_history), 0, "Use strict check of inbound packets to security policy compliance"); SYSCTL_VNET_PCPUSTAT(_net_inet_ipsec, OID_AUTO, ipsecstats, struct ipsecstat, ipsec4stat, "IPsec IPv4 statistics."); #ifdef REGRESSION /* * When set to 1, IPsec will send packets with the same sequence number. * This allows to verify if the other side has proper replay attacks detection. */ VNET_DEFINE(int, ipsec_replay) = 0; SYSCTL_INT(_net_inet_ipsec, OID_AUTO, test_replay, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_replay), 0, "Emulate replay attack"); /* * When set 1, IPsec will send packets with corrupted HMAC. * This allows to verify if the other side properly detects modified packets. */ VNET_DEFINE(int, ipsec_integrity) = 0; SYSCTL_INT(_net_inet_ipsec, OID_AUTO, test_integrity, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_integrity), 0, "Emulate man-in-the-middle attack"); #endif #ifdef INET6 VNET_PCPUSTAT_DEFINE(struct ipsecstat, ipsec6stat); VNET_PCPUSTAT_SYSINIT(ipsec6stat); #ifdef VIMAGE VNET_PCPUSTAT_SYSUNINIT(ipsec6stat); #endif /* VIMAGE */ VNET_DEFINE(int, ip6_esp_trans_deflev) = IPSEC_LEVEL_USE; VNET_DEFINE(int, ip6_esp_net_deflev) = IPSEC_LEVEL_USE; VNET_DEFINE(int, ip6_ah_trans_deflev) = IPSEC_LEVEL_USE; VNET_DEFINE(int, ip6_ah_net_deflev) = IPSEC_LEVEL_USE; VNET_DEFINE(int, ip6_ipsec_ecn) = 0; /* ECN ignore(-1)/forbidden(0)/allowed(1) */ SYSCTL_DECL(_net_inet6_ipsec6); /* net.inet6.ipsec6 */ SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEF_POLICY, def_policy, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(def_policy).policy, 0, "IPsec default policy."); SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEF_ESP_TRANSLEV, esp_trans_deflev, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_esp_trans_deflev), 0, "Default ESP transport mode level."); SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEF_ESP_NETLEV, esp_net_deflev, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_esp_net_deflev), 0, "Default ESP tunnel mode level."); SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEF_AH_TRANSLEV, ah_trans_deflev, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_ah_trans_deflev), 0, "AH transfer mode default level."); SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEF_AH_NETLEV, ah_net_deflev, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_ah_net_deflev), 0, "AH tunnel mode default level."); SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_ECN, ecn, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_ipsec_ecn), 0, "Explicit Congestion Notification handling."); SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEBUG, debug, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_debug), 0, "Enable IPsec debugging output when set."); SYSCTL_VNET_PCPUSTAT(_net_inet6_ipsec6, IPSECCTL_STATS, ipsecstats, struct ipsecstat, ipsec6stat, "IPsec IPv6 statistics."); #endif /* INET6 */ static int ipsec_in_reject(struct secpolicy *, struct inpcb *, const struct mbuf *); static void ipsec_setspidx_inpcb(struct inpcb *, struct secpolicyindex *); static void ipsec4_get_ulp(const struct mbuf *m, struct secpolicyindex *, int); static void ipsec4_setspidx_ipaddr(const struct mbuf *, struct secpolicyindex *); #ifdef INET6 static void ipsec6_get_ulp(const struct mbuf *m, struct secpolicyindex *, int); static void ipsec6_setspidx_ipaddr(const struct mbuf *, struct secpolicyindex *); #endif static void vshiftl(unsigned char *, int, int); MALLOC_DEFINE(M_IPSEC_INPCB, "inpcbpolicy", "inpcb-resident ipsec policy"); /* * Return a held reference to the default SP. */ static struct secpolicy * key_allocsp_default(void) { struct secpolicy *sp; sp = &V_def_policy; if (sp->policy != IPSEC_POLICY_DISCARD && sp->policy != IPSEC_POLICY_NONE) { ipseclog((LOG_INFO, "fixed system default policy: %d->%d\n", sp->policy, IPSEC_POLICY_NONE)); sp->policy = IPSEC_POLICY_NONE; } key_addref(sp); return (sp); } static struct secpolicy * ipsec_checkpolicy(struct secpolicy *sp, struct inpcb *inp, int *error) { uint32_t genid; if (inp != NULL && (inp->inp_sp->flags & INP_OUTBOUND_POLICY) == 0 && inp->inp_sp->sp_out == NULL) { /* * Save found OUTBOUND policy into PCB SP cache. */ genid = key_getspgen(); inp->inp_sp->sp_out = sp; if (genid != inp->inp_sp->genid) { /* Reset INBOUND cached policy if genid is changed */ if ((inp->inp_sp->flags & INP_INBOUND_POLICY) == 0) inp->inp_sp->sp_in = NULL; inp->inp_sp->genid = genid; } KEYDBG(IPSEC_STAMP, printf("%s: PCB(%p): cached SP(%p)\n", __func__, inp, sp)); } switch (sp->policy) { default: printf("%s: invalid policy %u\n", __func__, sp->policy); /* FALLTHROUGH */ case IPSEC_POLICY_DISCARD: *error = -EINVAL; /* Packet is discarded by caller. */ /* FALLTHROUGH */ case IPSEC_POLICY_BYPASS: case IPSEC_POLICY_NONE: key_freesp(&sp); sp = NULL; /* NB: force NULL result. */ break; case IPSEC_POLICY_IPSEC: break; } KEYDBG(IPSEC_DUMP, printf("%s: get SP(%p), error %d\n", __func__, sp, *error)); return (sp); } static struct secpolicy * ipsec_getpcbpolicy(struct inpcb *inp, u_int dir) { struct secpolicy *sp; int flags; if (inp == NULL || inp->inp_sp == NULL) return (NULL); flags = inp->inp_sp->flags; if (dir == IPSEC_DIR_OUTBOUND) { sp = inp->inp_sp->sp_out; flags &= INP_OUTBOUND_POLICY; } else { sp = inp->inp_sp->sp_in; flags &= INP_INBOUND_POLICY; } /* * Check flags. If we have PCB SP, just return it. * Otherwise we need to check that cached SP entry isn't stale. */ if (flags == 0) { if (sp == NULL) return (NULL); if (inp->inp_sp->genid != key_getspgen()) { /* * Invalidate the cache. * Do not touch policy if it was set by PCB. */ if ((inp->inp_sp->flags & INP_INBOUND_POLICY) == 0) inp->inp_sp->sp_in = NULL; if ((inp->inp_sp->flags & INP_OUTBOUND_POLICY) == 0) inp->inp_sp->sp_out = NULL; return (NULL); } KEYDBG(IPSEC_STAMP, printf("%s: PCB(%p): cache hit SP(%p)\n", __func__, inp, sp)); /* Return referenced cached policy */ } IPSEC_ASSERT(sp != NULL, ("null SP, but flags is 0x%04x", flags)); key_addref(sp); return (sp); } static void ipsec_setspidx_inpcb(struct inpcb *inp, struct secpolicyindex *spidx) { #ifdef INET6 if (inp->inp_vflag & INP_IPV6) { bzero(&spidx->src.sin6, sizeof(spidx->src.sin6)); spidx->src.sin6.sin6_family = AF_INET6; spidx->src.sin6.sin6_len = sizeof(struct sockaddr_in6); spidx->src.sin6.sin6_addr = inp->in6p_laddr; spidx->src.sin6.sin6_port = inp->inp_lport; if (IN6_IS_SCOPE_LINKLOCAL(&inp->in6p_laddr)) { /* XXXAE: use in6p_zoneid */ spidx->src.sin6.sin6_addr.s6_addr16[1] = 0; spidx->src.sin6.sin6_scope_id = ntohs( inp->in6p_laddr.s6_addr16[1]); } spidx->prefs = sizeof(struct in6_addr) << 3; bzero(&spidx->dst.sin6, sizeof(spidx->dst.sin6)); spidx->dst.sin6.sin6_family = AF_INET6; spidx->dst.sin6.sin6_len = sizeof(struct sockaddr_in6); spidx->dst.sin6.sin6_addr = inp->in6p_faddr; spidx->dst.sin6.sin6_port = inp->inp_fport; if (IN6_IS_SCOPE_LINKLOCAL(&inp->in6p_faddr)) { /* XXXAE: use in6p_zoneid */ spidx->dst.sin6.sin6_addr.s6_addr16[1] = 0; spidx->dst.sin6.sin6_scope_id = ntohs( inp->in6p_faddr.s6_addr16[1]); } spidx->prefd = sizeof(struct in6_addr) << 3; } #endif #ifdef INET if (inp->inp_vflag & INP_IPV4) { bzero(&spidx->src.sin, sizeof(spidx->src.sin)); spidx->src.sin.sin_family = AF_INET; spidx->src.sin.sin_len = sizeof(struct sockaddr_in); spidx->src.sin.sin_addr = inp->inp_laddr; spidx->src.sin.sin_port = inp->inp_lport; spidx->prefs = sizeof(struct in_addr) << 3; bzero(&spidx->dst.sin, sizeof(spidx->dst.sin)); spidx->dst.sin.sin_family = AF_INET; spidx->dst.sin.sin_len = sizeof(struct sockaddr_in); spidx->dst.sin.sin_addr = inp->inp_faddr; spidx->dst.sin.sin_port = inp->inp_fport; spidx->prefd = sizeof(struct in_addr) << 3; } #endif spidx->ul_proto = inp->inp_ip_p; KEYDBG(IPSEC_DUMP, printf("%s: ", __func__); kdebug_secpolicyindex(spidx, NULL)); } #ifdef INET static void ipsec4_get_ulp(const struct mbuf *m, struct secpolicyindex *spidx, int needport) { uint8_t nxt; int off; /* Sanity check. */ IPSEC_ASSERT(m->m_pkthdr.len >= sizeof(struct ip), ("packet too short")); if (m->m_len >= sizeof (struct ip)) { const struct ip *ip = mtod(m, const struct ip *); if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) goto done; off = ip->ip_hl << 2; nxt = ip->ip_p; } else { struct ip ih; m_copydata(m, 0, sizeof (struct ip), (caddr_t) &ih); if (ih.ip_off & htons(IP_MF | IP_OFFMASK)) goto done; off = ih.ip_hl << 2; nxt = ih.ip_p; } while (off < m->m_pkthdr.len) { struct ip6_ext ip6e; struct tcphdr th; struct udphdr uh; switch (nxt) { case IPPROTO_TCP: spidx->ul_proto = nxt; if (!needport) goto done_proto; if (off + sizeof(struct tcphdr) > m->m_pkthdr.len) goto done; m_copydata(m, off, sizeof (th), (caddr_t) &th); spidx->src.sin.sin_port = th.th_sport; spidx->dst.sin.sin_port = th.th_dport; return; case IPPROTO_UDP: spidx->ul_proto = nxt; if (!needport) goto done_proto; if (off + sizeof(struct udphdr) > m->m_pkthdr.len) goto done; m_copydata(m, off, sizeof (uh), (caddr_t) &uh); spidx->src.sin.sin_port = uh.uh_sport; spidx->dst.sin.sin_port = uh.uh_dport; return; case IPPROTO_AH: if (off + sizeof(ip6e) > m->m_pkthdr.len) goto done; /* XXX Sigh, this works but is totally bogus. */ m_copydata(m, off, sizeof(ip6e), (caddr_t) &ip6e); off += (ip6e.ip6e_len + 2) << 2; nxt = ip6e.ip6e_nxt; break; case IPPROTO_ICMP: default: /* XXX Intermediate headers??? */ spidx->ul_proto = nxt; goto done_proto; } } done: spidx->ul_proto = IPSEC_ULPROTO_ANY; done_proto: spidx->src.sin.sin_port = IPSEC_PORT_ANY; spidx->dst.sin.sin_port = IPSEC_PORT_ANY; KEYDBG(IPSEC_DUMP, printf("%s: ", __func__); kdebug_secpolicyindex(spidx, NULL)); } /* Assumes that m is sane. */ static void ipsec4_setspidx_ipaddr(const struct mbuf *m, struct secpolicyindex *spidx) { static const struct sockaddr_in template = { sizeof (struct sockaddr_in), AF_INET, 0, { 0 }, { 0, 0, 0, 0, 0, 0, 0, 0 } }; spidx->src.sin = template; spidx->dst.sin = template; if (m->m_len < sizeof (struct ip)) { m_copydata(m, offsetof(struct ip, ip_src), sizeof (struct in_addr), (caddr_t) &spidx->src.sin.sin_addr); m_copydata(m, offsetof(struct ip, ip_dst), sizeof (struct in_addr), (caddr_t) &spidx->dst.sin.sin_addr); } else { const struct ip *ip = mtod(m, const struct ip *); spidx->src.sin.sin_addr = ip->ip_src; spidx->dst.sin.sin_addr = ip->ip_dst; } spidx->prefs = sizeof(struct in_addr) << 3; spidx->prefd = sizeof(struct in_addr) << 3; } static struct secpolicy * ipsec4_getpolicy(const struct mbuf *m, struct inpcb *inp, u_int dir) { struct secpolicyindex spidx; struct secpolicy *sp; sp = ipsec_getpcbpolicy(inp, dir); if (sp == NULL && key_havesp(dir)) { /* Make an index to look for a policy. */ ipsec4_setspidx_ipaddr(m, &spidx); /* Fill ports in spidx if we have inpcb. */ ipsec4_get_ulp(m, &spidx, inp != NULL); spidx.dir = dir; sp = key_allocsp(&spidx, dir); } if (sp == NULL) /* No SP found, use system default. */ sp = key_allocsp_default(); return (sp); } /* * Check security policy for *OUTBOUND* IPv4 packet. */ struct secpolicy * ipsec4_checkpolicy(const struct mbuf *m, struct inpcb *inp, int *error) { struct secpolicy *sp; *error = 0; sp = ipsec4_getpolicy(m, inp, IPSEC_DIR_OUTBOUND); if (sp != NULL) sp = ipsec_checkpolicy(sp, inp, error); if (sp == NULL) { switch (*error) { case 0: /* No IPsec required: BYPASS or NONE */ break; case -EINVAL: IPSECSTAT_INC(ips_out_polvio); break; default: IPSECSTAT_INC(ips_out_inval); } } KEYDBG(IPSEC_STAMP, printf("%s: using SP(%p), error %d\n", __func__, sp, *error)); if (sp != NULL) KEYDBG(IPSEC_DATA, kdebug_secpolicy(sp)); return (sp); } /* * Check IPv4 packet against *INBOUND* security policy. * This function is called from tcp_input(), udp_input(), * rip_input() and sctp_input(). */ int ipsec4_in_reject(const struct mbuf *m, struct inpcb *inp) { struct secpolicy *sp; int result; sp = ipsec4_getpolicy(m, inp, IPSEC_DIR_INBOUND); result = ipsec_in_reject(sp, inp, m); key_freesp(&sp); if (result != 0) IPSECSTAT_INC(ips_in_polvio); return (result); } #endif /* INET */ #ifdef INET6 static void ipsec6_get_ulp(const struct mbuf *m, struct secpolicyindex *spidx, int needport) { struct tcphdr th; struct udphdr uh; struct icmp6_hdr ih; int off, nxt; IPSEC_ASSERT(m->m_pkthdr.len >= sizeof(struct ip6_hdr), ("packet too short")); /* Set default. */ spidx->ul_proto = IPSEC_ULPROTO_ANY; spidx->src.sin6.sin6_port = IPSEC_PORT_ANY; spidx->dst.sin6.sin6_port = IPSEC_PORT_ANY; nxt = -1; off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt); if (off < 0 || m->m_pkthdr.len < off) return; switch (nxt) { case IPPROTO_TCP: spidx->ul_proto = nxt; if (!needport) break; if (off + sizeof(struct tcphdr) > m->m_pkthdr.len) break; m_copydata(m, off, sizeof(th), (caddr_t)&th); spidx->src.sin6.sin6_port = th.th_sport; spidx->dst.sin6.sin6_port = th.th_dport; break; case IPPROTO_UDP: spidx->ul_proto = nxt; if (!needport) break; if (off + sizeof(struct udphdr) > m->m_pkthdr.len) break; m_copydata(m, off, sizeof(uh), (caddr_t)&uh); spidx->src.sin6.sin6_port = uh.uh_sport; spidx->dst.sin6.sin6_port = uh.uh_dport; break; case IPPROTO_ICMPV6: spidx->ul_proto = nxt; if (off + sizeof(struct icmp6_hdr) > m->m_pkthdr.len) break; m_copydata(m, off, sizeof(ih), (caddr_t)&ih); spidx->src.sin6.sin6_port = htons((uint16_t)ih.icmp6_type); spidx->dst.sin6.sin6_port = htons((uint16_t)ih.icmp6_code); break; default: /* XXX Intermediate headers??? */ spidx->ul_proto = nxt; break; } KEYDBG(IPSEC_DUMP, printf("%s: ", __func__); kdebug_secpolicyindex(spidx, NULL)); } /* Assumes that m is sane. */ static void ipsec6_setspidx_ipaddr(const struct mbuf *m, struct secpolicyindex *spidx) { struct ip6_hdr ip6buf; const struct ip6_hdr *ip6 = NULL; struct sockaddr_in6 *sin6; if (m->m_len >= sizeof(*ip6)) ip6 = mtod(m, const struct ip6_hdr *); else { m_copydata(m, 0, sizeof(ip6buf), (caddr_t)&ip6buf); ip6 = &ip6buf; } sin6 = (struct sockaddr_in6 *)&spidx->src; bzero(sin6, sizeof(*sin6)); sin6->sin6_family = AF_INET6; sin6->sin6_len = sizeof(struct sockaddr_in6); bcopy(&ip6->ip6_src, &sin6->sin6_addr, sizeof(ip6->ip6_src)); if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) { sin6->sin6_addr.s6_addr16[1] = 0; sin6->sin6_scope_id = ntohs(ip6->ip6_src.s6_addr16[1]); } spidx->prefs = sizeof(struct in6_addr) << 3; sin6 = (struct sockaddr_in6 *)&spidx->dst; bzero(sin6, sizeof(*sin6)); sin6->sin6_family = AF_INET6; sin6->sin6_len = sizeof(struct sockaddr_in6); bcopy(&ip6->ip6_dst, &sin6->sin6_addr, sizeof(ip6->ip6_dst)); if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) { sin6->sin6_addr.s6_addr16[1] = 0; sin6->sin6_scope_id = ntohs(ip6->ip6_dst.s6_addr16[1]); } spidx->prefd = sizeof(struct in6_addr) << 3; } static struct secpolicy * ipsec6_getpolicy(const struct mbuf *m, struct inpcb *inp, u_int dir) { struct secpolicyindex spidx; struct secpolicy *sp; sp = ipsec_getpcbpolicy(inp, dir); if (sp == NULL && key_havesp(dir)) { /* Make an index to look for a policy. */ ipsec6_setspidx_ipaddr(m, &spidx); /* Fill ports in spidx if we have inpcb. */ ipsec6_get_ulp(m, &spidx, inp != NULL); spidx.dir = dir; sp = key_allocsp(&spidx, dir); } if (sp == NULL) /* No SP found, use system default. */ sp = key_allocsp_default(); return (sp); } /* * Check security policy for *OUTBOUND* IPv6 packet. */ struct secpolicy * ipsec6_checkpolicy(const struct mbuf *m, struct inpcb *inp, int *error) { struct secpolicy *sp; *error = 0; sp = ipsec6_getpolicy(m, inp, IPSEC_DIR_OUTBOUND); if (sp != NULL) sp = ipsec_checkpolicy(sp, inp, error); if (sp == NULL) { switch (*error) { case 0: /* No IPsec required: BYPASS or NONE */ break; case -EINVAL: IPSEC6STAT_INC(ips_out_polvio); break; default: IPSEC6STAT_INC(ips_out_inval); } } KEYDBG(IPSEC_STAMP, printf("%s: using SP(%p), error %d\n", __func__, sp, *error)); if (sp != NULL) KEYDBG(IPSEC_DATA, kdebug_secpolicy(sp)); return (sp); } /* * Check IPv6 packet against inbound security policy. * This function is called from tcp6_input(), udp6_input(), * rip6_input() and sctp_input(). */ int ipsec6_in_reject(const struct mbuf *m, struct inpcb *inp) { struct secpolicy *sp; int result; sp = ipsec6_getpolicy(m, inp, IPSEC_DIR_INBOUND); result = ipsec_in_reject(sp, inp, m); key_freesp(&sp); if (result) IPSEC6STAT_INC(ips_in_polvio); return (result); } #endif int ipsec_run_hhooks(struct ipsec_ctx_data *ctx, int type) { int idx; switch (ctx->af) { #ifdef INET case AF_INET: idx = HHOOK_IPSEC_INET; break; #endif #ifdef INET6 case AF_INET6: idx = HHOOK_IPSEC_INET6; break; #endif default: return (EPFNOSUPPORT); } if (type == HHOOK_TYPE_IPSEC_IN) HHOOKS_RUN_IF(V_ipsec_hhh_in[idx], ctx, NULL); else HHOOKS_RUN_IF(V_ipsec_hhh_out[idx], ctx, NULL); if (*ctx->mp == NULL) return (EACCES); return (0); } /* Initialize PCB policy. */ int ipsec_init_pcbpolicy(struct inpcb *inp) { IPSEC_ASSERT(inp != NULL, ("null inp")); IPSEC_ASSERT(inp->inp_sp == NULL, ("inp_sp already initialized")); inp->inp_sp = malloc(sizeof(struct inpcbpolicy), M_IPSEC_INPCB, M_NOWAIT | M_ZERO); if (inp->inp_sp == NULL) { ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__)); return (ENOBUFS); } return (0); } /* Delete PCB policy. */ int ipsec_delete_pcbpolicy(struct inpcb *inp) { if (inp->inp_sp == NULL) return (0); if (inp->inp_sp->flags & INP_INBOUND_POLICY) key_freesp(&inp->inp_sp->sp_in); if (inp->inp_sp->flags & INP_OUTBOUND_POLICY) key_freesp(&inp->inp_sp->sp_out); free(inp->inp_sp, M_IPSEC_INPCB); inp->inp_sp = NULL; return (0); } /* Deep-copy a policy in PCB. */ static struct secpolicy * ipsec_deepcopy_pcbpolicy(struct secpolicy *src) { struct secpolicy *dst; int i; if (src == NULL) return (NULL); IPSEC_ASSERT(src->state == IPSEC_SPSTATE_PCB, ("SP isn't PCB")); dst = key_newsp(); if (dst == NULL) return (NULL); dst->policy = src->policy; dst->state = src->state; dst->priority = src->priority; /* Do not touch the refcnt field. */ /* Copy IPsec request chain. */ for (i = 0; i < src->tcount; i++) { dst->req[i] = ipsec_newisr(); if (dst->req[i] == NULL) { key_freesp(&dst); return (NULL); } bcopy(src->req[i], dst->req[i], sizeof(struct ipsecrequest)); dst->tcount++; } KEYDBG(IPSEC_DUMP, printf("%s: copied SP(%p) -> SP(%p)\n", __func__, src, dst); kdebug_secpolicy(dst)); return (dst); } /* Copy old IPsec policy into new. */ int ipsec_copy_pcbpolicy(struct inpcb *old, struct inpcb *new) { struct secpolicy *sp; /* * old->inp_sp can be NULL if PCB was created when an IPsec * support was unavailable. This is not an error, we don't have * policies in this PCB, so nothing to copy. */ if (old->inp_sp == NULL) return (0); IPSEC_ASSERT(new->inp_sp != NULL, ("new inp_sp is NULL")); INP_WLOCK_ASSERT(new); if (old->inp_sp->flags & INP_INBOUND_POLICY) { sp = ipsec_deepcopy_pcbpolicy(old->inp_sp->sp_in); if (sp == NULL) return (ENOBUFS); } else sp = NULL; if (new->inp_sp->flags & INP_INBOUND_POLICY) key_freesp(&new->inp_sp->sp_in); new->inp_sp->sp_in = sp; if (sp != NULL) new->inp_sp->flags |= INP_INBOUND_POLICY; else new->inp_sp->flags &= ~INP_INBOUND_POLICY; if (old->inp_sp->flags & INP_OUTBOUND_POLICY) { sp = ipsec_deepcopy_pcbpolicy(old->inp_sp->sp_out); if (sp == NULL) return (ENOBUFS); } else sp = NULL; if (new->inp_sp->flags & INP_OUTBOUND_POLICY) key_freesp(&new->inp_sp->sp_out); new->inp_sp->sp_out = sp; if (sp != NULL) new->inp_sp->flags |= INP_OUTBOUND_POLICY; else new->inp_sp->flags &= ~INP_OUTBOUND_POLICY; return (0); } static int ipsec_set_pcbpolicy(struct inpcb *inp, struct ucred *cred, void *request, size_t len) { struct sadb_x_policy *xpl; struct secpolicy **spp, *newsp; int error, flags; xpl = (struct sadb_x_policy *)request; /* Select direction. */ switch (xpl->sadb_x_policy_dir) { case IPSEC_DIR_INBOUND: spp = &inp->inp_sp->sp_in; flags = INP_INBOUND_POLICY; break; case IPSEC_DIR_OUTBOUND: spp = &inp->inp_sp->sp_out; flags = INP_OUTBOUND_POLICY; break; default: ipseclog((LOG_ERR, "%s: invalid direction=%u\n", __func__, xpl->sadb_x_policy_dir)); return (EINVAL); } /* * Privileged sockets are allowed to set own security policy * and configure IPsec bypass. Unprivileged sockets only can * have ENTRUST policy. */ switch (xpl->sadb_x_policy_type) { case IPSEC_POLICY_IPSEC: case IPSEC_POLICY_BYPASS: if (cred != NULL && priv_check_cred(cred, PRIV_NETINET_IPSEC, 0) != 0) return (EACCES); /* Allocate new SP entry. */ newsp = key_msg2sp(xpl, len, &error); if (newsp == NULL) return (error); newsp->state = IPSEC_SPSTATE_PCB; break; case IPSEC_POLICY_ENTRUST: /* We just use NULL pointer for ENTRUST policy */ newsp = NULL; break; default: /* Other security policy types aren't allowed for PCB */ return (EINVAL); } /* Clear old SP and set new SP. */ if (*spp != NULL) key_freesp(spp); *spp = newsp; KEYDBG(IPSEC_DUMP, printf("%s: new SP(%p)\n", __func__, newsp)); if (newsp == NULL) inp->inp_sp->flags &= ~flags; else { inp->inp_sp->flags |= flags; KEYDBG(IPSEC_DUMP, kdebug_secpolicy(newsp)); } return (0); } static int ipsec_get_pcbpolicy(struct inpcb *inp, void *request, size_t *len) { struct sadb_x_policy *xpl; struct secpolicy *sp; int error, flags; xpl = (struct sadb_x_policy *)request; flags = inp->inp_sp->flags; /* Select direction. */ switch (xpl->sadb_x_policy_dir) { case IPSEC_DIR_INBOUND: sp = inp->inp_sp->sp_in; flags &= INP_INBOUND_POLICY; break; case IPSEC_DIR_OUTBOUND: sp = inp->inp_sp->sp_out; flags &= INP_OUTBOUND_POLICY; break; default: ipseclog((LOG_ERR, "%s: invalid direction=%u\n", __func__, xpl->sadb_x_policy_dir)); return (EINVAL); } if (flags == 0) { /* Return ENTRUST policy */ xpl->sadb_x_policy_exttype = SADB_X_EXT_POLICY; xpl->sadb_x_policy_type = IPSEC_POLICY_ENTRUST; xpl->sadb_x_policy_id = 0; xpl->sadb_x_policy_priority = 0; xpl->sadb_x_policy_len = PFKEY_UNIT64(sizeof(*xpl)); *len = sizeof(*xpl); return (0); } IPSEC_ASSERT(sp != NULL, ("sp is NULL, but flags is 0x%04x", inp->inp_sp->flags)); key_addref(sp); error = key_sp2msg(sp, request, len); key_freesp(&sp); if (error == EINVAL) return (error); /* * We return "success", but user should check *len. * *len will be set to size of valid data and * sadb_x_policy_len will contain needed size. */ return (0); } /* Handle socket option control request for PCB */ int ipsec_control_pcbpolicy(struct inpcb *inp, struct sockopt *sopt) { void *optdata; size_t optlen; int error; if (inp->inp_sp == NULL) return (ENOPROTOOPT); /* Limit maximum request size to PAGE_SIZE */ optlen = sopt->sopt_valsize; if (optlen < sizeof(struct sadb_x_policy) || optlen > PAGE_SIZE) return (EINVAL); optdata = malloc(optlen, M_TEMP, sopt->sopt_td ? M_WAITOK: M_NOWAIT); if (optdata == NULL) return (ENOBUFS); /* * We need a hint from the user, what policy is requested - input * or output? User should specify it in the buffer, even for * setsockopt(). */ error = sooptcopyin(sopt, optdata, optlen, optlen); if (error == 0) { if (sopt->sopt_dir == SOPT_SET) error = ipsec_set_pcbpolicy(inp, sopt->sopt_td ? sopt->sopt_td->td_ucred: NULL, optdata, optlen); else { error = ipsec_get_pcbpolicy(inp, optdata, &optlen); if (error == 0) error = sooptcopyout(sopt, optdata, optlen); } } free(optdata, M_TEMP); return (error); } struct ipsecrequest * ipsec_newisr(void) { return (malloc(sizeof(struct ipsecrequest), M_IPSEC_SR, M_NOWAIT | M_ZERO)); } void ipsec_delisr(struct ipsecrequest *p) { free(p, M_IPSEC_SR); } /* * Return current level. * Either IPSEC_LEVEL_USE or IPSEC_LEVEL_REQUIRE are always returned. */ u_int ipsec_get_reqlevel(struct secpolicy *sp, u_int idx) { struct ipsecrequest *isr; u_int esp_trans_deflev, esp_net_deflev; u_int ah_trans_deflev, ah_net_deflev; u_int level = 0; IPSEC_ASSERT(idx < sp->tcount, ("Wrong IPsec request index %d", idx)); /* XXX Note that we have ipseclog() expanded here - code sync issue. */ #define IPSEC_CHECK_DEFAULT(lev) \ (((lev) != IPSEC_LEVEL_USE && (lev) != IPSEC_LEVEL_REQUIRE && \ (lev) != IPSEC_LEVEL_UNIQUE) \ ? (V_ipsec_debug ? \ log(LOG_INFO, "fixed system default level " #lev ":%d->%d\n",\ (lev), IPSEC_LEVEL_REQUIRE) : 0), \ (lev) = IPSEC_LEVEL_REQUIRE, (lev) : (lev)) /* * IPsec VTI uses unique security policy with fake spidx filled * with zeroes. Just return IPSEC_LEVEL_REQUIRE instead of doing * full level lookup for such policies. */ if (sp->state == IPSEC_SPSTATE_IFNET) { IPSEC_ASSERT(sp->req[idx]->level == IPSEC_LEVEL_UNIQUE, ("Wrong IPsec request level %d", sp->req[idx]->level)); return (IPSEC_LEVEL_REQUIRE); } /* Set default level. */ switch (sp->spidx.src.sa.sa_family) { #ifdef INET case AF_INET: esp_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip4_esp_trans_deflev); esp_net_deflev = IPSEC_CHECK_DEFAULT(V_ip4_esp_net_deflev); ah_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip4_ah_trans_deflev); ah_net_deflev = IPSEC_CHECK_DEFAULT(V_ip4_ah_net_deflev); break; #endif #ifdef INET6 case AF_INET6: esp_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip6_esp_trans_deflev); esp_net_deflev = IPSEC_CHECK_DEFAULT(V_ip6_esp_net_deflev); ah_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip6_ah_trans_deflev); ah_net_deflev = IPSEC_CHECK_DEFAULT(V_ip6_ah_net_deflev); break; #endif /* INET6 */ default: panic("%s: unknown af %u", __func__, sp->spidx.src.sa.sa_family); } #undef IPSEC_CHECK_DEFAULT isr = sp->req[idx]; /* Set level. */ switch (isr->level) { case IPSEC_LEVEL_DEFAULT: switch (isr->saidx.proto) { case IPPROTO_ESP: if (isr->saidx.mode == IPSEC_MODE_TUNNEL) level = esp_net_deflev; else level = esp_trans_deflev; break; case IPPROTO_AH: if (isr->saidx.mode == IPSEC_MODE_TUNNEL) level = ah_net_deflev; else level = ah_trans_deflev; break; case IPPROTO_IPCOMP: /* * We don't really care, as IPcomp document says that * we shouldn't compress small packets. */ level = IPSEC_LEVEL_USE; break; default: panic("%s: Illegal protocol defined %u\n", __func__, isr->saidx.proto); } break; case IPSEC_LEVEL_USE: case IPSEC_LEVEL_REQUIRE: level = isr->level; break; case IPSEC_LEVEL_UNIQUE: level = IPSEC_LEVEL_REQUIRE; break; default: panic("%s: Illegal IPsec level %u\n", __func__, isr->level); } return (level); } static int ipsec_check_history(const struct mbuf *m, struct secpolicy *sp, u_int idx) { struct xform_history *xh; struct m_tag *mtag; mtag = NULL; while ((mtag = m_tag_find(__DECONST(struct mbuf *, m), PACKET_TAG_IPSEC_IN_DONE, mtag)) != NULL) { xh = (struct xform_history *)(mtag + 1); KEYDBG(IPSEC_DATA, char buf[IPSEC_ADDRSTRLEN]; printf("%s: mode %s proto %u dst %s\n", __func__, kdebug_secasindex_mode(xh->mode), xh->proto, ipsec_address(&xh->dst, buf, sizeof(buf)))); if (xh->proto != sp->req[idx]->saidx.proto) continue; /* If SA had IPSEC_MODE_ANY, consider this as match. */ if (xh->mode != sp->req[idx]->saidx.mode && xh->mode != IPSEC_MODE_ANY) continue; /* * For transport mode IPsec request doesn't contain * addresses. We need to use address from spidx. */ if (sp->req[idx]->saidx.mode == IPSEC_MODE_TRANSPORT) { if (key_sockaddrcmp_withmask(&xh->dst.sa, &sp->spidx.dst.sa, sp->spidx.prefd) != 0) continue; } else { if (key_sockaddrcmp(&xh->dst.sa, &sp->req[idx]->saidx.dst.sa, 0) != 0) continue; } return (0); /* matched */ } return (1); } /* * Check security policy requirements against the actual * packet contents. Return one if the packet should be * reject as "invalid"; otherwiser return zero to have the * packet treated as "valid". * * OUT: * 0: valid * 1: invalid */ static int ipsec_in_reject(struct secpolicy *sp, struct inpcb *inp, const struct mbuf *m) { uint32_t genid; int i; KEYDBG(IPSEC_STAMP, printf("%s: PCB(%p): using SP(%p)\n", __func__, inp, sp)); KEYDBG(IPSEC_DATA, kdebug_secpolicy(sp)); if (inp != NULL && (inp->inp_sp->flags & INP_INBOUND_POLICY) == 0 && inp->inp_sp->sp_in == NULL) { /* * Save found INBOUND policy into PCB SP cache. */ genid = key_getspgen(); inp->inp_sp->sp_in = sp; if (genid != inp->inp_sp->genid) { /* Reset OUTBOUND cached policy if genid is changed */ if ((inp->inp_sp->flags & INP_OUTBOUND_POLICY) == 0) inp->inp_sp->sp_out = NULL; inp->inp_sp->genid = genid; } KEYDBG(IPSEC_STAMP, printf("%s: PCB(%p): cached SP(%p)\n", __func__, inp, sp)); } /* Check policy. */ switch (sp->policy) { case IPSEC_POLICY_DISCARD: return (1); case IPSEC_POLICY_BYPASS: case IPSEC_POLICY_NONE: return (0); } IPSEC_ASSERT(sp->policy == IPSEC_POLICY_IPSEC, ("invalid policy %u", sp->policy)); /* * ipsec[46]_common_input_cb after each transform adds * PACKET_TAG_IPSEC_IN_DONE mbuf tag. It contains SPI, proto, mode * and destination address from saidx. We can compare info from * these tags with requirements in SP. */ for (i = 0; i < sp->tcount; i++) { /* * Do not check IPcomp, since IPcomp document * says that we shouldn't compress small packets. * IPComp policy should always be treated as being * in "use" level. */ if (sp->req[i]->saidx.proto == IPPROTO_IPCOMP || ipsec_get_reqlevel(sp, i) != IPSEC_LEVEL_REQUIRE) continue; if (V_check_policy_history != 0 && ipsec_check_history(m, sp, i) != 0) return (1); else switch (sp->req[i]->saidx.proto) { case IPPROTO_ESP: if ((m->m_flags & M_DECRYPTED) == 0) { KEYDBG(IPSEC_DUMP, printf("%s: ESP m_flags:%x\n", __func__, m->m_flags)); return (1); } break; case IPPROTO_AH: if ((m->m_flags & M_AUTHIPHDR) == 0) { KEYDBG(IPSEC_DUMP, printf("%s: AH m_flags:%x\n", __func__, m->m_flags)); return (1); } break; } } return (0); /* Valid. */ } /* * Compute the byte size to be occupied by IPsec header. * In case it is tunnelled, it includes the size of outer IP header. */ static size_t ipsec_hdrsiz_internal(struct secpolicy *sp) { size_t size; int i; KEYDBG(IPSEC_STAMP, printf("%s: using SP(%p)\n", __func__, sp)); KEYDBG(IPSEC_DATA, kdebug_secpolicy(sp)); switch (sp->policy) { case IPSEC_POLICY_DISCARD: case IPSEC_POLICY_BYPASS: case IPSEC_POLICY_NONE: return (0); } IPSEC_ASSERT(sp->policy == IPSEC_POLICY_IPSEC, ("invalid policy %u", sp->policy)); /* * XXX: for each transform we need to lookup suitable SA * and use info from SA to calculate headers size. * XXX: for NAT-T we need to cosider UDP header size. */ size = 0; for (i = 0; i < sp->tcount; i++) { switch (sp->req[i]->saidx.proto) { case IPPROTO_ESP: size += esp_hdrsiz(NULL); break; case IPPROTO_AH: size += ah_hdrsiz(NULL); break; case IPPROTO_IPCOMP: size += sizeof(struct ipcomp); break; } if (sp->req[i]->saidx.mode == IPSEC_MODE_TUNNEL) { switch (sp->req[i]->saidx.dst.sa.sa_family) { #ifdef INET case AF_INET: size += sizeof(struct ip); break; #endif #ifdef INET6 case AF_INET6: size += sizeof(struct ip6_hdr); break; #endif default: ipseclog((LOG_ERR, "%s: unknown AF %d in " "IPsec tunnel SA\n", __func__, sp->req[i]->saidx.dst.sa.sa_family)); break; } } } return (size); } /* * Compute ESP/AH header size for protocols with PCB, including * outer IP header. Currently only tcp_output() uses it. */ size_t ipsec_hdrsiz_inpcb(struct inpcb *inp) { struct secpolicyindex spidx; struct secpolicy *sp; size_t sz; sp = ipsec_getpcbpolicy(inp, IPSEC_DIR_OUTBOUND); if (sp == NULL && key_havesp(IPSEC_DIR_OUTBOUND)) { ipsec_setspidx_inpcb(inp, &spidx); spidx.dir = IPSEC_DIR_OUTBOUND; sp = key_allocsp(&spidx, IPSEC_DIR_OUTBOUND); } if (sp == NULL) sp = key_allocsp_default(); sz = ipsec_hdrsiz_internal(sp); key_freesp(&sp); return (sz); } -/* - * This function is called from ipsec_hdrsiz_tcp(), ip_ipsec_mtu(), - * disabled ip6_ipsec_mtu() and ip6_forward(). - */ -size_t -ipsec_hdrsiz(const struct mbuf *m, u_int dir, struct inpcb *inp) -{ - struct secpolicy *sp; - int error; - size_t size; - - if (!key_havesp(dir)) - return 0; - - IPSEC_ASSERT(m != NULL, ("null mbuf")); - - /* Get SP for this packet. */ - if (inp == NULL) - sp = ipsec_getpolicybyaddr(m, dir, &error); - else - sp = ipsec_getpolicybysock(m, dir, inp, &error); - - if (sp != NULL) { - size = ipsec_hdrsiz_internal(sp); - KEYDEBUG(KEYDEBUG_IPSEC_DATA, - printf("%s: size:%lu.\n", __func__, - (unsigned long)size)); - - KEY_FREESP(&sp); - } else { - size = 0; /* XXX Should be panic? - * -> No, we are called w/o knowing if - * IPsec processing is needed. */ - } - return (size); -} - /* * Check the variable replay window. * ipsec_chkreplay() performs replay check before ICV verification. * ipsec_updatereplay() updates replay bitmap. This must be called after * ICV verification (it also performs replay check, which is usually done * beforehand). * 0 (zero) is returned if packet disallowed, 1 if packet permitted. * * Based on RFC 2401. */ int ipsec_chkreplay(u_int32_t seq, struct secasvar *sav) { const struct secreplay *replay; u_int32_t diff; int fr; u_int32_t wsizeb; /* Constant: bits of window size. */ int frlast; /* Constant: last frame. */ IPSEC_ASSERT(sav != NULL, ("Null SA")); IPSEC_ASSERT(sav->replay != NULL, ("Null replay state")); replay = sav->replay; if (replay->wsize == 0) return (1); /* No need to check replay. */ /* Constant. */ frlast = replay->wsize - 1; wsizeb = replay->wsize << 3; /* Sequence number of 0 is invalid. */ if (seq == 0) return (0); /* First time is always okay. */ if (replay->count == 0) return (1); if (seq > replay->lastseq) { /* Larger sequences are okay. */ return (1); } else { /* seq is equal or less than lastseq. */ diff = replay->lastseq - seq; /* Over range to check, i.e. too old or wrapped. */ if (diff >= wsizeb) return (0); fr = frlast - diff / 8; /* This packet already seen? */ if ((replay->bitmap)[fr] & (1 << (diff % 8))) return (0); /* Out of order but good. */ return (1); } } /* * Check replay counter whether to update or not. * OUT: 0: OK * 1: NG */ int ipsec_updatereplay(u_int32_t seq, struct secasvar *sav) { char buf[128]; struct secreplay *replay; u_int32_t diff; int fr; u_int32_t wsizeb; /* Constant: bits of window size. */ int frlast; /* Constant: last frame. */ IPSEC_ASSERT(sav != NULL, ("Null SA")); IPSEC_ASSERT(sav->replay != NULL, ("Null replay state")); replay = sav->replay; if (replay->wsize == 0) goto ok; /* No need to check replay. */ /* Constant. */ frlast = replay->wsize - 1; wsizeb = replay->wsize << 3; /* Sequence number of 0 is invalid. */ if (seq == 0) return (1); /* First time. */ if (replay->count == 0) { replay->lastseq = seq; bzero(replay->bitmap, replay->wsize); (replay->bitmap)[frlast] = 1; goto ok; } if (seq > replay->lastseq) { /* seq is larger than lastseq. */ diff = seq - replay->lastseq; /* New larger sequence number. */ if (diff < wsizeb) { /* In window. */ /* Set bit for this packet. */ vshiftl(replay->bitmap, diff, replay->wsize); (replay->bitmap)[frlast] |= 1; } else { /* This packet has a "way larger". */ bzero(replay->bitmap, replay->wsize); (replay->bitmap)[frlast] = 1; } replay->lastseq = seq; /* Larger is good. */ } else { /* seq is equal or less than lastseq. */ diff = replay->lastseq - seq; /* Over range to check, i.e. too old or wrapped. */ if (diff >= wsizeb) return (1); fr = frlast - diff / 8; /* This packet already seen? */ if ((replay->bitmap)[fr] & (1 << (diff % 8))) return (1); /* Mark as seen. */ (replay->bitmap)[fr] |= (1 << (diff % 8)); /* Out of order but good. */ } ok: if (replay->count == ~0) { /* Set overflow flag. */ replay->overflow++; /* Don't increment, no more packets accepted. */ if ((sav->flags & SADB_X_EXT_CYCSEQ) == 0) return (1); ipseclog((LOG_WARNING, "%s: replay counter made %d cycle. %s\n", __func__, replay->overflow, ipsec_logsastr(sav, buf, sizeof(buf)))); } replay->count++; return (0); } int ipsec_updateid(struct secasvar *sav, uint64_t *new, uint64_t *old) { uint64_t tmp; /* * tdb_cryptoid is initialized by xform_init(). * Then it can be changed only when some crypto error occurred or * when SA is deleted. We stored used cryptoid in the xform_data * structure. In case when crypto error occurred and crypto * subsystem has reinited the session, it returns new cryptoid * and EAGAIN error code. * * This function will be called when we got EAGAIN from crypto * subsystem. * *new is cryptoid that was returned by crypto subsystem in * the crp_sid. * *old is the original cryptoid that we stored in xform_data. * * For first failed request *old == sav->tdb_cryptoid, then * we update sav->tdb_cryptoid and redo crypto_dispatch(). * For next failed request *old != sav->tdb_cryptoid, then * we store cryptoid from first request into the *new variable * and crp_sid from this second session will be returned via * *old pointer, so caller can release second session. * * XXXAE: check this more carefully. */ KEYDBG(IPSEC_STAMP, printf("%s: SA(%p) moves cryptoid %jd -> %jd\n", __func__, sav, (uintmax_t)(*old), (uintmax_t)(*new))); KEYDBG(IPSEC_DATA, kdebug_secasv(sav)); SECASVAR_LOCK(sav); if (sav->tdb_cryptoid != *old) { /* cryptoid was already updated */ tmp = *new; *new = sav->tdb_cryptoid; *old = tmp; SECASVAR_UNLOCK(sav); return (1); } sav->tdb_cryptoid = *new; SECASVAR_UNLOCK(sav); return (0); } /* * Shift variable length buffer to left. * IN: bitmap: pointer to the buffer * nbit: the number of to shift. * wsize: buffer size (bytes). */ static void vshiftl(unsigned char *bitmap, int nbit, int wsize) { int s, j, i; unsigned char over; for (j = 0; j < nbit; j += 8) { s = (nbit - j < 8) ? (nbit - j): 8; bitmap[0] <<= s; for (i = 1; i < wsize; i++) { over = (bitmap[i] >> (8 - s)); bitmap[i] <<= s; bitmap[i-1] |= over; } } } /* Return a printable string for the address. */ char* -ipsec_address(union sockaddr_union* sa, char *buf, socklen_t size) +ipsec_address(const union sockaddr_union* sa, char *buf, socklen_t size) { switch (sa->sa.sa_family) { #ifdef INET case AF_INET: return (inet_ntop(AF_INET, &sa->sin.sin_addr, buf, size)); #endif /* INET */ #ifdef INET6 case AF_INET6: if (IN6_IS_SCOPE_LINKLOCAL(&sa->sin6.sin6_addr)) { snprintf(buf, size, "%s%%%u", inet_ntop(AF_INET6, &sa->sin6.sin6_addr, buf, size), sa->sin6.sin6_scope_id); return (buf); } else return (inet_ntop(AF_INET6, &sa->sin6.sin6_addr, buf, size)); #endif /* INET6 */ case 0: return ("*"); default: return ("(unknown address family)"); } } char * ipsec_logsastr(struct secasvar *sav, char *buf, size_t size) { char sbuf[IPSEC_ADDRSTRLEN], dbuf[IPSEC_ADDRSTRLEN]; IPSEC_ASSERT(sav->sah->saidx.src.sa.sa_family == sav->sah->saidx.dst.sa.sa_family, ("address family mismatch")); snprintf(buf, size, "SA(SPI=%08lx src=%s dst=%s)", (u_long)ntohl(sav->spi), ipsec_address(&sav->sah->saidx.src, sbuf, sizeof(sbuf)), ipsec_address(&sav->sah->saidx.dst, dbuf, sizeof(dbuf))); return (buf); } void ipsec_dumpmbuf(const struct mbuf *m) { const u_char *p; int totlen; int i; totlen = 0; printf("---\n"); while (m) { p = mtod(m, const u_char *); for (i = 0; i < m->m_len; i++) { printf("%02x ", p[i]); totlen++; if (totlen % 16 == 0) printf("\n"); } m = m->m_next; } if (totlen % 16 != 0) printf("\n"); printf("---\n"); } static void def_policy_init(const void *unused __unused) { bzero(&V_def_policy, sizeof(struct secpolicy)); V_def_policy.policy = IPSEC_POLICY_NONE; V_def_policy.refcnt = 1; } VNET_SYSINIT(def_policy_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FIRST, def_policy_init, NULL); /* XXX This stuff doesn't belong here... */ static struct xformsw* xforms = NULL; /* * Register a transform; typically at system startup. */ void xform_register(struct xformsw* xsp) { xsp->xf_next = xforms; xforms = xsp; } /* * Initialize transform support in an sav. */ int xform_init(struct secasvar *sav, int xftype) { struct xformsw *xsp; if (sav->tdb_xform != NULL) /* Previously initialized. */ return (0); for (xsp = xforms; xsp; xsp = xsp->xf_next) if (xsp->xf_type == xftype) return ((*xsp->xf_init)(sav, xsp)); return (EINVAL); } Index: projects/ipsec/sys/netipsec/ipsec.h =================================================================== --- projects/ipsec/sys/netipsec/ipsec.h (revision 309109) +++ projects/ipsec/sys/netipsec/ipsec.h (revision 309110) @@ -1,358 +1,354 @@ /* $FreeBSD$ */ /* $KAME: ipsec.h,v 1.53 2001/11/20 08:32:38 itojun Exp $ */ /*- * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. * 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. Neither the name of the project 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 PROJECT 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 PROJECT 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. */ /* * IPsec controller part. */ #ifndef _NETIPSEC_IPSEC_H_ #define _NETIPSEC_IPSEC_H_ #if defined(_KERNEL) && !defined(_LKM) && !defined(KLD_MODULE) #include "opt_inet.h" #include "opt_ipsec.h" #endif #include #include #ifdef _KERNEL #include #include #include #define IPSEC_ASSERT(_c,_m) KASSERT(_c, _m) -#define IPSEC_IS_PRIVILEGED_SO(_so) \ - ((_so)->so_cred != NULL && \ - priv_check_cred((_so)->so_cred, PRIV_NETINET_IPSEC, 0) \ - == 0) - /* * Security Policy Index * Ensure that both address families in the "src" and "dst" are same. * When the value of the ul_proto is ICMPv6, the port field in "src" * specifies ICMPv6 type, and the port field in "dst" specifies ICMPv6 code. */ struct secpolicyindex { union sockaddr_union src; /* IP src address for SP */ union sockaddr_union dst; /* IP dst address for SP */ uint8_t ul_proto; /* upper layer Protocol */ uint8_t dir; /* direction of packet flow */ uint8_t prefs; /* prefix length in bits for src */ uint8_t prefd; /* prefix length in bits for dst */ }; /* Request for IPsec */ struct ipsecrequest { struct secasindex saidx;/* hint for search proper SA */ /* if __ss_len == 0 then no address specified.*/ u_int level; /* IPsec level defined below. */ }; /* Security Policy Data Base */ struct secpolicy { TAILQ_ENTRY(secpolicy) chain; LIST_ENTRY(secpolicy) idhash; LIST_ENTRY(secpolicy) drainq; struct secpolicyindex spidx; /* selector */ #define IPSEC_MAXREQ 4 struct ipsecrequest *req[IPSEC_MAXREQ]; u_int tcount; /* IPsec transforms count */ volatile u_int refcnt; /* reference count */ u_int policy; /* policy_type per pfkeyv2.h */ u_int state; #define IPSEC_SPSTATE_DEAD 0 #define IPSEC_SPSTATE_LARVAL 1 #define IPSEC_SPSTATE_ALIVE 2 #define IPSEC_SPSTATE_PCB 3 #define IPSEC_SPSTATE_IFNET 4 uint32_t priority; /* priority of this policy */ uint32_t id; /* It's unique number on the system. */ /* * lifetime handler. * the policy can be used without limitiation if both lifetime and * validtime are zero. * "lifetime" is passed by sadb_lifetime.sadb_lifetime_addtime. * "validtime" is passed by sadb_lifetime.sadb_lifetime_usetime. */ time_t created; /* time created the policy */ time_t lastused; /* updated every when kernel sends a packet */ long lifetime; /* duration of the lifetime of this policy */ long validtime; /* duration this policy is valid without use */ }; /* * PCB security policies. * Application can setup private security policies for socket. * Such policies can have IPSEC, BYPASS and ENTRUST type. * By default policies set to NULL, this mean that they have ENTRUST type. * When application sets BYPASS or IPSEC type policy, flags field * also updated. In case when flags is not set, the system could store * used security policy into the sp_in/sp_out pointer to speedup further * lookups. */ struct inpcbpolicy { struct secpolicy *sp_in; struct secpolicy *sp_out; uint32_t genid; uint16_t flags; #define INP_INBOUND_POLICY 0x0001 #define INP_OUTBOUND_POLICY 0x0002 uint16_t hdrsz; }; /* SP acquiring list table. */ struct secspacq { LIST_ENTRY(secspacq) chain; struct secpolicyindex spidx; time_t created; /* for lifetime */ int count; /* for lifetime */ /* XXX: here is mbuf place holder to be sent ? */ }; #endif /* _KERNEL */ /* buffer size for formatted output of ipsec address */ #define IPSEC_ADDRSTRLEN (INET6_ADDRSTRLEN + 11) /* according to IANA assignment, port 0x0000 and proto 0xff are reserved. */ #define IPSEC_PORT_ANY 0 #define IPSEC_ULPROTO_ANY 255 #define IPSEC_PROTO_ANY 255 /* mode of security protocol */ /* NOTE: DON'T use IPSEC_MODE_ANY at SPD. It's only use in SAD */ #define IPSEC_MODE_ANY 0 /* i.e. wildcard. */ #define IPSEC_MODE_TRANSPORT 1 #define IPSEC_MODE_TUNNEL 2 #define IPSEC_MODE_TCPMD5 3 /* TCP MD5 mode */ /* * Direction of security policy. * NOTE: Since INVALID is used just as flag. * The other are used for loop counter too. */ #define IPSEC_DIR_ANY 0 #define IPSEC_DIR_INBOUND 1 #define IPSEC_DIR_OUTBOUND 2 #define IPSEC_DIR_MAX 3 #define IPSEC_DIR_INVALID 4 /* Policy level */ /* * IPSEC, ENTRUST and BYPASS are allowed for setsockopt() in PCB, * DISCARD, IPSEC and NONE are allowed for setkey() in SPD. * DISCARD and NONE are allowed for system default. */ #define IPSEC_POLICY_DISCARD 0 /* discarding packet */ #define IPSEC_POLICY_NONE 1 /* through IPsec engine */ #define IPSEC_POLICY_IPSEC 2 /* do IPsec */ #define IPSEC_POLICY_ENTRUST 3 /* consulting SPD if present. */ #define IPSEC_POLICY_BYPASS 4 /* only for privileged socket. */ /* Security protocol level */ #define IPSEC_LEVEL_DEFAULT 0 /* reference to system default */ #define IPSEC_LEVEL_USE 1 /* use SA if present. */ #define IPSEC_LEVEL_REQUIRE 2 /* require SA. */ #define IPSEC_LEVEL_UNIQUE 3 /* unique SA. */ #define IPSEC_MANUAL_REQID_MAX 0x3fff /* * if security policy level == unique, this id * indicate to a relative SA for use, else is * zero. * 1 - 0x3fff are reserved for manual keying. * 0 are reserved for above reason. Others is * for kernel use. * Note that this id doesn't identify SA * by only itself. */ #define IPSEC_REPLAYWSIZE 32 /* statistics for ipsec processing */ struct ipsecstat { uint64_t ips_in_polvio; /* input: sec policy violation */ uint64_t ips_in_nomem; /* input: no memory available */ uint64_t ips_in_inval; /* input: generic error */ uint64_t ips_out_polvio; /* output: sec policy violation */ uint64_t ips_out_nosa; /* output: SA unavailable */ uint64_t ips_out_nomem; /* output: no memory available */ uint64_t ips_out_noroute; /* output: no route available */ uint64_t ips_out_inval; /* output: generic error */ uint64_t ips_out_bundlesa; /* output: bundled SA processed */ uint64_t ips_mbcoalesced; /* mbufs coalesced during clone */ uint64_t ips_clcoalesced; /* clusters coalesced during clone */ uint64_t ips_clcopied; /* clusters copied during clone */ uint64_t ips_mbinserted; /* mbufs inserted during makespace */ /* * Temporary statistics for performance analysis. */ /* See where ESP/AH/IPCOMP header land in mbuf on input */ uint64_t ips_input_front; uint64_t ips_input_middle; uint64_t ips_input_end; }; /* * Definitions for IPsec & Key sysctl operations. */ #define IPSECCTL_STATS 1 /* stats */ #define IPSECCTL_DEF_POLICY 2 #define IPSECCTL_DEF_ESP_TRANSLEV 3 /* int; ESP transport mode */ #define IPSECCTL_DEF_ESP_NETLEV 4 /* int; ESP tunnel mode */ #define IPSECCTL_DEF_AH_TRANSLEV 5 /* int; AH transport mode */ #define IPSECCTL_DEF_AH_NETLEV 6 /* int; AH tunnel mode */ #if 0 /* obsolete, do not reuse */ #define IPSECCTL_INBOUND_CALL_IKE 7 #endif #define IPSECCTL_AH_CLEARTOS 8 #define IPSECCTL_AH_OFFSETMASK 9 #define IPSECCTL_DFBIT 10 #define IPSECCTL_ECN 11 #define IPSECCTL_DEBUG 12 #define IPSECCTL_ESP_RANDPAD 13 #ifdef _KERNEL #include struct ipsec_ctx_data; #define IPSEC_INIT_CTX(_ctx, _mp, _sav, _af, _enc) do { \ (_ctx)->mp = (_mp); \ (_ctx)->sav = (_sav); \ (_ctx)->af = (_af); \ (_ctx)->enc = (_enc); \ } while(0) int ipsec_run_hhooks(struct ipsec_ctx_data *ctx, int direction); VNET_DECLARE(int, ipsec_debug); #define V_ipsec_debug VNET(ipsec_debug) #ifdef REGRESSION VNET_DECLARE(int, ipsec_replay); VNET_DECLARE(int, ipsec_integrity); #define V_ipsec_replay VNET(ipsec_replay) #define V_ipsec_integrity VNET(ipsec_integrity) #endif VNET_PCPUSTAT_DECLARE(struct ipsecstat, ipsec4stat); VNET_DECLARE(int, ip4_esp_trans_deflev); VNET_DECLARE(int, ip4_esp_net_deflev); VNET_DECLARE(int, ip4_ah_trans_deflev); VNET_DECLARE(int, ip4_ah_net_deflev); VNET_DECLARE(int, ip4_ah_offsetmask); VNET_DECLARE(int, ip4_ipsec_dfbit); VNET_DECLARE(int, ip4_ipsec_ecn); VNET_DECLARE(int, ip4_esp_randpad); VNET_DECLARE(int, crypto_support); #define IPSECSTAT_INC(name) \ VNET_PCPUSTAT_ADD(struct ipsecstat, ipsec4stat, name, 1) #define V_ip4_esp_trans_deflev VNET(ip4_esp_trans_deflev) #define V_ip4_esp_net_deflev VNET(ip4_esp_net_deflev) #define V_ip4_ah_trans_deflev VNET(ip4_ah_trans_deflev) #define V_ip4_ah_net_deflev VNET(ip4_ah_net_deflev) #define V_ip4_ah_offsetmask VNET(ip4_ah_offsetmask) #define V_ip4_ipsec_dfbit VNET(ip4_ipsec_dfbit) #define V_ip4_ipsec_ecn VNET(ip4_ipsec_ecn) #define V_ip4_esp_randpad VNET(ip4_esp_randpad) #define V_crypto_support VNET(crypto_support) #define ipseclog(x) do { if (V_ipsec_debug) log x; } while (0) /* for openbsd compatibility */ #define DPRINTF(x) do { if (V_ipsec_debug) printf x; } while (0) struct inpcb; +struct m_tag; struct secasvar; struct sockopt; +union sockaddr_union; struct ipsecrequest *ipsec_newisr(void); void ipsec_delisr(struct ipsecrequest *); struct secpolicy *ipsec4_checkpolicy(const struct mbuf *, struct inpcb *, int *); u_int ipsec_get_reqlevel(struct secpolicy *, u_int); int ipsec4_in_reject(const struct mbuf *, struct inpcb *); size_t ipsec_hdrsiz_inpcb(struct inpcb *); int ipsec_init_pcbpolicy(struct inpcb *); int ipsec_delete_pcbpolicy(struct inpcb *); int ipsec_copy_pcbpolicy(struct inpcb *, struct inpcb *); int ipsec_control_pcbpolicy(struct inpcb *, struct sockopt *); -extern int ipsec_chkreplay(u_int32_t, struct secasvar *); -extern int ipsec_updatereplay(u_int32_t, struct secasvar *); +int ipsec_chkreplay(uint32_t, struct secasvar *); +int ipsec_updatereplay(uint32_t, struct secasvar *); +int ipsec_updateid(struct secasvar *, uint64_t *, uint64_t *); -extern size_t ipsec_hdrsiz(const struct mbuf *, u_int, struct inpcb *); +char *ipsec_address(const union sockaddr_union *, char *, socklen_t); +char *ipsec_logsastr(struct secasvar *, char *, size_t); -union sockaddr_union; -extern char *ipsec_address(union sockaddr_union *, char *, socklen_t); -extern char *ipsec_logsastr(struct secasvar *, char *, size_t); - extern void ipsec_dumpmbuf(const struct mbuf *); -struct m_tag; extern int ah4_input(struct mbuf **mp, int *offp, int proto); extern void ah4_ctlinput(int cmd, struct sockaddr *sa, void *); extern int esp4_input(struct mbuf **mp, int *offp, int proto); extern void esp4_ctlinput(int cmd, struct sockaddr *sa, void *); extern int ipcomp4_input(struct mbuf **mp, int *offp, int proto); extern int ipsec_common_input(struct mbuf *m, int, int, int, int); extern int ipsec4_common_input_cb(struct mbuf *m, struct secasvar *sav, int skip, int protoff); -extern int ipsec4_process_packet(struct mbuf *, struct ipsecrequest *); -extern int ipsec_process_done(struct mbuf *, struct ipsecrequest *); +extern int ipsec4_process_packet(struct mbuf *, struct secpolicy *, + struct inpcb *); +extern int ipsec_process_done(struct mbuf *, struct secpolicy *, + struct secasvar *, u_int); extern void m_checkalignment(const char* where, struct mbuf *m0, int off, int len); extern struct mbuf *m_makespace(struct mbuf *m0, int skip, int hlen, int *off); extern caddr_t m_pad(struct mbuf *m, int n); extern int m_striphdr(struct mbuf *m, int skip, int hlen); #endif /* _KERNEL */ #ifndef _KERNEL extern caddr_t ipsec_set_policy(char *, int); extern int ipsec_get_policylen(caddr_t); extern char *ipsec_dump_policy(caddr_t, char *); extern const char *ipsec_strerror(void); #endif /* ! KERNEL */ #endif /* _NETIPSEC_IPSEC_H_ */ Index: projects/ipsec/sys/netipsec/ipsec6.h =================================================================== --- projects/ipsec/sys/netipsec/ipsec6.h (revision 309109) +++ projects/ipsec/sys/netipsec/ipsec6.h (revision 309110) @@ -1,72 +1,75 @@ /* $FreeBSD$ */ /* $KAME: ipsec.h,v 1.44 2001/03/23 08:08:47 itojun Exp $ */ /*- * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. * 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. Neither the name of the project 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 PROJECT 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 PROJECT 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. */ /* * IPsec controller part. */ #ifndef _NETIPSEC_IPSEC6_H_ #define _NETIPSEC_IPSEC6_H_ #include #include #ifdef _KERNEL #include VNET_PCPUSTAT_DECLARE(struct ipsecstat, ipsec6stat); VNET_DECLARE(int, ip6_esp_trans_deflev); VNET_DECLARE(int, ip6_esp_net_deflev); VNET_DECLARE(int, ip6_ah_trans_deflev); VNET_DECLARE(int, ip6_ah_net_deflev); VNET_DECLARE(int, ip6_ipsec_ecn); #define IPSEC6STAT_INC(name) \ VNET_PCPUSTAT_ADD(struct ipsecstat, ipsec6stat, name, 1) #define V_ip6_esp_trans_deflev VNET(ip6_esp_trans_deflev) #define V_ip6_esp_net_deflev VNET(ip6_esp_net_deflev) #define V_ip6_ah_trans_deflev VNET(ip6_ah_trans_deflev) #define V_ip6_ah_net_deflev VNET(ip6_ah_net_deflev) #define V_ip6_ipsec_ecn VNET(ip6_ipsec_ecn) struct inpcb; extern int ipsec6_in_reject(const struct mbuf *, struct inpcb *); +struct secpolicy *ipsec6_checkpolicy(const struct mbuf *, + struct inpcb *, int *); struct m_tag; extern int ipsec6_common_input(struct mbuf **mp, int *offp, int proto); extern int ipsec6_common_input_cb(struct mbuf *m, struct secasvar *sav, int skip, int protoff); extern void esp6_ctlinput(int, struct sockaddr *, void *); -extern int ipsec6_process_packet(struct mbuf *, struct ipsecrequest *); +int ipsec6_process_packet(struct mbuf *, struct secpolicy *, + struct inpcb *); #endif /*_KERNEL*/ #endif /*_NETIPSEC_IPSEC6_H_*/ Index: projects/ipsec/sys/netipsec/key.c =================================================================== --- projects/ipsec/sys/netipsec/key.c (revision 309109) +++ projects/ipsec/sys/netipsec/key.c (revision 309110) @@ -1,7632 +1,7548 @@ /* $FreeBSD$ */ /* $KAME: key.c,v 1.191 2001/06/27 10:46:49 sakane Exp $ */ /*- * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. * 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. Neither the name of the project 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 PROJECT 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 PROJECT 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. */ /* * This code is referd to RFC 2367 */ #include "opt_inet.h" #include "opt_inet6.h" #include "opt_ipsec.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef INET6 #include #include #include #endif /* INET6 */ #if defined(INET) || defined(INET6) #include #endif #ifdef INET6 #include #endif /* INET6 */ #include #include #include #include #include #include #ifdef INET6 #include #endif #include #include /* randomness */ #include #define FULLMASK 0xff #define _BITS(bytes) ((bytes) << 3) /* * Note on SA reference counting: * - SAs that are not in DEAD state will have (total external reference + 1) * following value in reference count field. they cannot be freed and are * referenced from SA header. * - SAs that are in DEAD state will have (total external reference) * in reference count field. they are ready to be freed. reference from * SA header will be removed in key_delsav(), when the reference count * field hits 0 (= no external reference other than from SA header. */ VNET_DEFINE(u_int32_t, key_debug_level) = 0; static VNET_DEFINE(u_int, key_spi_trycnt) = 1000; static VNET_DEFINE(u_int32_t, key_spi_minval) = 0x100; static VNET_DEFINE(u_int32_t, key_spi_maxval) = 0x0fffffff; /* XXX */ static VNET_DEFINE(u_int32_t, policy_id) = 0; /*interval to initialize randseed,1(m)*/ static VNET_DEFINE(u_int, key_int_random) = 60; /* interval to expire acquiring, 30(s)*/ static VNET_DEFINE(u_int, key_larval_lifetime) = 30; /* counter for blocking SADB_ACQUIRE.*/ static VNET_DEFINE(int, key_blockacq_count) = 10; /* lifetime for blocking SADB_ACQUIRE.*/ static VNET_DEFINE(int, key_blockacq_lifetime) = 20; /* preferred old sa rather than new sa.*/ static VNET_DEFINE(int, key_preferred_oldsa) = 1; #define V_key_spi_trycnt VNET(key_spi_trycnt) #define V_key_spi_minval VNET(key_spi_minval) #define V_key_spi_maxval VNET(key_spi_maxval) #define V_policy_id VNET(policy_id) #define V_key_int_random VNET(key_int_random) #define V_key_larval_lifetime VNET(key_larval_lifetime) #define V_key_blockacq_count VNET(key_blockacq_count) #define V_key_blockacq_lifetime VNET(key_blockacq_lifetime) #define V_key_preferred_oldsa VNET(key_preferred_oldsa) static VNET_DEFINE(u_int32_t, acq_seq) = 0; #define V_acq_seq VNET(acq_seq) static VNET_DEFINE(uint32_t, sp_genid) = 0; #define V_sp_genid VNET(sp_genid) /* SPD */ TAILQ_HEAD(secpolicy_queue, secpolicy); LIST_HEAD(secpolicy_list, secpolicy); static VNET_DEFINE(struct secpolicy_queue, sptree[IPSEC_DIR_MAX]); static struct rmlock sptree_lock; #define V_sptree VNET(sptree) #define SPTREE_LOCK_INIT() rm_init(&sptree_lock, "sptree") #define SPTREE_LOCK_DESTROY() rm_destroy(&sptree_lock) #define SPTREE_RLOCK_TRACKER struct rm_priotracker sptree_tracker #define SPTREE_RLOCK() rm_rlock(&sptree_lock, &sptree_tracker) #define SPTREE_RUNLOCK() rm_runlock(&sptree_lock, &sptree_tracker) #define SPTREE_RLOCK_ASSERT() rm_assert(&sptree_lock, RA_RLOCKED) #define SPTREE_WLOCK() rm_wlock(&sptree_lock) #define SPTREE_WUNLOCK() rm_wunlock(&sptree_lock) #define SPTREE_WLOCK_ASSERT() rm_assert(&sptree_lock, RA_WLOCKED) #define SPTREE_UNLOCK_ASSERT() rm_assert(&sptree_lock, RA_UNLOCKED) /* Hash table for lookup SP using unique id */ static VNET_DEFINE(struct secpolicy_list *, sphashtbl); static VNET_DEFINE(u_long, sphash_mask); #define V_sphashtbl VNET(sphashtbl) #define V_sphash_mask VNET(sphash_mask) #define SPHASH_NHASH_LOG2 7 #define SPHASH_NHASH (1 << SPHASH_NHASH_LOG2) #define SPHASH_HASHVAL(id) (key_u32hash(id) & V_sphash_mask) #define SPHASH_HASH(id) &V_sphashtbl[SPHASH_HASHVAL(id)] /* SAD */ TAILQ_HEAD(secashead_queue, secashead); LIST_HEAD(secashead_list, secashead); static VNET_DEFINE(struct secashead_queue, sahtree); static struct rmlock sahtree_lock; #define V_sahtree VNET(sahtree) #define SAHTREE_LOCK_INIT() rm_init(&sahtree_lock, "sahtree") #define SAHTREE_LOCK_DESTROY() rm_destroy(&sahtree_lock) #define SAHTREE_RLOCK_TRACKER struct rm_priotracker sahtree_tracker #define SAHTREE_RLOCK() rm_rlock(&sahtree_lock, &sahtree_tracker) #define SAHTREE_RUNLOCK() rm_runlock(&sahtree_lock, &sahtree_tracker) #define SAHTREE_RLOCK_ASSERT() rm_assert(&sahtree_lock, RA_RLOCKED) #define SAHTREE_WLOCK() rm_wlock(&sahtree_lock) #define SAHTREE_WUNLOCK() rm_wunlock(&sahtree_lock) #define SAHTREE_WLOCK_ASSERT() rm_assert(&sahtree_lock, RA_WLOCKED) #define SAHTREE_UNLOCK_ASSERT() rm_assert(&sahtree_lock, RA_UNLOCKED) /* Hash table for lookup in SAD using SA addresses */ static VNET_DEFINE(struct secashead_list *, sahaddrhashtbl); static VNET_DEFINE(u_long, sahaddrhash_mask); #define V_sahaddrhashtbl VNET(sahaddrhashtbl) #define V_sahaddrhash_mask VNET(sahaddrhash_mask) #define SAHHASH_NHASH_LOG2 7 #define SAHHASH_NHASH (1 << SAHHASH_NHASH_LOG2) #define SAHADDRHASH_HASHVAL(saidx) \ (key_saidxhash(saidx) & V_sahaddrhash_mask) #define SAHADDRHASH_HASH(saidx) \ &V_sahaddrhashtbl[SAHADDRHASH_HASHVAL(saidx)] /* Hash table for lookup in SAD using SPI */ LIST_HEAD(secasvar_list, secasvar); static VNET_DEFINE(struct secasvar_list *, savhashtbl); static VNET_DEFINE(u_long, savhash_mask); #define V_savhashtbl VNET(savhashtbl) #define V_savhash_mask VNET(savhash_mask) #define SAVHASH_NHASH_LOG2 7 #define SAVHASH_NHASH (1 << SAVHASH_NHASH_LOG2) #define SAVHASH_HASHVAL(spi) (key_u32hash(spi) & V_savhash_mask) #define SAVHASH_HASH(spi) &V_savhashtbl[SAVHASH_HASHVAL(spi)] static uint32_t key_saidxhash(const struct secasindex *saidx) { uint32_t hval; hval = fnv_32_buf(&saidx->proto, sizeof(saidx->proto), FNV1_32_INIT); switch (saidx->dst.sa.sa_family) { #ifdef INET case AF_INET: hval = fnv_32_buf(&saidx->src.sin.sin_addr, sizeof(in_addr_t), hval); hval = fnv_32_buf(&saidx->dst.sin.sin_addr, sizeof(in_addr_t), hval); break; #endif #ifdef INET6 case AF_INET6: hval = fnv_32_buf(&saidx->src.sin6.sin6_addr, sizeof(struct in6_addr), hval); hval = fnv_32_buf(&saidx->dst.sin6.sin6_addr, sizeof(struct in6_addr), hval); break; #endif default: hval = 0; ipseclog((LOG_DEBUG, "%s: unknown address family %d", __func__, saidx->dst.sa.sa_family)); } return (hval); } static uint32_t key_u32hash(uint32_t val) { return (fnv_32_buf(&val, sizeof(val), FNV1_32_INIT)); } /* registed list */ static VNET_DEFINE(LIST_HEAD(_regtree, secreg), regtree[SADB_SATYPE_MAX + 1]); #define V_regtree VNET(regtree) static struct mtx regtree_lock; #define REGTREE_LOCK_INIT() \ mtx_init(®tree_lock, "regtree", "fast ipsec regtree", MTX_DEF) #define REGTREE_LOCK_DESTROY() mtx_destroy(®tree_lock) #define REGTREE_LOCK() mtx_lock(®tree_lock) #define REGTREE_UNLOCK() mtx_unlock(®tree_lock) #define REGTREE_LOCK_ASSERT() mtx_assert(®tree_lock, MA_OWNED) /* Acquiring list */ LIST_HEAD(secacq_list, secacq); static VNET_DEFINE(struct secacq_list, acqtree); #define V_acqtree VNET(acqtree) static struct mtx acq_lock; #define ACQ_LOCK_INIT() \ mtx_init(&acq_lock, "acqtree", "ipsec SA acquiring list", MTX_DEF) #define ACQ_LOCK_DESTROY() mtx_destroy(&acq_lock) #define ACQ_LOCK() mtx_lock(&acq_lock) #define ACQ_UNLOCK() mtx_unlock(&acq_lock) #define ACQ_LOCK_ASSERT() mtx_assert(&acq_lock, MA_OWNED) /* Hash table for lookup in ACQ list using SA addresses */ static VNET_DEFINE(struct secacq_list *, acqaddrhashtbl); static VNET_DEFINE(u_long, acqaddrhash_mask); #define V_acqaddrhashtbl VNET(acqaddrhashtbl) #define V_acqaddrhash_mask VNET(acqaddrhash_mask) /* Hash table for lookup in ACQ list using SEQ number */ static VNET_DEFINE(struct secacq_list *, acqseqhashtbl); static VNET_DEFINE(u_long, acqseqhash_mask); #define V_acqseqhashtbl VNET(acqseqhashtbl) #define V_acqseqhash_mask VNET(acqseqhash_mask) #define ACQHASH_NHASH_LOG2 7 #define ACQHASH_NHASH (1 << ACQHASH_NHASH_LOG2) #define ACQADDRHASH_HASHVAL(saidx) \ (key_saidxhash(saidx) & V_acqaddrhash_mask) #define ACQSEQHASH_HASHVAL(seq) \ (key_u32hash(seq) & V_acqseqhash_mask) #define ACQADDRHASH_HASH(saidx) \ &V_acqaddrhashtbl[ACQADDRHASH_HASHVAL(saidx)] #define ACQSEQHASH_HASH(seq) \ &V_acqseqhashtbl[ACQSEQHASH_HASHVAL(seq)] /* SP acquiring list */ static VNET_DEFINE(LIST_HEAD(_spacqtree, secspacq), spacqtree); #define V_spacqtree VNET(spacqtree) static struct mtx spacq_lock; #define SPACQ_LOCK_INIT() \ mtx_init(&spacq_lock, "spacqtree", \ "fast ipsec security policy acquire list", MTX_DEF) #define SPACQ_LOCK_DESTROY() mtx_destroy(&spacq_lock) #define SPACQ_LOCK() mtx_lock(&spacq_lock) #define SPACQ_UNLOCK() mtx_unlock(&spacq_lock) #define SPACQ_LOCK_ASSERT() mtx_assert(&spacq_lock, MA_OWNED) static const int minsize[] = { sizeof(struct sadb_msg), /* SADB_EXT_RESERVED */ sizeof(struct sadb_sa), /* SADB_EXT_SA */ sizeof(struct sadb_lifetime), /* SADB_EXT_LIFETIME_CURRENT */ sizeof(struct sadb_lifetime), /* SADB_EXT_LIFETIME_HARD */ sizeof(struct sadb_lifetime), /* SADB_EXT_LIFETIME_SOFT */ sizeof(struct sadb_address), /* SADB_EXT_ADDRESS_SRC */ sizeof(struct sadb_address), /* SADB_EXT_ADDRESS_DST */ sizeof(struct sadb_address), /* SADB_EXT_ADDRESS_PROXY */ sizeof(struct sadb_key), /* SADB_EXT_KEY_AUTH */ sizeof(struct sadb_key), /* SADB_EXT_KEY_ENCRYPT */ sizeof(struct sadb_ident), /* SADB_EXT_IDENTITY_SRC */ sizeof(struct sadb_ident), /* SADB_EXT_IDENTITY_DST */ sizeof(struct sadb_sens), /* SADB_EXT_SENSITIVITY */ sizeof(struct sadb_prop), /* SADB_EXT_PROPOSAL */ sizeof(struct sadb_supported), /* SADB_EXT_SUPPORTED_AUTH */ sizeof(struct sadb_supported), /* SADB_EXT_SUPPORTED_ENCRYPT */ sizeof(struct sadb_spirange), /* SADB_EXT_SPIRANGE */ 0, /* SADB_X_EXT_KMPRIVATE */ sizeof(struct sadb_x_policy), /* SADB_X_EXT_POLICY */ sizeof(struct sadb_x_sa2), /* SADB_X_SA2 */ sizeof(struct sadb_x_nat_t_type),/* SADB_X_EXT_NAT_T_TYPE */ sizeof(struct sadb_x_nat_t_port),/* SADB_X_EXT_NAT_T_SPORT */ sizeof(struct sadb_x_nat_t_port),/* SADB_X_EXT_NAT_T_DPORT */ sizeof(struct sadb_address), /* SADB_X_EXT_NAT_T_OAI */ sizeof(struct sadb_address), /* SADB_X_EXT_NAT_T_OAR */ sizeof(struct sadb_x_nat_t_frag),/* SADB_X_EXT_NAT_T_FRAG */ }; static const int maxsize[] = { sizeof(struct sadb_msg), /* SADB_EXT_RESERVED */ sizeof(struct sadb_sa), /* SADB_EXT_SA */ sizeof(struct sadb_lifetime), /* SADB_EXT_LIFETIME_CURRENT */ sizeof(struct sadb_lifetime), /* SADB_EXT_LIFETIME_HARD */ sizeof(struct sadb_lifetime), /* SADB_EXT_LIFETIME_SOFT */ 0, /* SADB_EXT_ADDRESS_SRC */ 0, /* SADB_EXT_ADDRESS_DST */ 0, /* SADB_EXT_ADDRESS_PROXY */ 0, /* SADB_EXT_KEY_AUTH */ 0, /* SADB_EXT_KEY_ENCRYPT */ 0, /* SADB_EXT_IDENTITY_SRC */ 0, /* SADB_EXT_IDENTITY_DST */ 0, /* SADB_EXT_SENSITIVITY */ 0, /* SADB_EXT_PROPOSAL */ 0, /* SADB_EXT_SUPPORTED_AUTH */ 0, /* SADB_EXT_SUPPORTED_ENCRYPT */ sizeof(struct sadb_spirange), /* SADB_EXT_SPIRANGE */ 0, /* SADB_X_EXT_KMPRIVATE */ 0, /* SADB_X_EXT_POLICY */ sizeof(struct sadb_x_sa2), /* SADB_X_SA2 */ sizeof(struct sadb_x_nat_t_type),/* SADB_X_EXT_NAT_T_TYPE */ sizeof(struct sadb_x_nat_t_port),/* SADB_X_EXT_NAT_T_SPORT */ sizeof(struct sadb_x_nat_t_port),/* SADB_X_EXT_NAT_T_DPORT */ 0, /* SADB_X_EXT_NAT_T_OAI */ 0, /* SADB_X_EXT_NAT_T_OAR */ sizeof(struct sadb_x_nat_t_frag),/* SADB_X_EXT_NAT_T_FRAG */ }; #define SADB_CHECKLEN(_mhp, _ext) \ ((_mhp)->extlen[(_ext)] < minsize[(_ext)] || (maxsize[(_ext)] != 0 && \ ((_mhp)->extlen[(_ext)] > maxsize[(_ext)]))) #define SADB_CHECKHDR(_mhp, _ext) ((_mhp)->ext[(_ext)] == NULL) static VNET_DEFINE(int, ipsec_esp_keymin) = 256; static VNET_DEFINE(int, ipsec_esp_auth) = 0; static VNET_DEFINE(int, ipsec_ah_keymin) = 128; #define V_ipsec_esp_keymin VNET(ipsec_esp_keymin) #define V_ipsec_esp_auth VNET(ipsec_esp_auth) #define V_ipsec_ah_keymin VNET(ipsec_ah_keymin) #ifdef SYSCTL_DECL SYSCTL_DECL(_net_key); #endif SYSCTL_INT(_net_key, KEYCTL_DEBUG_LEVEL, debug, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_debug_level), 0, ""); /* max count of trial for the decision of spi value */ SYSCTL_INT(_net_key, KEYCTL_SPI_TRY, spi_trycnt, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_spi_trycnt), 0, ""); /* minimum spi value to allocate automatically. */ SYSCTL_INT(_net_key, KEYCTL_SPI_MIN_VALUE, spi_minval, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_spi_minval), 0, ""); /* maximun spi value to allocate automatically. */ SYSCTL_INT(_net_key, KEYCTL_SPI_MAX_VALUE, spi_maxval, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_spi_maxval), 0, ""); /* interval to initialize randseed */ SYSCTL_INT(_net_key, KEYCTL_RANDOM_INT, int_random, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_int_random), 0, ""); /* lifetime for larval SA */ SYSCTL_INT(_net_key, KEYCTL_LARVAL_LIFETIME, larval_lifetime, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_larval_lifetime), 0, ""); /* counter for blocking to send SADB_ACQUIRE to IKEd */ SYSCTL_INT(_net_key, KEYCTL_BLOCKACQ_COUNT, blockacq_count, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_blockacq_count), 0, ""); /* lifetime for blocking to send SADB_ACQUIRE to IKEd */ SYSCTL_INT(_net_key, KEYCTL_BLOCKACQ_LIFETIME, blockacq_lifetime, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_blockacq_lifetime), 0, ""); /* ESP auth */ SYSCTL_INT(_net_key, KEYCTL_ESP_AUTH, esp_auth, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_esp_auth), 0, ""); /* minimum ESP key length */ SYSCTL_INT(_net_key, KEYCTL_ESP_KEYMIN, esp_keymin, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_esp_keymin), 0, ""); /* minimum AH key length */ SYSCTL_INT(_net_key, KEYCTL_AH_KEYMIN, ah_keymin, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_ah_keymin), 0, ""); /* perfered old SA rather than new SA */ SYSCTL_INT(_net_key, KEYCTL_PREFERED_OLDSA, preferred_oldsa, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_preferred_oldsa), 0, ""); #define __LIST_CHAINED(elm) \ (!((elm)->chain.le_next == NULL && (elm)->chain.le_prev == NULL)) MALLOC_DEFINE(M_IPSEC_SA, "secasvar", "ipsec security association"); MALLOC_DEFINE(M_IPSEC_SAH, "sahead", "ipsec sa head"); MALLOC_DEFINE(M_IPSEC_SP, "ipsecpolicy", "ipsec security policy"); MALLOC_DEFINE(M_IPSEC_SR, "ipsecrequest", "ipsec security request"); MALLOC_DEFINE(M_IPSEC_MISC, "ipsec-misc", "ipsec miscellaneous"); MALLOC_DEFINE(M_IPSEC_SAQ, "ipsec-saq", "ipsec sa acquire"); MALLOC_DEFINE(M_IPSEC_SAR, "ipsec-reg", "ipsec sa acquire"); static VNET_DEFINE(uma_zone_t, key_lft_zone); #define V_key_lft_zone VNET(key_lft_zone) /* * set parameters into secpolicyindex buffer. * Must allocate secpolicyindex buffer passed to this function. */ #define KEY_SETSECSPIDX(_dir, s, d, ps, pd, ulp, idx) \ do { \ bzero((idx), sizeof(struct secpolicyindex)); \ (idx)->dir = (_dir); \ (idx)->prefs = (ps); \ (idx)->prefd = (pd); \ (idx)->ul_proto = (ulp); \ bcopy((s), &(idx)->src, ((const struct sockaddr *)(s))->sa_len); \ bcopy((d), &(idx)->dst, ((const struct sockaddr *)(d))->sa_len); \ } while (0) /* * set parameters into secasindex buffer. * Must allocate secasindex buffer before calling this function. */ #define KEY_SETSECASIDX(p, m, r, s, d, idx) \ do { \ bzero((idx), sizeof(struct secasindex)); \ (idx)->proto = (p); \ (idx)->mode = (m); \ (idx)->reqid = (r); \ bcopy((s), &(idx)->src, ((const struct sockaddr *)(s))->sa_len); \ bcopy((d), &(idx)->dst, ((const struct sockaddr *)(d))->sa_len); \ } while (0) /* key statistics */ struct _keystat { u_long getspi_count; /* the avarage of count to try to get new SPI */ } keystat; struct sadb_msghdr { struct sadb_msg *msg; struct sadb_ext *ext[SADB_EXT_MAX + 1]; int extoff[SADB_EXT_MAX + 1]; int extlen[SADB_EXT_MAX + 1]; }; #ifndef IPSEC_DEBUG2 static struct callout key_timer; #endif static void key_unlink(struct secpolicy *); static struct secpolicy *key_getsp(struct secpolicyindex *); static struct secpolicy *key_getspbyid(u_int32_t); static struct mbuf *key_gather_mbuf(struct mbuf *, const struct sadb_msghdr *, int, int, ...); static int key_spdadd(struct socket *, struct mbuf *, const struct sadb_msghdr *); static uint32_t key_getnewspid(void); static int key_spddelete(struct socket *, struct mbuf *, const struct sadb_msghdr *); static int key_spddelete2(struct socket *, struct mbuf *, const struct sadb_msghdr *); static int key_spdget(struct socket *, struct mbuf *, const struct sadb_msghdr *); static int key_spdflush(struct socket *, struct mbuf *, const struct sadb_msghdr *); static int key_spddump(struct socket *, struct mbuf *, const struct sadb_msghdr *); static struct mbuf *key_setdumpsp(struct secpolicy *, u_int8_t, u_int32_t, u_int32_t); static struct mbuf *key_sp2mbuf(struct secpolicy *); static size_t key_getspreqmsglen(struct secpolicy *); static int key_spdexpire(struct secpolicy *); static struct secashead *key_newsah(struct secasindex *); static void key_freesah(struct secashead **); static void key_delsah(struct secashead *); static struct secasvar *key_newsav(const struct sadb_msghdr *, struct secasindex *, uint32_t, int *); static void key_delsav(struct secasvar *); static void key_unlinksav(struct secasvar *); static struct secashead *key_getsah(struct secasindex *); static int key_checkspidup(uint32_t); static struct secasvar *key_getsavbyspi(uint32_t); static int key_setnatt(struct secasvar *, const struct sadb_msghdr *); static int key_setsaval(struct secasvar *, const struct sadb_msghdr *); static int key_updatelifetimes(struct secasvar *, const struct sadb_msghdr *); static struct mbuf *key_setdumpsa(struct secasvar *, u_int8_t, u_int8_t, u_int32_t, u_int32_t); static struct mbuf *key_setsadbmsg(u_int8_t, u_int16_t, u_int8_t, u_int32_t, pid_t, u_int16_t); static struct mbuf *key_setsadbsa(struct secasvar *); static struct mbuf *key_setsadbaddr(u_int16_t, const struct sockaddr *, u_int8_t, u_int16_t); #ifdef IPSEC_NAT_T static struct mbuf *key_setsadbxport(u_int16_t, u_int16_t); static struct mbuf *key_setsadbxtype(u_int16_t); #endif static void key_porttosaddr(struct sockaddr *, u_int16_t); #define KEY_PORTTOSADDR(saddr, port) \ key_porttosaddr((struct sockaddr *)(saddr), (port)) static struct mbuf *key_setsadbxsa2(u_int8_t, u_int32_t, u_int32_t); static struct mbuf *key_setsadbxpolicy(u_int16_t, u_int8_t, u_int32_t, u_int32_t); static struct seckey *key_dup_keymsg(const struct sadb_key *, size_t, struct malloc_type *); static struct seclifetime *key_dup_lifemsg(const struct sadb_lifetime *src, struct malloc_type *); -#ifdef INET6 -static int key_ismyaddr6(struct sockaddr_in6 *); -#endif /* flags for key_cmpsaidx() */ #define CMP_HEAD 1 /* protocol, addresses. */ #define CMP_MODE_REQID 2 /* additionally HEAD, reqid, mode. */ #define CMP_REQID 3 /* additionally HEAD, reaid. */ #define CMP_EXACTLY 4 /* all elements. */ static int key_cmpsaidx(const struct secasindex *, const struct secasindex *, int); static int key_cmpspidx_exactly(struct secpolicyindex *, struct secpolicyindex *); static int key_cmpspidx_withmask(struct secpolicyindex *, struct secpolicyindex *); static int key_bbcmp(const void *, const void *, u_int); static uint8_t key_satype2proto(uint8_t); static uint8_t key_proto2satype(uint8_t); static int key_getspi(struct socket *, struct mbuf *, const struct sadb_msghdr *); static u_int32_t key_do_getnewspi(struct sadb_spirange *, struct secasindex *); static int key_update(struct socket *, struct mbuf *, const struct sadb_msghdr *); #ifdef IPSEC_DOSEQCHECK static struct secasvar *key_getsavbyseq(struct secashead *, u_int32_t); #endif static int key_add(struct socket *, struct mbuf *, const struct sadb_msghdr *); static int key_setident(struct secashead *, const struct sadb_msghdr *); static struct mbuf *key_getmsgbuf_x1(struct mbuf *, const struct sadb_msghdr *); static int key_delete(struct socket *, struct mbuf *, const struct sadb_msghdr *); static int key_delete_all(struct socket *, struct mbuf *, const struct sadb_msghdr *, struct secasindex *); static int key_get(struct socket *, struct mbuf *, const struct sadb_msghdr *); static void key_getcomb_setlifetime(struct sadb_comb *); static struct mbuf *key_getcomb_esp(void); static struct mbuf *key_getcomb_ah(void); static struct mbuf *key_getcomb_ipcomp(void); static struct mbuf *key_getprop(const struct secasindex *); static int key_acquire(const struct secasindex *, struct secpolicy *); static uint32_t key_newacq(const struct secasindex *, int *); static uint32_t key_getacq(const struct secasindex *, int *); static int key_acqdone(const struct secasindex *, uint32_t); static int key_acqreset(uint32_t); static struct secspacq *key_newspacq(struct secpolicyindex *); static struct secspacq *key_getspacq(struct secpolicyindex *); static int key_acquire2(struct socket *, struct mbuf *, const struct sadb_msghdr *); static int key_register(struct socket *, struct mbuf *, const struct sadb_msghdr *); static int key_expire(struct secasvar *, int); static int key_flush(struct socket *, struct mbuf *, const struct sadb_msghdr *); static int key_dump(struct socket *, struct mbuf *, const struct sadb_msghdr *); static int key_promisc(struct socket *, struct mbuf *, const struct sadb_msghdr *); static int key_senderror(struct socket *, struct mbuf *, int); static int key_validate_ext(const struct sadb_ext *, int); static int key_align(struct mbuf *, struct sadb_msghdr *); static struct mbuf *key_setlifetime(struct seclifetime *src, u_int16_t exttype); static struct mbuf *key_setkey(struct seckey *src, u_int16_t exttype); #if 0 static const char *key_getfqdn(void); static const char *key_getuserfqdn(void); #endif #define DBG_IPSEC_INITREF(t, p) do { \ refcount_init(&(p)->refcnt, 1); \ KEYDBG(KEY_STAMP, \ printf("%s: Initialize refcnt %s(%p) = %u\n", \ __func__, #t, (p), (p)->refcnt)); \ } while (0) #define DBG_IPSEC_ADDREF(t, p) do { \ refcount_acquire(&(p)->refcnt); \ KEYDBG(KEY_STAMP, \ printf("%s: Acquire refcnt %s(%p) -> %u\n", \ __func__, #t, (p), (p)->refcnt)); \ } while (0) #define DBG_IPSEC_DELREF(t, p) do { \ KEYDBG(KEY_STAMP, \ printf("%s: Release refcnt %s(%p) -> %u\n", \ __func__, #t, (p), (p)->refcnt - 1)); \ refcount_release(&(p)->refcnt); \ } while (0) #define IPSEC_INITREF(t, p) refcount_init(&(p)->refcnt, 1) #define IPSEC_ADDREF(t, p) refcount_acquire(&(p)->refcnt) #define IPSEC_DELREF(t, p) refcount_release(&(p)->refcnt) #define SP_INITREF(p) IPSEC_INITREF(SP, p) #define SP_ADDREF(p) IPSEC_ADDREF(SP, p) #define SP_DELREF(p) IPSEC_DELREF(SP, p) #define SAH_INITREF(p) IPSEC_INITREF(SAH, p) #define SAH_ADDREF(p) IPSEC_ADDREF(SAH, p) #define SAH_DELREF(p) IPSEC_DELREF(SAH, p) #define SAV_INITREF(p) IPSEC_INITREF(SAV, p) #define SAV_ADDREF(p) IPSEC_ADDREF(SAV, p) #define SAV_DELREF(p) IPSEC_DELREF(SAV, p) /* * Update the refcnt while holding the SPTREE lock. */ void key_addref(struct secpolicy *sp) { SP_ADDREF(sp); } /* * Return 0 when there are known to be no SP's for the specified * direction. Otherwise return 1. This is used by IPsec code * to optimize performance. */ int key_havesp(u_int dir) { return (dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND ? TAILQ_FIRST(&V_sptree[dir]) != NULL : 1); } /* %%% IPsec policy management */ /* * Return current SPDB generation. */ uint32_t key_getspgen(void) { return (V_sp_genid); } static int key_checksockaddrs(struct sockaddr *src, struct sockaddr *dst) { /* family match */ if (src->sa_family != dst->sa_family) return (EINVAL); /* sa_len match */ if (src->sa_len != dst->sa_len) return (EINVAL); switch (src->sa_family) { #ifdef INET case AF_INET: if (src->sa_len != sizeof(struct sockaddr_in)) return (EINVAL); break; #endif #ifdef INET6 case AF_INET6: if (src->sa_len != sizeof(struct sockaddr_in6)) return (EINVAL); break; #endif default: return (EAFNOSUPPORT); } return (0); } /* * allocating a SP for OUTBOUND or INBOUND packet. * Must call key_freesp() later. * OUT: NULL: not found * others: found and return the pointer. */ struct secpolicy * key_allocsp(struct secpolicyindex *spidx, u_int dir) { SPTREE_RLOCK_TRACKER; struct secpolicy *sp; IPSEC_ASSERT(spidx != NULL, ("null spidx")); IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND, ("invalid direction %u", dir)); SPTREE_RLOCK(); TAILQ_FOREACH(sp, &V_sptree[dir], chain) { if (key_cmpspidx_withmask(&sp->spidx, spidx)) { SP_ADDREF(sp); break; } } SPTREE_RUNLOCK(); if (sp != NULL) { /* found a SPD entry */ sp->lastused = time_second; KEYDBG(IPSEC_STAMP, printf("%s: return SP(%p)\n", __func__, sp)); KEYDBG(IPSEC_DATA, kdebug_secpolicy(sp)); } else { KEYDBG(IPSEC_DATA, printf("%s: lookup failed for ", __func__); kdebug_secpolicyindex(spidx, NULL)); } return (sp); } /* * Allocating an SA entry for an *OUTBOUND* packet. * OUT: positive: corresponding SA for given saidx found. * NULL: SA not found, but will be acquired, check *error * for acquiring status. */ struct secasvar * key_allocsa_policy(struct secpolicy *sp, const struct secasindex *saidx, int *error) { SAHTREE_RLOCK_TRACKER; struct secashead *sah; struct secasvar *sav; IPSEC_ASSERT(saidx != NULL, ("null saidx")); IPSEC_ASSERT(saidx->mode == IPSEC_MODE_TRANSPORT || saidx->mode == IPSEC_MODE_TUNNEL, ("unexpected policy %u", saidx->mode)); /* * We check new SA in the IPsec request because a different * SA may be involved each time this request is checked, either * because new SAs are being configured, or this request is * associated with an unconnected datagram socket, or this request * is associated with a system default policy. */ SAHTREE_RLOCK(); LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) { KEYDBG(IPSEC_DUMP, printf("%s: checking SAH\n", __func__); kdebug_secash(sah, " ")); if (key_cmpsaidx(&sah->saidx, saidx, CMP_MODE_REQID)) break; } if (sah != NULL) { /* * Allocate the oldest SA available according to * draft-jenkins-ipsec-rekeying-03. */ if (V_key_preferred_oldsa) sav = TAILQ_LAST(&sah->savtree_alive, secasvar_queue); else sav = TAILQ_FIRST(&sah->savtree_alive); if (sav != NULL) SAV_ADDREF(sav); } else sav = NULL; SAHTREE_RUNLOCK(); if (sav != NULL) { *error = 0; KEYDBG(IPSEC_STAMP, printf("%s: chosen SA(%p) for SP(%p)\n", __func__, sav, sp)); KEYDBG(IPSEC_DATA, kdebug_secasv(sav)); return (sav); /* return referenced SA */ } /* there is no SA */ *error = key_acquire(saidx, sp); if ((*error) != 0) ipseclog((LOG_DEBUG, "%s: error %d returned from key_acquire()\n", __func__, *error)); KEYDBG(IPSEC_STAMP, printf("%s: acquire SA for SP(%p), error %d\n", __func__, sp, *error)); KEYDBG(IPSEC_DATA, kdebug_secasindex(saidx, NULL)); return (NULL); } /* * allocating a usable SA entry for a *INBOUND* packet. * Must call key_freesav() later. * OUT: positive: pointer to a usable sav (i.e. MATURE or DYING state). * NULL: not found, or error occurred. * * In the comparison, no source address is used--for RFC2401 conformance. * To quote, from section 4.1: * A security association is uniquely identified by a triple consisting * of a Security Parameter Index (SPI), an IP Destination Address, and a * security protocol (AH or ESP) identifier. * Note that, however, we do need to keep source address in IPsec SA. * IKE specification and PF_KEY specification do assume that we * keep source address in IPsec SA. We see a tricky situation here. */ struct secasvar * key_allocsa(union sockaddr_union *dst, uint8_t proto, uint32_t spi) { SAHTREE_RLOCK_TRACKER; struct secasvar *sav; int chkport; IPSEC_ASSERT(dst != NULL, ("null dst address")); chkport = 0; SAHTREE_RLOCK(); LIST_FOREACH(sav, SAVHASH_HASH(spi), spihash) { if (sav->spi == spi) break; } /* * We use single SPI namespace for all protocols, so it is * impossible to have SPI duplicates in the SAVHASH. * XXXAE: this breaks TCP_SIGNATURE. */ if (sav != NULL) { #ifdef IPSEC_NAT_T /* * Really only check ports when this is a NAT-T * SA. Otherwise other lookups providing ports * might suffer. */ chkport = (sav->natt_type != 0 && dst->sa.sa_family == AF_INET && dst->sa.sa_len == sizeof(struct sockaddr_in) && dst->sin.sin_port != 0); #endif if (sav->state != SADB_SASTATE_LARVAL && sav->sah->saidx.proto == proto && key_sockaddrcmp(&dst->sa, &sav->sah->saidx.dst.sa, chkport) == 0) SAV_ADDREF(sav); else sav = NULL; } SAHTREE_RUNLOCK(); if (sav == NULL) { KEYDBG(IPSEC_STAMP, char buf[IPSEC_ADDRSTRLEN]; printf("%s: SA not found for spi %u proto %u dst %s\n", __func__, ntohl(spi), proto, ipsec_address(dst, buf, sizeof(buf)))); } else { KEYDBG(IPSEC_STAMP, printf("%s: return SA(%p)\n", __func__, sav)); KEYDBG(IPSEC_DATA, kdebug_secasv(sav)); } return (sav); } struct secasvar * key_allocsa_tunnel(union sockaddr_union *src, union sockaddr_union *dst, uint8_t proto) { SAHTREE_RLOCK_TRACKER; struct secasindex saidx; struct secashead *sah; struct secasvar *sav; IPSEC_ASSERT(src != NULL, ("null src address")); IPSEC_ASSERT(dst != NULL, ("null dst address")); KEY_SETSECASIDX(proto, IPSEC_MODE_TUNNEL, 0, &src->sa, &dst->sa, &saidx); sav = NULL; SAHTREE_RLOCK(); LIST_FOREACH(sah, SAHADDRHASH_HASH(&saidx), addrhash) { if (IPSEC_MODE_TUNNEL != sah->saidx.mode) continue; if (proto != sah->saidx.proto) continue; if (key_sockaddrcmp(&src->sa, &sav->sah->saidx.src.sa, 0) != 0) continue; if (key_sockaddrcmp(&dst->sa, &sav->sah->saidx.dst.sa, 0) != 0) continue; /* XXXAE: is key_preferred_oldsa reasonably?*/ if (V_key_preferred_oldsa) sav = TAILQ_LAST(&sah->savtree_alive, secasvar_queue); else sav = TAILQ_FIRST(&sah->savtree_alive); if (sav != NULL) { SAV_ADDREF(sav); break; } } SAHTREE_RUNLOCK(); KEYDBG(IPSEC_STAMP, printf("%s: return SA(%p)\n", __func__, sav)); if (sav != NULL) KEYDBG(IPSEC_DATA, kdebug_secasv(sav)); return (sav); } /* * Must be called after calling key_allocsp(). */ void key_freesp(struct secpolicy **spp) { struct secpolicy *sp = *spp; IPSEC_ASSERT(sp != NULL, ("null sp")); if (SP_DELREF(sp) == 0) return; KEYDBG(IPSEC_STAMP, printf("%s: last reference to SP(%p)\n", __func__, sp)); KEYDBG(IPSEC_DATA, kdebug_secpolicy(sp)); *spp = NULL; while (sp->tcount > 0) ipsec_delisr(sp->req[--sp->tcount]); free(sp, M_IPSEC_SP); } static void key_unlink(struct secpolicy *sp) { IPSEC_ASSERT(sp->spidx.dir == IPSEC_DIR_INBOUND || sp->spidx.dir == IPSEC_DIR_OUTBOUND, ("invalid direction %u", sp->spidx.dir)); SPTREE_UNLOCK_ASSERT(); KEYDBG(KEY_STAMP, printf("%s: SP(%p)\n", __func__, sp)); SPTREE_WLOCK(); if (sp->state != IPSEC_SPSTATE_ALIVE) { /* SP is already unlinked */ SPTREE_WUNLOCK(); return; } sp->state = IPSEC_SPSTATE_DEAD; TAILQ_REMOVE(&V_sptree[sp->spidx.dir], sp, chain); LIST_REMOVE(sp, idhash); V_sp_genid++; SPTREE_WUNLOCK(); key_freesp(&sp); } /* * insert a secpolicy into the SP database. Lower priorities first */ static void key_insertsp(struct secpolicy *newsp) { struct secpolicy *sp; SPTREE_WLOCK_ASSERT(); TAILQ_FOREACH(sp, &V_sptree[newsp->spidx.dir], chain) { if (newsp->priority < sp->priority) { TAILQ_INSERT_BEFORE(sp, newsp, chain); goto done; } } TAILQ_INSERT_TAIL(&V_sptree[newsp->spidx.dir], newsp, chain); done: LIST_INSERT_HEAD(SPHASH_HASH(newsp->id), newsp, idhash); newsp->state = IPSEC_SPSTATE_ALIVE; V_sp_genid++; } -void -key_addrefsa(struct secasvar *sav, const char* where, int tag) -{ - - IPSEC_ASSERT(sav != NULL, ("null sav")); - IPSEC_ASSERT(sav->refcnt > 0, ("refcount must exist")); - - SAV_ADDREF(sav); -} - /* * Must be called after calling key_allocsa(). * This function is called by key_freesp() to free some SA allocated * for a policy. */ void key_freesav(struct secasvar **psav) { struct secasvar *sav = *psav; IPSEC_ASSERT(sav != NULL, ("null sav")); if (SAV_DELREF(sav) == 0) return; KEYDBG(IPSEC_STAMP, printf("%s: last reference to SA(%p)\n", __func__, sav)); *psav = NULL; key_delsav(sav); } /* * Unlink SA from SAH and SPI hash under SAHTREE_WLOCK. * Expect that SA has extra reference due to lookup. * Release this references, also release SAH reference after unlink. */ static void key_unlinksav(struct secasvar *sav) { struct secashead *sah; KEYDBG(KEY_STAMP, printf("%s: SA(%p)\n", __func__, sav)); SAHTREE_UNLOCK_ASSERT(); SAHTREE_WLOCK(); if (sav->state == SADB_SASTATE_DEAD) { /* SA is already unlinked */ SAHTREE_WUNLOCK(); return; } /* Unlink from SAH */ if (sav->state == SADB_SASTATE_LARVAL) TAILQ_REMOVE(&sav->sah->savtree_larval, sav, chain); else TAILQ_REMOVE(&sav->sah->savtree_alive, sav, chain); /* Unlink from SPI hash */ LIST_REMOVE(sav, spihash); sav->state = SADB_SASTATE_DEAD; sah = sav->sah; SAHTREE_WUNLOCK(); key_freesav(&sav); /* Since we are unlinked, release reference to SAH */ key_freesah(&sah); } /* %%% SPD management */ /* * search SPD * OUT: NULL : not found * others : found, pointer to a SP. */ static struct secpolicy * key_getsp(struct secpolicyindex *spidx) { SPTREE_RLOCK_TRACKER; struct secpolicy *sp; IPSEC_ASSERT(spidx != NULL, ("null spidx")); SPTREE_RLOCK(); TAILQ_FOREACH(sp, &V_sptree[spidx->dir], chain) { if (key_cmpspidx_exactly(spidx, &sp->spidx)) { SP_ADDREF(sp); break; } } SPTREE_RUNLOCK(); return sp; } /* * get SP by index. * OUT: NULL : not found * others : found, pointer to referenced SP. */ static struct secpolicy * key_getspbyid(uint32_t id) { SPTREE_RLOCK_TRACKER; struct secpolicy *sp; SPTREE_RLOCK(); LIST_FOREACH(sp, SPHASH_HASH(id), idhash) { if (sp->id == id) { SP_ADDREF(sp); break; } } SPTREE_RUNLOCK(); return (sp); } struct secpolicy * key_newsp(void) { struct secpolicy *sp; sp = malloc(sizeof(*sp), M_IPSEC_SP, M_NOWAIT | M_ZERO); if (sp != NULL) SP_INITREF(sp); return (sp); } /* * create secpolicy structure from sadb_x_policy structure. * NOTE: `state', `secpolicyindex' and 'id' in secpolicy structure * are not set, so must be set properly later. */ struct secpolicy * key_msg2sp(struct sadb_x_policy *xpl0, size_t len, int *error) { struct secpolicy *newsp; IPSEC_ASSERT(xpl0 != NULL, ("null xpl0")); IPSEC_ASSERT(len >= sizeof(*xpl0), ("policy too short: %zu", len)); if (len != PFKEY_EXTLEN(xpl0)) { ipseclog((LOG_DEBUG, "%s: Invalid msg length.\n", __func__)); *error = EINVAL; return NULL; } if ((newsp = key_newsp()) == NULL) { *error = ENOBUFS; return NULL; } newsp->spidx.dir = xpl0->sadb_x_policy_dir; newsp->policy = xpl0->sadb_x_policy_type; newsp->priority = xpl0->sadb_x_policy_priority; newsp->tcount = 0; /* check policy */ switch (xpl0->sadb_x_policy_type) { case IPSEC_POLICY_DISCARD: case IPSEC_POLICY_NONE: case IPSEC_POLICY_ENTRUST: case IPSEC_POLICY_BYPASS: break; case IPSEC_POLICY_IPSEC: { struct sadb_x_ipsecrequest *xisr; struct ipsecrequest *isr; int tlen; /* validity check */ if (PFKEY_EXTLEN(xpl0) < sizeof(*xpl0)) { ipseclog((LOG_DEBUG, "%s: Invalid msg length.\n", __func__)); key_freesp(&newsp); *error = EINVAL; return NULL; } tlen = PFKEY_EXTLEN(xpl0) - sizeof(*xpl0); xisr = (struct sadb_x_ipsecrequest *)(xpl0 + 1); while (tlen > 0) { /* length check */ if (xisr->sadb_x_ipsecrequest_len < sizeof(*xisr)) { ipseclog((LOG_DEBUG, "%s: invalid ipsecrequest " "length.\n", __func__)); key_freesp(&newsp); *error = EINVAL; return NULL; } if (newsp->tcount >= IPSEC_MAXREQ) { ipseclog((LOG_DEBUG, "%s: too many ipsecrequests.\n", __func__)); key_freesp(&newsp); *error = EINVAL; return (NULL); } /* allocate request buffer */ /* NB: data structure is zero'd */ isr = ipsec_newisr(); if (isr == NULL) { ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__)); key_freesp(&newsp); *error = ENOBUFS; return NULL; } newsp->req[newsp->tcount++] = isr; /* set values */ switch (xisr->sadb_x_ipsecrequest_proto) { case IPPROTO_ESP: case IPPROTO_AH: case IPPROTO_IPCOMP: break; default: ipseclog((LOG_DEBUG, "%s: invalid proto type=%u\n", __func__, xisr->sadb_x_ipsecrequest_proto)); key_freesp(&newsp); *error = EPROTONOSUPPORT; return NULL; } isr->saidx.proto = (uint8_t)xisr->sadb_x_ipsecrequest_proto; switch (xisr->sadb_x_ipsecrequest_mode) { case IPSEC_MODE_TRANSPORT: case IPSEC_MODE_TUNNEL: break; case IPSEC_MODE_ANY: default: ipseclog((LOG_DEBUG, "%s: invalid mode=%u\n", __func__, xisr->sadb_x_ipsecrequest_mode)); key_freesp(&newsp); *error = EINVAL; return NULL; } isr->saidx.mode = xisr->sadb_x_ipsecrequest_mode; switch (xisr->sadb_x_ipsecrequest_level) { case IPSEC_LEVEL_DEFAULT: case IPSEC_LEVEL_USE: case IPSEC_LEVEL_REQUIRE: break; case IPSEC_LEVEL_UNIQUE: /* validity check */ /* * If range violation of reqid, kernel will * update it, don't refuse it. */ if (xisr->sadb_x_ipsecrequest_reqid > IPSEC_MANUAL_REQID_MAX) { ipseclog((LOG_DEBUG, "%s: reqid=%d range " "violation, updated by kernel.\n", __func__, xisr->sadb_x_ipsecrequest_reqid)); xisr->sadb_x_ipsecrequest_reqid = 0; } /* allocate new reqid id if reqid is zero. */ if (xisr->sadb_x_ipsecrequest_reqid == 0) { u_int32_t reqid; if ((reqid = key_newreqid()) == 0) { key_freesp(&newsp); *error = ENOBUFS; return NULL; } isr->saidx.reqid = reqid; xisr->sadb_x_ipsecrequest_reqid = reqid; } else { /* set it for manual keying. */ isr->saidx.reqid = xisr->sadb_x_ipsecrequest_reqid; } break; default: ipseclog((LOG_DEBUG, "%s: invalid level=%u\n", __func__, xisr->sadb_x_ipsecrequest_level)); key_freesp(&newsp); *error = EINVAL; return NULL; } isr->level = xisr->sadb_x_ipsecrequest_level; /* set IP addresses if there */ /* XXXAE: those are needed only for tunnel mode */ if (xisr->sadb_x_ipsecrequest_len > sizeof(*xisr)) { struct sockaddr *paddr; paddr = (struct sockaddr *)(xisr + 1); /* validity check */ if (paddr->sa_len > sizeof(isr->saidx.src)) { ipseclog((LOG_DEBUG, "%s: invalid " "request address length.\n", __func__)); key_freesp(&newsp); *error = EINVAL; return NULL; } bcopy(paddr, &isr->saidx.src, paddr->sa_len); paddr = (struct sockaddr *)((caddr_t)paddr + paddr->sa_len); /* validity check */ if (paddr->sa_len > sizeof(isr->saidx.dst)) { ipseclog((LOG_DEBUG, "%s: invalid " "request address length.\n", __func__)); key_freesp(&newsp); *error = EINVAL; return NULL; } /* AF family should match */ if (paddr->sa_family != isr->saidx.src.sa.sa_family) { ipseclog((LOG_DEBUG, "%s: address " "family doesn't match.\n", __func__)); key_freesp(&newsp); *error = EINVAL; return (NULL); } bcopy(paddr, &isr->saidx.dst, paddr->sa_len); } tlen -= xisr->sadb_x_ipsecrequest_len; /* validity check */ if (tlen < 0) { ipseclog((LOG_DEBUG, "%s: becoming tlen < 0.\n", __func__)); key_freesp(&newsp); *error = EINVAL; return NULL; } xisr = (struct sadb_x_ipsecrequest *)((caddr_t)xisr + xisr->sadb_x_ipsecrequest_len); } /* XXXAE: LARVAL SP */ if (newsp->tcount < 1) { ipseclog((LOG_DEBUG, "%s: valid IPSEC transforms " "not found.\n", __func__)); key_freesp(&newsp); *error = EINVAL; return (NULL); } } break; default: ipseclog((LOG_DEBUG, "%s: invalid policy type.\n", __func__)); key_freesp(&newsp); *error = EINVAL; return NULL; } *error = 0; return (newsp); } uint32_t key_newreqid(void) { static uint32_t auto_reqid = IPSEC_MANUAL_REQID_MAX + 1; if (auto_reqid == ~0) auto_reqid = IPSEC_MANUAL_REQID_MAX + 1; else auto_reqid++; /* XXX should be unique check */ return (auto_reqid); } /* * copy secpolicy struct to sadb_x_policy structure indicated. */ static struct mbuf * key_sp2mbuf(struct secpolicy *sp) { struct mbuf *m; size_t tlen; tlen = key_getspreqmsglen(sp); m = m_get2(tlen, M_NOWAIT, MT_DATA, 0); if (m == NULL) return (NULL); m_align(m, tlen); m->m_len = tlen; if (key_sp2msg(sp, m->m_data, &tlen) != 0) { m_freem(m); return (NULL); } return (m); } int key_sp2msg(struct secpolicy *sp, void *request, size_t *len) { struct sadb_x_ipsecrequest *xisr; struct sadb_x_policy *xpl; struct ipsecrequest *isr; size_t xlen, ilen; caddr_t p; int error, i; IPSEC_ASSERT(sp != NULL, ("null policy")); xlen = sizeof(*xpl); if (*len < xlen) return (EINVAL); error = 0; bzero(request, *len); xpl = (struct sadb_x_policy *)request; xpl->sadb_x_policy_exttype = SADB_X_EXT_POLICY; xpl->sadb_x_policy_type = sp->policy; xpl->sadb_x_policy_dir = sp->spidx.dir; xpl->sadb_x_policy_id = sp->id; xpl->sadb_x_policy_priority = sp->priority; /* if is the policy for ipsec ? */ if (sp->policy == IPSEC_POLICY_IPSEC) { p = (caddr_t)xpl + sizeof(*xpl); for (i = 0; i < sp->tcount; i++) { isr = sp->req[i]; ilen = PFKEY_ALIGN8(sizeof(*xisr) + isr->saidx.src.sa.sa_len + isr->saidx.dst.sa.sa_len); xlen += ilen; if (xlen > *len) { error = ENOBUFS; /* Calculate needed size */ continue; } xisr = (struct sadb_x_ipsecrequest *)p; xisr->sadb_x_ipsecrequest_len = ilen; xisr->sadb_x_ipsecrequest_proto = isr->saidx.proto; xisr->sadb_x_ipsecrequest_mode = isr->saidx.mode; xisr->sadb_x_ipsecrequest_level = isr->level; xisr->sadb_x_ipsecrequest_reqid = isr->saidx.reqid; p += sizeof(*xisr); bcopy(&isr->saidx.src, p, isr->saidx.src.sa.sa_len); p += isr->saidx.src.sa.sa_len; bcopy(&isr->saidx.dst, p, isr->saidx.dst.sa.sa_len); p += isr->saidx.dst.sa.sa_len; } } xpl->sadb_x_policy_len = PFKEY_UNIT64(xlen); if (error == 0) *len = xlen; else *len = sizeof(*xpl); return (error); } /* m will not be freed nor modified */ static struct mbuf * key_gather_mbuf(struct mbuf *m, const struct sadb_msghdr *mhp, int ndeep, int nitem, ...) { va_list ap; int idx; int i; struct mbuf *result = NULL, *n; int len; IPSEC_ASSERT(m != NULL, ("null mbuf")); IPSEC_ASSERT(mhp != NULL, ("null msghdr")); va_start(ap, nitem); for (i = 0; i < nitem; i++) { idx = va_arg(ap, int); if (idx < 0 || idx > SADB_EXT_MAX) goto fail; /* don't attempt to pull empty extension */ if (idx == SADB_EXT_RESERVED && mhp->msg == NULL) continue; if (idx != SADB_EXT_RESERVED && (mhp->ext[idx] == NULL || mhp->extlen[idx] == 0)) continue; if (idx == SADB_EXT_RESERVED) { len = PFKEY_ALIGN8(sizeof(struct sadb_msg)); IPSEC_ASSERT(len <= MHLEN, ("header too big %u", len)); MGETHDR(n, M_NOWAIT, MT_DATA); if (!n) goto fail; n->m_len = len; n->m_next = NULL; m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, caddr_t)); } else if (i < ndeep) { len = mhp->extlen[idx]; n = m_get2(len, M_NOWAIT, MT_DATA, 0); if (n == NULL) goto fail; m_align(n, len); n->m_len = len; m_copydata(m, mhp->extoff[idx], mhp->extlen[idx], mtod(n, caddr_t)); } else { n = m_copym(m, mhp->extoff[idx], mhp->extlen[idx], M_NOWAIT); } if (n == NULL) goto fail; if (result) m_cat(result, n); else result = n; } va_end(ap); if ((result->m_flags & M_PKTHDR) != 0) { result->m_pkthdr.len = 0; for (n = result; n; n = n->m_next) result->m_pkthdr.len += n->m_len; } return result; fail: m_freem(result); va_end(ap); return NULL; } /* * SADB_X_SPDADD, SADB_X_SPDSETIDX or SADB_X_SPDUPDATE processing * add an entry to SP database, when received * * from the user(?). * Adding to SP database, * and send * * to the socket which was send. * * SPDADD set a unique policy entry. * SPDSETIDX like SPDADD without a part of policy requests. * SPDUPDATE replace a unique policy entry. * * XXXAE: serialize this in PF_KEY to avoid races. * m will always be freed. */ static int key_spdadd(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp) { struct secpolicyindex spidx; struct sadb_address *src0, *dst0; struct sadb_x_policy *xpl0, *xpl; struct sadb_lifetime *lft = NULL; struct secpolicy *newsp; int error; IPSEC_ASSERT(so != NULL, ("null socket")); IPSEC_ASSERT(m != NULL, ("null mbuf")); IPSEC_ASSERT(mhp != NULL, ("null msghdr")); IPSEC_ASSERT(mhp->msg != NULL, ("null msg")); if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) || SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) || SADB_CHECKHDR(mhp, SADB_X_EXT_POLICY)) { ipseclog((LOG_DEBUG, "%s: invalid message: missing required header.\n", __func__)); return key_senderror(so, m, EINVAL); } if (SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) || SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST) || SADB_CHECKLEN(mhp, SADB_X_EXT_POLICY)) { ipseclog((LOG_DEBUG, "%s: invalid message: wrong header size.\n", __func__)); return key_senderror(so, m, EINVAL); } if (!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD)) { if (SADB_CHECKLEN(mhp, SADB_EXT_LIFETIME_HARD)) { ipseclog((LOG_DEBUG, "%s: invalid message: wrong header size.\n", __func__)); return key_senderror(so, m, EINVAL); } lft = (struct sadb_lifetime *)mhp->ext[SADB_EXT_LIFETIME_HARD]; } src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC]; dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST]; xpl0 = (struct sadb_x_policy *)mhp->ext[SADB_X_EXT_POLICY]; /* * Note: do not parse SADB_X_EXT_NAT_T_* here: * we are processing traffic endpoints. */ /* check the direciton */ switch (xpl0->sadb_x_policy_dir) { case IPSEC_DIR_INBOUND: case IPSEC_DIR_OUTBOUND: break; default: ipseclog((LOG_DEBUG, "%s: invalid SP direction.\n", __func__)); return key_senderror(so, m, EINVAL); } /* key_spdadd() accepts DISCARD, NONE and IPSEC. */ if (xpl0->sadb_x_policy_type != IPSEC_POLICY_DISCARD && xpl0->sadb_x_policy_type != IPSEC_POLICY_NONE && xpl0->sadb_x_policy_type != IPSEC_POLICY_IPSEC) { ipseclog((LOG_DEBUG, "%s: invalid policy type.\n", __func__)); return key_senderror(so, m, EINVAL); } /* policy requests are mandatory when action is ipsec. */ if (xpl0->sadb_x_policy_type == IPSEC_POLICY_IPSEC && mhp->extlen[SADB_X_EXT_POLICY] <= sizeof(*xpl0)) { ipseclog((LOG_DEBUG, "%s: policy requests required.\n", __func__)); return key_senderror(so, m, EINVAL); } error = key_checksockaddrs((struct sockaddr *)(src0 + 1), (struct sockaddr *)(dst0 + 1)); if (error != 0 || src0->sadb_address_proto != dst0->sadb_address_proto) { ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__)); return key_senderror(so, m, error); } /* make secindex */ KEY_SETSECSPIDX(xpl0->sadb_x_policy_dir, src0 + 1, dst0 + 1, src0->sadb_address_prefixlen, dst0->sadb_address_prefixlen, src0->sadb_address_proto, &spidx); /* Checking there is SP already or not. */ newsp = key_getsp(&spidx); if (newsp != NULL) { if (mhp->msg->sadb_msg_type == SADB_X_SPDUPDATE) { KEYDBG(KEY_STAMP, printf("%s: unlink SP(%p) for SPDUPDATE\n", __func__, newsp)); KEYDBG(KEY_DATA, kdebug_secpolicy(newsp)); key_unlink(newsp); key_freesp(&newsp); } else { key_freesp(&newsp); ipseclog((LOG_DEBUG, "%s: a SP entry exists already.", __func__)); return (key_senderror(so, m, EEXIST)); } } /* allocate new SP entry */ if ((newsp = key_msg2sp(xpl0, PFKEY_EXTLEN(xpl0), &error)) == NULL) { return key_senderror(so, m, error); } newsp->lastused = newsp->created = time_second; newsp->lifetime = lft ? lft->sadb_lifetime_addtime : 0; newsp->validtime = lft ? lft->sadb_lifetime_usetime : 0; bcopy(&spidx, &newsp->spidx, sizeof(spidx)); /* XXXAE: there is race between key_getsp() and key_insertsp() */ SPTREE_WLOCK(); if ((newsp->id = key_getnewspid()) == 0) { SPTREE_WUNLOCK(); key_freesp(&newsp); return key_senderror(so, m, ENOBUFS); } key_insertsp(newsp); SPTREE_WUNLOCK(); KEYDBG(KEY_STAMP, printf("%s: SP(%p)\n", __func__, newsp)); KEYDBG(KEY_DATA, kdebug_secpolicy(newsp)); { struct mbuf *n, *mpolicy; struct sadb_msg *newmsg; int off; /* * Note: do not send SADB_X_EXT_NAT_T_* here: * we are sending traffic endpoints. */ /* create new sadb_msg to reply. */ if (lft) { n = key_gather_mbuf(m, mhp, 2, 5, SADB_EXT_RESERVED, SADB_X_EXT_POLICY, SADB_EXT_LIFETIME_HARD, SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST); } else { n = key_gather_mbuf(m, mhp, 2, 4, SADB_EXT_RESERVED, SADB_X_EXT_POLICY, SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST); } if (!n) return key_senderror(so, m, ENOBUFS); if (n->m_len < sizeof(*newmsg)) { n = m_pullup(n, sizeof(*newmsg)); if (!n) return key_senderror(so, m, ENOBUFS); } newmsg = mtod(n, struct sadb_msg *); newmsg->sadb_msg_errno = 0; newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len); off = 0; mpolicy = m_pulldown(n, PFKEY_ALIGN8(sizeof(struct sadb_msg)), sizeof(*xpl), &off); if (mpolicy == NULL) { /* n is already freed */ return key_senderror(so, m, ENOBUFS); } xpl = (struct sadb_x_policy *)(mtod(mpolicy, caddr_t) + off); if (xpl->sadb_x_policy_exttype != SADB_X_EXT_POLICY) { m_freem(n); return key_senderror(so, m, EINVAL); } xpl->sadb_x_policy_id = newsp->id; m_freem(m); return key_sendup_mbuf(so, n, KEY_SENDUP_ALL); } } /* * get new policy id. * OUT: * 0: failure. * others: success. */ static uint32_t key_getnewspid(void) { struct secpolicy *sp; uint32_t newid = 0; int count = V_key_spi_trycnt; /* XXX */ SPTREE_WLOCK_ASSERT(); while (count--) { if (V_policy_id == ~0) /* overflowed */ newid = V_policy_id = 1; else newid = ++V_policy_id; LIST_FOREACH(sp, SPHASH_HASH(newid), idhash) { if (sp->id == newid) break; } if (sp == NULL) break; } if (count == 0 || newid == 0) { ipseclog((LOG_DEBUG, "%s: failed to allocate policy id.\n", __func__)); return (0); } return (newid); } /* * SADB_SPDDELETE processing * receive * * from the user(?), and set SADB_SASTATE_DEAD, * and send, * * to the ikmpd. * policy(*) including direction of policy. * * m will always be freed. */ static int key_spddelete(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp) { struct secpolicyindex spidx; struct sadb_address *src0, *dst0; struct sadb_x_policy *xpl0; struct secpolicy *sp; IPSEC_ASSERT(so != NULL, ("null so")); IPSEC_ASSERT(m != NULL, ("null mbuf")); IPSEC_ASSERT(mhp != NULL, ("null msghdr")); IPSEC_ASSERT(mhp->msg != NULL, ("null msg")); if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) || SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) || SADB_CHECKHDR(mhp, SADB_X_EXT_POLICY)) { ipseclog((LOG_DEBUG, "%s: invalid message: missing required header.\n", __func__)); return key_senderror(so, m, EINVAL); } if (SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) || SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST) || SADB_CHECKLEN(mhp, SADB_X_EXT_POLICY)) { ipseclog((LOG_DEBUG, "%s: invalid message: wrong header size.\n", __func__)); return key_senderror(so, m, EINVAL); } src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC]; dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST]; xpl0 = (struct sadb_x_policy *)mhp->ext[SADB_X_EXT_POLICY]; /* check the direciton */ switch (xpl0->sadb_x_policy_dir) { case IPSEC_DIR_INBOUND: case IPSEC_DIR_OUTBOUND: break; default: ipseclog((LOG_DEBUG, "%s: invalid SP direction.\n", __func__)); return key_senderror(so, m, EINVAL); } /* Only DISCARD, NONE and IPSEC are allowed */ if (xpl0->sadb_x_policy_type != IPSEC_POLICY_DISCARD && xpl0->sadb_x_policy_type != IPSEC_POLICY_NONE && xpl0->sadb_x_policy_type != IPSEC_POLICY_IPSEC) { ipseclog((LOG_DEBUG, "%s: invalid policy type.\n", __func__)); return key_senderror(so, m, EINVAL); } /* * Note: do not parse SADB_X_EXT_NAT_T_* here: * we are processing traffic endpoints. */ if (key_checksockaddrs((struct sockaddr *)(src0 + 1), (struct sockaddr *)(dst0 + 1)) != 0 || src0->sadb_address_proto != dst0->sadb_address_proto) { ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__)); return key_senderror(so, m, EINVAL); } /* make secindex */ KEY_SETSECSPIDX(xpl0->sadb_x_policy_dir, src0 + 1, dst0 + 1, src0->sadb_address_prefixlen, dst0->sadb_address_prefixlen, src0->sadb_address_proto, &spidx); /* Is there SP in SPD ? */ if ((sp = key_getsp(&spidx)) == NULL) { ipseclog((LOG_DEBUG, "%s: no SP found.\n", __func__)); return key_senderror(so, m, EINVAL); } /* save policy id to buffer to be returned. */ xpl0->sadb_x_policy_id = sp->id; KEYDBG(KEY_STAMP, printf("%s: SP(%p)\n", __func__, sp)); KEYDBG(KEY_DATA, kdebug_secpolicy(sp)); key_unlink(sp); key_freesp(&sp); { struct mbuf *n; struct sadb_msg *newmsg; /* * Note: do not send SADB_X_EXT_NAT_T_* here: * we are sending traffic endpoints. */ /* create new sadb_msg to reply. */ n = key_gather_mbuf(m, mhp, 1, 4, SADB_EXT_RESERVED, SADB_X_EXT_POLICY, SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST); if (!n) return key_senderror(so, m, ENOBUFS); newmsg = mtod(n, struct sadb_msg *); newmsg->sadb_msg_errno = 0; newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len); m_freem(m); return key_sendup_mbuf(so, n, KEY_SENDUP_ALL); } } /* * SADB_SPDDELETE2 processing * receive * * from the user(?), and set SADB_SASTATE_DEAD, * and send, * * to the ikmpd. * policy(*) including direction of policy. * * m will always be freed. */ static int key_spddelete2(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp) { struct secpolicy *sp; uint32_t id; IPSEC_ASSERT(so != NULL, ("null socket")); IPSEC_ASSERT(m != NULL, ("null mbuf")); IPSEC_ASSERT(mhp != NULL, ("null msghdr")); IPSEC_ASSERT(mhp->msg != NULL, ("null msg")); if (SADB_CHECKHDR(mhp, SADB_X_EXT_POLICY) || SADB_CHECKLEN(mhp, SADB_X_EXT_POLICY)) { ipseclog((LOG_DEBUG, "%s: invalid message is passed.\n", __func__)); return key_senderror(so, m, EINVAL); } id = ((struct sadb_x_policy *) mhp->ext[SADB_X_EXT_POLICY])->sadb_x_policy_id; /* Is there SP in SPD ? */ if ((sp = key_getspbyid(id)) == NULL) { ipseclog((LOG_DEBUG, "%s: no SP found for id %u.\n", __func__, id)); return key_senderror(so, m, EINVAL); } KEYDBG(KEY_STAMP, printf("%s: SP(%p)\n", __func__, sp)); KEYDBG(KEY_DATA, kdebug_secpolicy(sp)); key_unlink(sp); key_freesp(&sp); { struct mbuf *n, *nn; struct sadb_msg *newmsg; int off, len; /* create new sadb_msg to reply. */ len = PFKEY_ALIGN8(sizeof(struct sadb_msg)); MGETHDR(n, M_NOWAIT, MT_DATA); if (n && len > MHLEN) { if (!(MCLGET(n, M_NOWAIT))) { m_freem(n); n = NULL; } } if (!n) return key_senderror(so, m, ENOBUFS); n->m_len = len; n->m_next = NULL; off = 0; m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, caddr_t) + off); off += PFKEY_ALIGN8(sizeof(struct sadb_msg)); IPSEC_ASSERT(off == len, ("length inconsistency (off %u len %u)", off, len)); n->m_next = m_copym(m, mhp->extoff[SADB_X_EXT_POLICY], mhp->extlen[SADB_X_EXT_POLICY], M_NOWAIT); if (!n->m_next) { m_freem(n); return key_senderror(so, m, ENOBUFS); } n->m_pkthdr.len = 0; for (nn = n; nn; nn = nn->m_next) n->m_pkthdr.len += nn->m_len; newmsg = mtod(n, struct sadb_msg *); newmsg->sadb_msg_errno = 0; newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len); m_freem(m); return key_sendup_mbuf(so, n, KEY_SENDUP_ALL); } } /* * SADB_X_SPDGET processing * receive * * from the user(?), * and send, * * to the ikmpd. * policy(*) including direction of policy. * * m will always be freed. */ static int key_spdget(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp) { struct secpolicy *sp; struct mbuf *n; uint32_t id; IPSEC_ASSERT(so != NULL, ("null socket")); IPSEC_ASSERT(m != NULL, ("null mbuf")); IPSEC_ASSERT(mhp != NULL, ("null msghdr")); IPSEC_ASSERT(mhp->msg != NULL, ("null msg")); if (SADB_CHECKHDR(mhp, SADB_X_EXT_POLICY) || SADB_CHECKLEN(mhp, SADB_X_EXT_POLICY)) { ipseclog((LOG_DEBUG, "%s: invalid message is passed.\n", __func__)); return key_senderror(so, m, EINVAL); } id = ((struct sadb_x_policy *) mhp->ext[SADB_X_EXT_POLICY])->sadb_x_policy_id; /* Is there SP in SPD ? */ if ((sp = key_getspbyid(id)) == NULL) { ipseclog((LOG_DEBUG, "%s: no SP found for id %u.\n", __func__, id)); return key_senderror(so, m, ENOENT); } n = key_setdumpsp(sp, SADB_X_SPDGET, mhp->msg->sadb_msg_seq, mhp->msg->sadb_msg_pid); key_freesp(&sp); if (n != NULL) { m_freem(m); return key_sendup_mbuf(so, n, KEY_SENDUP_ONE); } else return key_senderror(so, m, ENOBUFS); } /* * SADB_X_SPDACQUIRE processing. * Acquire policy and SA(s) for a *OUTBOUND* packet. * send * * to KMD, and expect to receive * with SADB_X_SPDACQUIRE if error occurred, * or * * with SADB_X_SPDUPDATE from KMD by PF_KEY. * policy(*) is without policy requests. * * 0 : succeed * others: error number */ int key_spdacquire(struct secpolicy *sp) { struct mbuf *result = NULL, *m; struct secspacq *newspacq; IPSEC_ASSERT(sp != NULL, ("null secpolicy")); IPSEC_ASSERT(sp->req == NULL, ("policy exists")); IPSEC_ASSERT(sp->policy == IPSEC_POLICY_IPSEC, ("policy not IPSEC %u", sp->policy)); /* Get an entry to check whether sent message or not. */ newspacq = key_getspacq(&sp->spidx); if (newspacq != NULL) { if (V_key_blockacq_count < newspacq->count) { /* reset counter and do send message. */ newspacq->count = 0; } else { /* increment counter and do nothing. */ newspacq->count++; SPACQ_UNLOCK(); return (0); } SPACQ_UNLOCK(); } else { /* make new entry for blocking to send SADB_ACQUIRE. */ newspacq = key_newspacq(&sp->spidx); if (newspacq == NULL) return ENOBUFS; } /* create new sadb_msg to reply. */ m = key_setsadbmsg(SADB_X_SPDACQUIRE, 0, 0, 0, 0, 0); if (!m) return ENOBUFS; result = m; result->m_pkthdr.len = 0; for (m = result; m; m = m->m_next) result->m_pkthdr.len += m->m_len; mtod(result, struct sadb_msg *)->sadb_msg_len = PFKEY_UNIT64(result->m_pkthdr.len); return key_sendup_mbuf(NULL, m, KEY_SENDUP_REGISTERED); } /* * SADB_SPDFLUSH processing * receive * * from the user, and free all entries in secpctree. * and send, * * to the user. * NOTE: what to do is only marking SADB_SASTATE_DEAD. * * m will always be freed. */ static int key_spdflush(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp) { struct secpolicy_queue drainq; struct sadb_msg *newmsg; struct secpolicy *sp, *nextsp; u_int dir; IPSEC_ASSERT(so != NULL, ("null socket")); IPSEC_ASSERT(m != NULL, ("null mbuf")); IPSEC_ASSERT(mhp != NULL, ("null msghdr")); IPSEC_ASSERT(mhp->msg != NULL, ("null msg")); if (m->m_len != PFKEY_ALIGN8(sizeof(struct sadb_msg))) return key_senderror(so, m, EINVAL); TAILQ_INIT(&drainq); SPTREE_WLOCK(); for (dir = 0; dir < IPSEC_DIR_MAX; dir++) { TAILQ_CONCAT(&drainq, &V_sptree[dir], chain); } /* * We need to set state to DEAD for each policy to be sure, * that another thread won't try to unlink it. * Also remove SP from sphash. */ TAILQ_FOREACH(sp, &drainq, chain) { sp->state = IPSEC_SPSTATE_DEAD; LIST_REMOVE(sp, idhash); } V_sp_genid++; SPTREE_WUNLOCK(); sp = TAILQ_FIRST(&drainq); while (sp != NULL) { nextsp = TAILQ_NEXT(sp, chain); - KEY_FREESP(&sp); + key_freesp(&sp); sp = nextsp; } if (sizeof(struct sadb_msg) > m->m_len + M_TRAILINGSPACE(m)) { ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__)); return key_senderror(so, m, ENOBUFS); } if (m->m_next) m_freem(m->m_next); m->m_next = NULL; m->m_pkthdr.len = m->m_len = PFKEY_ALIGN8(sizeof(struct sadb_msg)); newmsg = mtod(m, struct sadb_msg *); newmsg->sadb_msg_errno = 0; newmsg->sadb_msg_len = PFKEY_UNIT64(m->m_pkthdr.len); return key_sendup_mbuf(so, m, KEY_SENDUP_ALL); } /* * SADB_SPDDUMP processing * receive * * from the user, and dump all SP leaves * and send, * ..... * to the ikmpd. * * m will always be freed. */ static int key_spddump(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp) { SPTREE_RLOCK_TRACKER; struct secpolicy *sp; int cnt; u_int dir; struct mbuf *n; IPSEC_ASSERT(so != NULL, ("null socket")); IPSEC_ASSERT(m != NULL, ("null mbuf")); IPSEC_ASSERT(mhp != NULL, ("null msghdr")); IPSEC_ASSERT(mhp->msg != NULL, ("null msg")); /* search SPD entry and get buffer size. */ cnt = 0; SPTREE_RLOCK(); for (dir = 0; dir < IPSEC_DIR_MAX; dir++) { TAILQ_FOREACH(sp, &V_sptree[dir], chain) { cnt++; } } if (cnt == 0) { SPTREE_RUNLOCK(); return key_senderror(so, m, ENOENT); } for (dir = 0; dir < IPSEC_DIR_MAX; dir++) { TAILQ_FOREACH(sp, &V_sptree[dir], chain) { --cnt; n = key_setdumpsp(sp, SADB_X_SPDDUMP, cnt, mhp->msg->sadb_msg_pid); if (n) key_sendup_mbuf(so, n, KEY_SENDUP_ONE); } } SPTREE_RUNLOCK(); m_freem(m); return 0; } static struct mbuf * key_setdumpsp(struct secpolicy *sp, u_int8_t type, u_int32_t seq, u_int32_t pid) { struct mbuf *result = NULL, *m; struct seclifetime lt; m = key_setsadbmsg(type, 0, SADB_SATYPE_UNSPEC, seq, pid, sp->refcnt); if (!m) goto fail; result = m; /* * Note: do not send SADB_X_EXT_NAT_T_* here: * we are sending traffic endpoints. */ m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC, &sp->spidx.src.sa, sp->spidx.prefs, sp->spidx.ul_proto); if (!m) goto fail; m_cat(result, m); m = key_setsadbaddr(SADB_EXT_ADDRESS_DST, &sp->spidx.dst.sa, sp->spidx.prefd, sp->spidx.ul_proto); if (!m) goto fail; m_cat(result, m); - m = key_sp2msg(sp); + m = key_sp2mbuf(sp); if (!m) goto fail; m_cat(result, m); if(sp->lifetime){ lt.addtime=sp->created; lt.usetime= sp->lastused; m = key_setlifetime(<, SADB_EXT_LIFETIME_CURRENT); if (!m) goto fail; m_cat(result, m); lt.addtime=sp->lifetime; lt.usetime= sp->validtime; m = key_setlifetime(<, SADB_EXT_LIFETIME_HARD); if (!m) goto fail; m_cat(result, m); } if ((result->m_flags & M_PKTHDR) == 0) goto fail; if (result->m_len < sizeof(struct sadb_msg)) { result = m_pullup(result, sizeof(struct sadb_msg)); if (result == NULL) goto fail; } result->m_pkthdr.len = 0; for (m = result; m; m = m->m_next) result->m_pkthdr.len += m->m_len; mtod(result, struct sadb_msg *)->sadb_msg_len = PFKEY_UNIT64(result->m_pkthdr.len); return result; fail: m_freem(result); return NULL; } - /* * get PFKEY message length for security policy and request. */ static size_t key_getspreqmsglen(struct secpolicy *sp) { size_t tlen, len; int i; tlen = sizeof(struct sadb_x_policy); /* if is the policy for ipsec ? */ if (sp->policy != IPSEC_POLICY_IPSEC) return (tlen); /* get length of ipsec requests */ for (i = 0; i < sp->tcount; i++) { len = sizeof(struct sadb_x_ipsecrequest) + sp->req[i]->saidx.src.sa.sa_len + sp->req[i]->saidx.dst.sa.sa_len; tlen += PFKEY_ALIGN8(len); } return (tlen); } /* * SADB_SPDEXPIRE processing * send * * to KMD by PF_KEY. * * OUT: 0 : succeed * others : error number */ static int key_spdexpire(struct secpolicy *sp) { struct sadb_lifetime *lt; struct mbuf *result = NULL, *m; int len, error = -1; IPSEC_ASSERT(sp != NULL, ("null secpolicy")); KEYDBG(KEY_STAMP, printf("%s: SP(%p)\n", __func__, sp)); KEYDBG(KEY_DATA, kdebug_secpolicy(sp)); /* set msg header */ m = key_setsadbmsg(SADB_X_SPDEXPIRE, 0, 0, 0, 0, 0); if (!m) { error = ENOBUFS; goto fail; } result = m; /* create lifetime extension (current and hard) */ len = PFKEY_ALIGN8(sizeof(*lt)) * 2; m = m_get2(len, M_NOWAIT, MT_DATA, 0); if (m == NULL) { error = ENOBUFS; goto fail; } m_align(m, len); m->m_len = len; bzero(mtod(m, caddr_t), len); lt = mtod(m, struct sadb_lifetime *); lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime)); lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_CURRENT; lt->sadb_lifetime_allocations = 0; lt->sadb_lifetime_bytes = 0; lt->sadb_lifetime_addtime = sp->created; lt->sadb_lifetime_usetime = sp->lastused; lt = (struct sadb_lifetime *)(mtod(m, caddr_t) + len / 2); lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime)); lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_HARD; lt->sadb_lifetime_allocations = 0; lt->sadb_lifetime_bytes = 0; lt->sadb_lifetime_addtime = sp->lifetime; lt->sadb_lifetime_usetime = sp->validtime; m_cat(result, m); /* * Note: do not send SADB_X_EXT_NAT_T_* here: * we are sending traffic endpoints. */ /* set sadb_address for source */ m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC, &sp->spidx.src.sa, sp->spidx.prefs, sp->spidx.ul_proto); if (!m) { error = ENOBUFS; goto fail; } m_cat(result, m); /* set sadb_address for destination */ m = key_setsadbaddr(SADB_EXT_ADDRESS_DST, &sp->spidx.dst.sa, sp->spidx.prefd, sp->spidx.ul_proto); if (!m) { error = ENOBUFS; goto fail; } m_cat(result, m); /* set secpolicy */ m = key_sp2mbuf(sp); if (!m) { error = ENOBUFS; goto fail; } m_cat(result, m); if ((result->m_flags & M_PKTHDR) == 0) { error = EINVAL; goto fail; } if (result->m_len < sizeof(struct sadb_msg)) { result = m_pullup(result, sizeof(struct sadb_msg)); if (result == NULL) { error = ENOBUFS; goto fail; } } result->m_pkthdr.len = 0; for (m = result; m; m = m->m_next) result->m_pkthdr.len += m->m_len; mtod(result, struct sadb_msg *)->sadb_msg_len = PFKEY_UNIT64(result->m_pkthdr.len); return key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED); fail: if (result) m_freem(result); return error; } /* %%% SAD management */ /* * allocating and initialize new SA head. * OUT: NULL : failure due to the lack of memory. * others : pointer to new SA head. */ static struct secashead * key_newsah(struct secasindex *saidx) { struct secashead *sah; sah = malloc(sizeof(struct secashead), M_IPSEC_SAH, M_NOWAIT | M_ZERO); if (sah == NULL) return (NULL); TAILQ_INIT(&sah->savtree_larval); TAILQ_INIT(&sah->savtree_alive); sah->saidx = *saidx; sah->state = SADB_SASTATE_DEAD; SAH_INITREF(sah); KEYDBG(KEY_STAMP, printf("%s: SAH(%p)\n", __func__, sah)); KEYDBG(KEY_DATA, kdebug_secash(sah, NULL)); return (sah); } static void key_freesah(struct secashead **psah) { struct secashead *sah = *psah; if (SAH_DELREF(sah) == 0) return; KEYDBG(KEY_STAMP, printf("%s: last reference to SAH(%p)\n", __func__, sah)); KEYDBG(KEY_DATA, kdebug_secash(sah, NULL)); *psah = NULL; key_delsah(sah); } static void key_delsah(struct secashead *sah) { IPSEC_ASSERT(sah != NULL, ("NULL sah")); IPSEC_ASSERT(sah->state == SADB_SASTATE_DEAD, ("Attempt to free non DEAD SAH %p", sah)); IPSEC_ASSERT(TAILQ_EMPTY(&sah->savtree_larval), ("Attempt to free SAH %p with LARVAL SA", sah)); IPSEC_ASSERT(TAILQ_EMPTY(&sah->savtree_alive), ("Attempt to free SAH %p with ALIVE SA", sah)); free(sah, M_IPSEC_SAH); } /* * allocating a new SA for key_add() and key_getspi() call, * and copy the values of mhp into new buffer. * When SAD message type is SADB_GETSPI set SA state to LARVAL. * For SADB_ADD create and initialize SA with MATURE state. * OUT: NULL : fail * others : pointer to new secasvar. */ static struct secasvar * key_newsav(const struct sadb_msghdr *mhp, struct secasindex *saidx, uint32_t spi, int *errp) { struct secashead *sah; struct secasvar *sav; int isnew; IPSEC_ASSERT(mhp != NULL, ("null msghdr")); IPSEC_ASSERT(mhp->msg != NULL, ("null msg")); IPSEC_ASSERT(mhp->msg->sadb_msg_type == SADB_GETSPI || mhp->msg->sadb_msg_type == SADB_ADD, ("wrong message type")); sav = NULL; sah = NULL; /* check SPI value */ switch (saidx->proto) { case IPPROTO_ESP: case IPPROTO_AH: /* * RFC 4302, 2.4. Security Parameters Index (SPI), SPI values * 1-255 reserved by IANA for future use, * 0 for implementation specific, local use. */ if (ntohl(spi) <= 255) { ipseclog((LOG_DEBUG, "%s: illegal range of SPI %u.\n", __func__, ntohl(spi))); *errp = EINVAL; goto done; } break; } sav = malloc(sizeof(struct secasvar), M_IPSEC_SA, M_NOWAIT | M_ZERO); if (sav == NULL) { ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__)); *errp = ENOBUFS; goto done; } sav->lft_c = uma_zalloc(V_key_lft_zone, M_NOWAIT); if (sav->lft_c == NULL) { ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__)); free(sav, M_IPSEC_SA), sav = NULL; *errp = ENOBUFS; goto done; } counter_u64_zero(sav->lft_c_allocations); counter_u64_zero(sav->lft_c_bytes); sav->spi = spi; sav->seq = mhp->msg->sadb_msg_seq; sav->state = SADB_SASTATE_LARVAL; sav->pid = (pid_t)mhp->msg->sadb_msg_pid; SAV_INITREF(sav); SECASVAR_LOCK_INIT(sav); again: sah = key_getsah(saidx); if (sah == NULL) { /* create a new SA index */ sah = key_newsah(saidx); if (sah == NULL) { ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__)); *errp = ENOBUFS; goto done; } isnew = 1; } else isnew = 0; sav->sah = sah; if (mhp->msg->sadb_msg_type == SADB_GETSPI) { sav->created = time_second; } else if (sav->state == SADB_SASTATE_LARVAL) { /* * Do not call key_setsaval() second time in case * of `goto again`. We will have MATURE state. */ *errp = key_setsaval(sav, mhp); if (*errp != 0) goto done; sav->state = SADB_SASTATE_MATURE; } SAHTREE_WLOCK(); /* * Check that existing SAH wasn't unlinked. * Since we didn't hold the SAHTREE lock, it is possible, * that callout handler or key_flush() or key_delete() could * unlink this SAH. */ if (isnew == 0 && sah->state == SADB_SASTATE_DEAD) { SAHTREE_WUNLOCK(); key_freesah(&sah); /* reference from key_getsah() */ goto again; } if (isnew != 0) { /* * Add new SAH into SADB. * * XXXAE: we can serialize key_add and key_getspi calls, so * several threads will not fight in the race. * Otherwise we should check under SAHTREE lock, that this * SAH would not added twice. */ TAILQ_INSERT_HEAD(&V_sahtree, sah, chain); /* Add new SAH into hash by addresses */ LIST_INSERT_HEAD(SAHADDRHASH_HASH(saidx), sah, addrhash); /* Now we are linked in the chain */ sah->state = SADB_SASTATE_MATURE; /* * SAV references this new SAH. * In case of existing SAH we reuse reference * from key_getsah(). */ SAH_ADDREF(sah); } /* Link SAV with SAH */ if (sav->state == SADB_SASTATE_MATURE) TAILQ_INSERT_HEAD(&sah->savtree_alive, sav, chain); else TAILQ_INSERT_HEAD(&sah->savtree_larval, sav, chain); /* Add SAV into SPI hash */ LIST_INSERT_HEAD(SAVHASH_HASH(sav->spi), sav, spihash); SAHTREE_WUNLOCK(); *errp = 0; /* success */ done: if (*errp != 0) { if (sav != NULL) { SECASVAR_LOCK_DESTROY(sav); uma_zfree(V_key_lft_zone, sav->lft_c); free(sav, M_IPSEC_SA), sav = NULL; } if (sah != NULL) key_freesah(&sah); } return (sav); } /* * free() SA variable entry. */ static void key_cleansav(struct secasvar *sav) { /* * Cleanup xform state. Note that zeroize'ing causes the * keys to be cleared; otherwise we must do it ourself. */ if (sav->tdb_xform != NULL) { sav->tdb_xform->xf_zeroize(sav); sav->tdb_xform = NULL; } else { if (sav->key_auth != NULL) bzero(sav->key_auth->key_data, _KEYLEN(sav->key_auth)); if (sav->key_enc != NULL) bzero(sav->key_enc->key_data, _KEYLEN(sav->key_enc)); } if (sav->key_auth != NULL) { if (sav->key_auth->key_data != NULL) free(sav->key_auth->key_data, M_IPSEC_MISC); free(sav->key_auth, M_IPSEC_MISC); sav->key_auth = NULL; } if (sav->key_enc != NULL) { if (sav->key_enc->key_data != NULL) free(sav->key_enc->key_data, M_IPSEC_MISC); free(sav->key_enc, M_IPSEC_MISC); sav->key_enc = NULL; } if (sav->replay != NULL) { free(sav->replay, M_IPSEC_MISC); sav->replay = NULL; } if (sav->lft_h != NULL) { free(sav->lft_h, M_IPSEC_MISC); sav->lft_h = NULL; } if (sav->lft_s != NULL) { free(sav->lft_s, M_IPSEC_MISC); sav->lft_s = NULL; } } /* * free() SA variable entry. */ static void key_delsav(struct secasvar *sav) { IPSEC_ASSERT(sav != NULL, ("null sav")); IPSEC_ASSERT(sav->state == SADB_SASTATE_DEAD, ("attempt to free non DEAD SA %p", sav)); IPSEC_ASSERT(sav->refcnt == 0, ("reference count %u > 0", sav->refcnt)); /* SA must be unlinked from the chain and hashtbl */ key_cleansav(sav); SECASVAR_LOCK_DESTROY(sav); uma_zfree(V_key_lft_zone, sav->lft_c); free(sav, M_IPSEC_SA); } /* * search SAH. * OUT: * NULL : not found * others : found, referenced pointer to a SAH. */ static struct secashead * key_getsah(struct secasindex *saidx) { SAHTREE_RLOCK_TRACKER; struct secashead *sah; SAHTREE_RLOCK(); LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) { if (key_cmpsaidx(&sah->saidx, saidx, CMP_MODE_REQID) != 0) { SAH_ADDREF(sah); break; } } SAHTREE_RUNLOCK(); return (sah); } /* * Check not to be duplicated SPI. * OUT: * 0 : not found * 1 : found SA with given SPI. */ static int key_checkspidup(uint32_t spi) { SAHTREE_RLOCK_TRACKER; struct secasvar *sav; /* Assume SPI is in network byte order */ SAHTREE_RLOCK(); LIST_FOREACH(sav, SAVHASH_HASH(spi), spihash) { if (sav->spi == spi) break; } SAHTREE_RUNLOCK(); return (sav != NULL); } /* * Search SA by SPI. * OUT: * NULL : not found * others : found, referenced pointer to a SA. */ static struct secasvar * key_getsavbyspi(uint32_t spi) { SAHTREE_RLOCK_TRACKER; struct secasvar *sav; /* Assume SPI is in network byte order */ SAHTREE_RLOCK(); LIST_FOREACH(sav, SAVHASH_HASH(spi), spihash) { if (sav->spi != spi) continue; SAV_ADDREF(sav); break; } SAHTREE_RUNLOCK(); return (sav); } static int key_updatelifetimes(struct secasvar *sav, const struct sadb_msghdr *mhp) { struct seclifetime *lft_h, *lft_s, *tmp; /* Lifetime extension is optional, check that it is present. */ if (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD) && SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT)) { /* * In case of SADB_UPDATE we may need to change * existing lifetimes. */ if (sav->state == SADB_SASTATE_MATURE) { lft_h = lft_s = NULL; goto reset; } return (0); } /* XXXAE: what should we do with CURRENT lifetime? */ /* Both HARD and SOFT extensions must present */ if ((SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD) && !SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT)) || (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT) && !SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD))) { ipseclog((LOG_DEBUG, "%s: invalid message: missing required header.\n", __func__)); return (EINVAL); } if (SADB_CHECKLEN(mhp, SADB_EXT_LIFETIME_HARD) || SADB_CHECKLEN(mhp, SADB_EXT_LIFETIME_SOFT)) { ipseclog((LOG_DEBUG, "%s: invalid message: wrong header size.\n", __func__)); return (EINVAL); } lft_h = key_dup_lifemsg((const struct sadb_lifetime *) mhp->ext[SADB_EXT_LIFETIME_HARD], M_IPSEC_MISC); if (lft_h == NULL) { PFKEYSTAT_INC(in_nomem); ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__)); return (ENOBUFS); } lft_s = key_dup_lifemsg((const struct sadb_lifetime *) mhp->ext[SADB_EXT_LIFETIME_SOFT], M_IPSEC_MISC); if (lft_s == NULL) { PFKEYSTAT_INC(in_nomem); free(lft_h, M_IPSEC_MISC); ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__)); return (ENOBUFS); } reset: if (sav->state != SADB_SASTATE_LARVAL) { /* * key_update() holds reference to this SA, * so it won't be deleted in meanwhile. */ SECASVAR_LOCK(sav); tmp = sav->lft_h; sav->lft_h = lft_h; lft_h = tmp; tmp = sav->lft_s; sav->lft_s = lft_s; lft_s = tmp; SECASVAR_UNLOCK(sav); if (lft_h != NULL) free(lft_h, M_IPSEC_MISC); if (lft_s != NULL) free(lft_s, M_IPSEC_MISC); return (0); } /* We can update lifetime without holding a lock */ IPSEC_ASSERT(sav->lft_h == NULL, ("lft_h is already initialized\n")); IPSEC_ASSERT(sav->lft_s == NULL, ("lft_s is already initialized\n")); sav->lft_h = lft_h; sav->lft_s = lft_s; return (0); } /* * copy SA values from PF_KEY message except *SPI, SEQ, PID and TYPE*. * You must update these if need. Expects only LARVAL SAs. * OUT: 0: success. * !0: failure. */ static int key_setsaval(struct secasvar *sav, const struct sadb_msghdr *mhp) { const struct sadb_sa *sa0; const struct sadb_key *key0; size_t len; int error; IPSEC_ASSERT(mhp != NULL, ("null msghdr")); IPSEC_ASSERT(mhp->msg != NULL, ("null msg")); IPSEC_ASSERT(sav->state == SADB_SASTATE_LARVAL, ("Attempt to update non LARVAL SA")); /* XXX rewrite */ error = key_setident(sav->sah, mhp); if (error != 0) goto fail; error = key_setnatt(sav, mhp); if (error != 0) goto fail; /* SA */ if (!SADB_CHECKHDR(mhp, SADB_EXT_SA)) { if (SADB_CHECKLEN(mhp, SADB_EXT_SA)) { error = EINVAL; goto fail; } sa0 = (const struct sadb_sa *)mhp->ext[SADB_EXT_SA]; sav->alg_auth = sa0->sadb_sa_auth; sav->alg_enc = sa0->sadb_sa_encrypt; sav->flags = sa0->sadb_sa_flags; /* replay window */ if ((sa0->sadb_sa_flags & SADB_X_EXT_OLD) == 0) { sav->replay = malloc(sizeof(struct secreplay) + sa0->sadb_sa_replay, M_IPSEC_MISC, M_NOWAIT | M_ZERO); if (sav->replay == NULL) { PFKEYSTAT_INC(in_nomem); ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__)); error = ENOBUFS; goto fail; } if (sa0->sadb_sa_replay != 0) sav->replay->bitmap = (caddr_t)(sav->replay+1); sav->replay->wsize = sa0->sadb_sa_replay; } } /* Authentication keys */ if (!SADB_CHECKHDR(mhp, SADB_EXT_KEY_AUTH)) { if (SADB_CHECKLEN(mhp, SADB_EXT_KEY_AUTH)) { error = EINVAL; goto fail; } error = 0; key0 = (const struct sadb_key *)mhp->ext[SADB_EXT_KEY_AUTH]; len = mhp->extlen[SADB_EXT_KEY_AUTH]; switch (mhp->msg->sadb_msg_satype) { case SADB_SATYPE_AH: case SADB_SATYPE_ESP: case SADB_X_SATYPE_TCPSIGNATURE: if (len == PFKEY_ALIGN8(sizeof(struct sadb_key)) && sav->alg_auth != SADB_X_AALG_NULL) error = EINVAL; break; case SADB_X_SATYPE_IPCOMP: default: error = EINVAL; break; } if (error) { ipseclog((LOG_DEBUG, "%s: invalid key_auth values.\n", __func__)); goto fail; } sav->key_auth = key_dup_keymsg(key0, len, M_IPSEC_MISC); if (sav->key_auth == NULL ) { ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__)); PFKEYSTAT_INC(in_nomem); error = ENOBUFS; goto fail; } } /* Encryption key */ if (!SADB_CHECKHDR(mhp, SADB_EXT_KEY_ENCRYPT)) { if (SADB_CHECKLEN(mhp, SADB_EXT_KEY_ENCRYPT)) { error = EINVAL; goto fail; } error = 0; key0 = (const struct sadb_key *)mhp->ext[SADB_EXT_KEY_ENCRYPT]; len = mhp->extlen[SADB_EXT_KEY_ENCRYPT]; switch (mhp->msg->sadb_msg_satype) { case SADB_SATYPE_ESP: if (len == PFKEY_ALIGN8(sizeof(struct sadb_key)) && sav->alg_enc != SADB_EALG_NULL) { error = EINVAL; break; } sav->key_enc = key_dup_keymsg(key0, len, M_IPSEC_MISC); if (sav->key_enc == NULL) { ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__)); PFKEYSTAT_INC(in_nomem); error = ENOBUFS; goto fail; } break; case SADB_X_SATYPE_IPCOMP: if (len != PFKEY_ALIGN8(sizeof(struct sadb_key))) error = EINVAL; sav->key_enc = NULL; /*just in case*/ break; case SADB_SATYPE_AH: case SADB_X_SATYPE_TCPSIGNATURE: default: error = EINVAL; break; } if (error) { ipseclog((LOG_DEBUG, "%s: invalid key_enc value.\n", __func__)); goto fail; } } /* set iv */ sav->ivlen = 0; switch (mhp->msg->sadb_msg_satype) { case SADB_SATYPE_AH: if (sav->flags & SADB_X_EXT_DERIV) { ipseclog((LOG_DEBUG, "%s: invalid flag (derived) " "given to AH SA.\n", __func__)); error = EINVAL; goto fail; } if (sav->alg_enc != SADB_EALG_NONE) { ipseclog((LOG_DEBUG, "%s: protocol and algorithm " "mismated.\n", __func__)); error = EINVAL; goto fail; } error = xform_init(sav, XF_AH); break; case SADB_SATYPE_ESP: if ((sav->flags & (SADB_X_EXT_OLD | SADB_X_EXT_DERIV)) == (SADB_X_EXT_OLD | SADB_X_EXT_DERIV)) { ipseclog((LOG_DEBUG, "%s: invalid flag (derived) " "given to old-esp.\n", __func__)); error = EINVAL; goto fail; } error = xform_init(sav, XF_ESP); break; case SADB_X_SATYPE_IPCOMP: if (sav->alg_auth != SADB_AALG_NONE) { ipseclog((LOG_DEBUG, "%s: protocol and algorithm " "mismated.\n", __func__)); error = EINVAL; goto fail; } if ((sav->flags & SADB_X_EXT_RAWCPI) == 0 && ntohl(sav->spi) >= 0x10000) { ipseclog((LOG_DEBUG, "%s: invalid cpi for IPComp.\n", __func__)); error = EINVAL; goto fail; } error = xform_init(sav, XF_IPCOMP); break; case SADB_X_SATYPE_TCPSIGNATURE: if (sav->alg_enc != SADB_EALG_NONE) { ipseclog((LOG_DEBUG, "%s: protocol and algorithm " "mismated.\n", __func__)); error = EINVAL; goto fail; } error = xform_init(sav, XF_TCPSIGNATURE); break; default: ipseclog((LOG_DEBUG, "%s: Invalid satype.\n", __func__)); error = EPROTONOSUPPORT; goto fail; } if (error) { ipseclog((LOG_DEBUG, "%s: unable to initialize SA type %u.\n", __func__, mhp->msg->sadb_msg_satype)); goto fail; } /* Initialize lifetime for CURRENT */ sav->firstused = 0; sav->created = time_second; /* lifetimes for HARD and SOFT */ error = key_updatelifetimes(sav, mhp); if (error == 0) return (0); fail: key_cleansav(sav); return (error); } /* * subroutine for SADB_GET and SADB_DUMP. */ static struct mbuf * key_setdumpsa(struct secasvar *sav, uint8_t type, uint8_t satype, uint32_t seq, uint32_t pid) { struct seclifetime lft_c; struct mbuf *result = NULL, *tres = NULL, *m; int i, dumporder[] = { SADB_EXT_SA, SADB_X_EXT_SA2, SADB_EXT_LIFETIME_HARD, SADB_EXT_LIFETIME_SOFT, SADB_EXT_LIFETIME_CURRENT, SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST, SADB_EXT_ADDRESS_PROXY, SADB_EXT_KEY_AUTH, SADB_EXT_KEY_ENCRYPT, SADB_EXT_IDENTITY_SRC, SADB_EXT_IDENTITY_DST, SADB_EXT_SENSITIVITY, #ifdef IPSEC_NAT_T SADB_X_EXT_NAT_T_TYPE, SADB_X_EXT_NAT_T_SPORT, SADB_X_EXT_NAT_T_DPORT, SADB_X_EXT_NAT_T_OAI, SADB_X_EXT_NAT_T_OAR, SADB_X_EXT_NAT_T_FRAG, #endif }; m = key_setsadbmsg(type, 0, satype, seq, pid, sav->refcnt); if (m == NULL) goto fail; result = m; for (i = nitems(dumporder) - 1; i >= 0; i--) { m = NULL; switch (dumporder[i]) { case SADB_EXT_SA: m = key_setsadbsa(sav); if (!m) goto fail; break; case SADB_X_EXT_SA2: m = key_setsadbxsa2(sav->sah->saidx.mode, sav->replay ? sav->replay->count : 0, sav->sah->saidx.reqid); if (!m) goto fail; break; case SADB_EXT_ADDRESS_SRC: m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC, &sav->sah->saidx.src.sa, FULLMASK, IPSEC_ULPROTO_ANY); if (!m) goto fail; break; case SADB_EXT_ADDRESS_DST: m = key_setsadbaddr(SADB_EXT_ADDRESS_DST, &sav->sah->saidx.dst.sa, FULLMASK, IPSEC_ULPROTO_ANY); if (!m) goto fail; break; case SADB_EXT_KEY_AUTH: if (!sav->key_auth) continue; m = key_setkey(sav->key_auth, SADB_EXT_KEY_AUTH); if (!m) goto fail; break; case SADB_EXT_KEY_ENCRYPT: if (!sav->key_enc) continue; m = key_setkey(sav->key_enc, SADB_EXT_KEY_ENCRYPT); if (!m) goto fail; break; case SADB_EXT_LIFETIME_CURRENT: lft_c.addtime = sav->created; lft_c.allocations = (uint32_t)counter_u64_fetch( sav->lft_c_allocations); lft_c.bytes = counter_u64_fetch(sav->lft_c_bytes); lft_c.usetime = sav->firstused; m = key_setlifetime(&lft_c, SADB_EXT_LIFETIME_CURRENT); if (!m) goto fail; break; case SADB_EXT_LIFETIME_HARD: if (!sav->lft_h) continue; m = key_setlifetime(sav->lft_h, SADB_EXT_LIFETIME_HARD); if (!m) goto fail; break; case SADB_EXT_LIFETIME_SOFT: if (!sav->lft_s) continue; m = key_setlifetime(sav->lft_s, SADB_EXT_LIFETIME_SOFT); if (!m) goto fail; break; #ifdef IPSEC_NAT_T case SADB_X_EXT_NAT_T_TYPE: m = key_setsadbxtype(sav->natt_type); if (!m) goto fail; break; case SADB_X_EXT_NAT_T_DPORT: m = key_setsadbxport( key_portfromsaddr(&sav->sah->saidx.dst.sa), SADB_X_EXT_NAT_T_DPORT); if (!m) goto fail; break; case SADB_X_EXT_NAT_T_SPORT: m = key_setsadbxport( key_portfromsaddr(&sav->sah->saidx.src.sa), SADB_X_EXT_NAT_T_SPORT); if (!m) goto fail; break; case SADB_X_EXT_NAT_T_OAI: case SADB_X_EXT_NAT_T_OAR: case SADB_X_EXT_NAT_T_FRAG: /* We do not (yet) support those. */ continue; #endif case SADB_EXT_ADDRESS_PROXY: case SADB_EXT_IDENTITY_SRC: case SADB_EXT_IDENTITY_DST: /* XXX: should we brought from SPD ? */ case SADB_EXT_SENSITIVITY: default: continue; } if (!m) goto fail; if (tres) m_cat(m, tres); tres = m; } m_cat(result, tres); tres = NULL; if (result->m_len < sizeof(struct sadb_msg)) { result = m_pullup(result, sizeof(struct sadb_msg)); if (result == NULL) goto fail; } result->m_pkthdr.len = 0; for (m = result; m; m = m->m_next) result->m_pkthdr.len += m->m_len; mtod(result, struct sadb_msg *)->sadb_msg_len = PFKEY_UNIT64(result->m_pkthdr.len); return result; fail: m_freem(result); m_freem(tres); return NULL; } /* * set data into sadb_msg. */ static struct mbuf * key_setsadbmsg(u_int8_t type, u_int16_t tlen, u_int8_t satype, u_int32_t seq, pid_t pid, u_int16_t reserved) { struct mbuf *m; struct sadb_msg *p; int len; len = PFKEY_ALIGN8(sizeof(struct sadb_msg)); if (len > MCLBYTES) return NULL; MGETHDR(m, M_NOWAIT, MT_DATA); if (m && len > MHLEN) { if (!(MCLGET(m, M_NOWAIT))) { m_freem(m); m = NULL; } } if (!m) return NULL; m->m_pkthdr.len = m->m_len = len; m->m_next = NULL; p = mtod(m, struct sadb_msg *); bzero(p, len); p->sadb_msg_version = PF_KEY_V2; p->sadb_msg_type = type; p->sadb_msg_errno = 0; p->sadb_msg_satype = satype; p->sadb_msg_len = PFKEY_UNIT64(tlen); p->sadb_msg_reserved = reserved; p->sadb_msg_seq = seq; p->sadb_msg_pid = (u_int32_t)pid; return m; } /* * copy secasvar data into sadb_address. */ static struct mbuf * key_setsadbsa(struct secasvar *sav) { struct mbuf *m; struct sadb_sa *p; int len; len = PFKEY_ALIGN8(sizeof(struct sadb_sa)); m = m_get2(len, M_NOWAIT, MT_DATA, 0); if (m == NULL) return (NULL); m_align(m, len); m->m_len = len; p = mtod(m, struct sadb_sa *); bzero(p, len); p->sadb_sa_len = PFKEY_UNIT64(len); p->sadb_sa_exttype = SADB_EXT_SA; p->sadb_sa_spi = sav->spi; p->sadb_sa_replay = (sav->replay != NULL ? sav->replay->wsize : 0); p->sadb_sa_state = sav->state; p->sadb_sa_auth = sav->alg_auth; p->sadb_sa_encrypt = sav->alg_enc; p->sadb_sa_flags = sav->flags; return m; } /* * set data into sadb_address. */ static struct mbuf * key_setsadbaddr(u_int16_t exttype, const struct sockaddr *saddr, u_int8_t prefixlen, u_int16_t ul_proto) { struct mbuf *m; struct sadb_address *p; size_t len; len = PFKEY_ALIGN8(sizeof(struct sadb_address)) + PFKEY_ALIGN8(saddr->sa_len); m = m_get2(len, M_NOWAIT, MT_DATA, 0); if (m == NULL) return (NULL); m_align(m, len); m->m_len = len; p = mtod(m, struct sadb_address *); bzero(p, len); p->sadb_address_len = PFKEY_UNIT64(len); p->sadb_address_exttype = exttype; p->sadb_address_proto = ul_proto; if (prefixlen == FULLMASK) { switch (saddr->sa_family) { case AF_INET: prefixlen = sizeof(struct in_addr) << 3; break; case AF_INET6: prefixlen = sizeof(struct in6_addr) << 3; break; default: ; /*XXX*/ } } p->sadb_address_prefixlen = prefixlen; p->sadb_address_reserved = 0; bcopy(saddr, mtod(m, caddr_t) + PFKEY_ALIGN8(sizeof(struct sadb_address)), saddr->sa_len); return m; } /* * set data into sadb_x_sa2. */ static struct mbuf * key_setsadbxsa2(u_int8_t mode, u_int32_t seq, u_int32_t reqid) { struct mbuf *m; struct sadb_x_sa2 *p; size_t len; len = PFKEY_ALIGN8(sizeof(struct sadb_x_sa2)); m = m_get2(len, M_NOWAIT, MT_DATA, 0); if (m == NULL) return (NULL); m_align(m, len); m->m_len = len; p = mtod(m, struct sadb_x_sa2 *); bzero(p, len); p->sadb_x_sa2_len = PFKEY_UNIT64(len); p->sadb_x_sa2_exttype = SADB_X_EXT_SA2; p->sadb_x_sa2_mode = mode; p->sadb_x_sa2_reserved1 = 0; p->sadb_x_sa2_reserved2 = 0; p->sadb_x_sa2_sequence = seq; p->sadb_x_sa2_reqid = reqid; return m; } #ifdef IPSEC_NAT_T /* * Set a type in sadb_x_nat_t_type. */ static struct mbuf * key_setsadbxtype(u_int16_t type) { struct mbuf *m; size_t len; struct sadb_x_nat_t_type *p; len = PFKEY_ALIGN8(sizeof(struct sadb_x_nat_t_type)); m = m_get2(len, M_NOWAIT, MT_DATA, 0); if (m == NULL) return (NULL); m_align(m, len); m->m_len = len; p = mtod(m, struct sadb_x_nat_t_type *); bzero(p, len); p->sadb_x_nat_t_type_len = PFKEY_UNIT64(len); p->sadb_x_nat_t_type_exttype = SADB_X_EXT_NAT_T_TYPE; p->sadb_x_nat_t_type_type = type; return (m); } /* * Set a port in sadb_x_nat_t_port. * In contrast to default RFC 2367 behaviour, port is in network byte order. */ static struct mbuf * key_setsadbxport(u_int16_t port, u_int16_t type) { struct mbuf *m; size_t len; struct sadb_x_nat_t_port *p; len = PFKEY_ALIGN8(sizeof(struct sadb_x_nat_t_port)); m = m_get2(len, M_NOWAIT, MT_DATA, 0); if (m == NULL) return (NULL); m_align(m, len); m->m_len = len; p = mtod(m, struct sadb_x_nat_t_port *); bzero(p, len); p->sadb_x_nat_t_port_len = PFKEY_UNIT64(len); p->sadb_x_nat_t_port_exttype = type; p->sadb_x_nat_t_port_port = port; return (m); } /* * Get port from sockaddr. Port is in network byte order. */ uint16_t key_portfromsaddr(struct sockaddr *sa) { switch (sa->sa_family) { #ifdef INET case AF_INET: return ((struct sockaddr_in *)sa)->sin_port; #endif #ifdef INET6 case AF_INET6: return ((struct sockaddr_in6 *)sa)->sin6_port; #endif } return (0); } #endif /* IPSEC_NAT_T */ /* * Set port in struct sockaddr. Port is in network byte order. */ static void key_porttosaddr(struct sockaddr *sa, uint16_t port) { switch (sa->sa_family) { #ifdef INET case AF_INET: ((struct sockaddr_in *)sa)->sin_port = port; break; #endif #ifdef INET6 case AF_INET6: ((struct sockaddr_in6 *)sa)->sin6_port = port; break; #endif default: ipseclog((LOG_DEBUG, "%s: unexpected address family %d.\n", __func__, sa->sa_family)); break; } } /* * set data into sadb_x_policy */ static struct mbuf * key_setsadbxpolicy(u_int16_t type, u_int8_t dir, u_int32_t id, u_int32_t priority) { struct mbuf *m; struct sadb_x_policy *p; size_t len; len = PFKEY_ALIGN8(sizeof(struct sadb_x_policy)); m = m_get2(len, M_NOWAIT, MT_DATA, 0); if (m == NULL) return (NULL); m_align(m, len); m->m_len = len; p = mtod(m, struct sadb_x_policy *); bzero(p, len); p->sadb_x_policy_len = PFKEY_UNIT64(len); p->sadb_x_policy_exttype = SADB_X_EXT_POLICY; p->sadb_x_policy_type = type; p->sadb_x_policy_dir = dir; p->sadb_x_policy_id = id; p->sadb_x_policy_priority = priority; return m; } /* %%% utilities */ /* Take a key message (sadb_key) from the socket and turn it into one * of the kernel's key structures (seckey). * * IN: pointer to the src * OUT: NULL no more memory */ struct seckey * -key_dup_keymsg(const struct sadb_key *src, u_int len, +key_dup_keymsg(const struct sadb_key *src, size_t len, struct malloc_type *type) { struct seckey *dst; - dst = (struct seckey *)malloc(sizeof(struct seckey), type, M_NOWAIT); + + dst = malloc(sizeof(*dst), type, M_NOWAIT); if (dst != NULL) { dst->bits = src->sadb_key_bits; - dst->key_data = (char *)malloc(len, type, M_NOWAIT); + dst->key_data = malloc(len, type, M_NOWAIT); if (dst->key_data != NULL) { - bcopy((const char *)src + sizeof(struct sadb_key), - dst->key_data, len); + bcopy((const char *)(src + 1), dst->key_data, len); } else { - ipseclog((LOG_DEBUG, "%s: No more memory.\n", - __func__)); + ipseclog((LOG_DEBUG, "%s: No more memory.\n", + __func__)); free(dst, type); dst = NULL; } } else { - ipseclog((LOG_DEBUG, "%s: No more memory.\n", - __func__)); + ipseclog((LOG_DEBUG, "%s: No more memory.\n", + __func__)); } - return dst; + return (dst); } /* Take a lifetime message (sadb_lifetime) passed in on a socket and * turn it into one of the kernel's lifetime structures (seclifetime). * * IN: pointer to the destination, source and malloc type * OUT: NULL, no more memory */ static struct seclifetime * key_dup_lifemsg(const struct sadb_lifetime *src, struct malloc_type *type) { struct seclifetime *dst; dst = malloc(sizeof(*dst), type, M_NOWAIT); if (dst == NULL) { ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__)); return (NULL); } dst->allocations = src->sadb_lifetime_allocations; dst->bytes = src->sadb_lifetime_bytes; dst->addtime = src->sadb_lifetime_addtime; dst->usetime = src->sadb_lifetime_usetime; return (dst); } -/* compare my own address - * OUT: 1: true, i.e. my address. - * 0: false - */ -int -key_ismyaddr(struct sockaddr *sa) -{ - - IPSEC_ASSERT(sa != NULL, ("null sockaddr")); - switch (sa->sa_family) { -#ifdef INET - case AF_INET: - return (in_localip(satosin(sa)->sin_addr)); -#endif -#ifdef INET6 - case AF_INET6: - return key_ismyaddr6((struct sockaddr_in6 *)sa); -#endif - } - - return 0; -} - -#ifdef INET6 /* - * compare my own address for IPv6. - * 1: ours - * 0: other - */ -static int -key_ismyaddr6(struct sockaddr_in6 *sin6) -{ - struct in6_addr in6; - - if (!IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr)) - return (in6_localip(&sin6->sin6_addr)); - - /* Convert address into kernel-internal form */ - in6 = sin6->sin6_addr; - in6.s6_addr16[1] = htons(sin6->sin6_scope_id & 0xffff); - return (in6_localip(&in6)); -} -#endif /*INET6*/ - -/* * compare two secasindex structure. * flag can specify to compare 2 saidxes. * compare two secasindex structure without both mode and reqid. * don't compare port. * IN: * saidx0: source, it can be in SAD. * saidx1: object. * OUT: * 1 : equal * 0 : not equal */ static int key_cmpsaidx(const struct secasindex *saidx0, const struct secasindex *saidx1, int flag) { int chkport = 0; /* sanity */ if (saidx0 == NULL && saidx1 == NULL) return 1; if (saidx0 == NULL || saidx1 == NULL) return 0; if (saidx0->proto != saidx1->proto) return 0; if (flag == CMP_EXACTLY) { if (saidx0->mode != saidx1->mode) return 0; if (saidx0->reqid != saidx1->reqid) return 0; if (bcmp(&saidx0->src, &saidx1->src, saidx0->src.sa.sa_len) != 0 || bcmp(&saidx0->dst, &saidx1->dst, saidx0->dst.sa.sa_len) != 0) return 0; } else { /* CMP_MODE_REQID, CMP_REQID, CMP_HEAD */ if (flag == CMP_MODE_REQID ||flag == CMP_REQID) { /* * If reqid of SPD is non-zero, unique SA is required. * The result must be of same reqid in this case. */ if (saidx1->reqid != 0 && saidx0->reqid != saidx1->reqid) return 0; } if (flag == CMP_MODE_REQID) { if (saidx0->mode != IPSEC_MODE_ANY && saidx0->mode != saidx1->mode) return 0; } #ifdef IPSEC_NAT_T /* * If NAT-T is enabled, check ports for tunnel mode. * Do not check ports if they are set to zero in the SPD. * Also do not do it for native transport mode, as there * is no port information available in the SP. */ if ((saidx1->mode == IPSEC_MODE_TUNNEL || (saidx1->mode == IPSEC_MODE_TRANSPORT && saidx1->proto == IPPROTO_ESP)) && saidx1->src.sa.sa_family == AF_INET && saidx1->dst.sa.sa_family == AF_INET && ((const struct sockaddr_in *)(&saidx1->src))->sin_port && ((const struct sockaddr_in *)(&saidx1->dst))->sin_port) chkport = 1; #endif /* IPSEC_NAT_T */ if (key_sockaddrcmp(&saidx0->src.sa, &saidx1->src.sa, chkport) != 0) { return 0; } if (key_sockaddrcmp(&saidx0->dst.sa, &saidx1->dst.sa, chkport) != 0) { return 0; } } return 1; } /* * compare two secindex structure exactly. * IN: * spidx0: source, it is often in SPD. * spidx1: object, it is often from PFKEY message. * OUT: * 1 : equal * 0 : not equal */ static int key_cmpspidx_exactly(struct secpolicyindex *spidx0, struct secpolicyindex *spidx1) { /* sanity */ if (spidx0 == NULL && spidx1 == NULL) return 1; if (spidx0 == NULL || spidx1 == NULL) return 0; if (spidx0->prefs != spidx1->prefs || spidx0->prefd != spidx1->prefd || spidx0->ul_proto != spidx1->ul_proto) return 0; return key_sockaddrcmp(&spidx0->src.sa, &spidx1->src.sa, 1) == 0 && key_sockaddrcmp(&spidx0->dst.sa, &spidx1->dst.sa, 1) == 0; } /* * compare two secindex structure with mask. * IN: * spidx0: source, it is often in SPD. * spidx1: object, it is often from IP header. * OUT: * 1 : equal * 0 : not equal */ static int key_cmpspidx_withmask(struct secpolicyindex *spidx0, struct secpolicyindex *spidx1) { /* sanity */ if (spidx0 == NULL && spidx1 == NULL) return 1; if (spidx0 == NULL || spidx1 == NULL) return 0; if (spidx0->src.sa.sa_family != spidx1->src.sa.sa_family || spidx0->dst.sa.sa_family != spidx1->dst.sa.sa_family || spidx0->src.sa.sa_len != spidx1->src.sa.sa_len || spidx0->dst.sa.sa_len != spidx1->dst.sa.sa_len) return 0; /* if spidx.ul_proto == IPSEC_ULPROTO_ANY, ignore. */ if (spidx0->ul_proto != (u_int16_t)IPSEC_ULPROTO_ANY && spidx0->ul_proto != spidx1->ul_proto) return 0; switch (spidx0->src.sa.sa_family) { case AF_INET: if (spidx0->src.sin.sin_port != IPSEC_PORT_ANY && spidx0->src.sin.sin_port != spidx1->src.sin.sin_port) return 0; if (!key_bbcmp(&spidx0->src.sin.sin_addr, &spidx1->src.sin.sin_addr, spidx0->prefs)) return 0; break; case AF_INET6: if (spidx0->src.sin6.sin6_port != IPSEC_PORT_ANY && spidx0->src.sin6.sin6_port != spidx1->src.sin6.sin6_port) return 0; /* * scope_id check. if sin6_scope_id is 0, we regard it * as a wildcard scope, which matches any scope zone ID. */ if (spidx0->src.sin6.sin6_scope_id && spidx1->src.sin6.sin6_scope_id && spidx0->src.sin6.sin6_scope_id != spidx1->src.sin6.sin6_scope_id) return 0; if (!key_bbcmp(&spidx0->src.sin6.sin6_addr, &spidx1->src.sin6.sin6_addr, spidx0->prefs)) return 0; break; default: /* XXX */ if (bcmp(&spidx0->src, &spidx1->src, spidx0->src.sa.sa_len) != 0) return 0; break; } switch (spidx0->dst.sa.sa_family) { case AF_INET: if (spidx0->dst.sin.sin_port != IPSEC_PORT_ANY && spidx0->dst.sin.sin_port != spidx1->dst.sin.sin_port) return 0; if (!key_bbcmp(&spidx0->dst.sin.sin_addr, &spidx1->dst.sin.sin_addr, spidx0->prefd)) return 0; break; case AF_INET6: if (spidx0->dst.sin6.sin6_port != IPSEC_PORT_ANY && spidx0->dst.sin6.sin6_port != spidx1->dst.sin6.sin6_port) return 0; /* * scope_id check. if sin6_scope_id is 0, we regard it * as a wildcard scope, which matches any scope zone ID. */ if (spidx0->dst.sin6.sin6_scope_id && spidx1->dst.sin6.sin6_scope_id && spidx0->dst.sin6.sin6_scope_id != spidx1->dst.sin6.sin6_scope_id) return 0; if (!key_bbcmp(&spidx0->dst.sin6.sin6_addr, &spidx1->dst.sin6.sin6_addr, spidx0->prefd)) return 0; break; default: /* XXX */ if (bcmp(&spidx0->dst, &spidx1->dst, spidx0->dst.sa.sa_len) != 0) return 0; break; } /* XXX Do we check other field ? e.g. flowinfo */ return 1; } #ifdef satosin #undef satosin #endif #define satosin(s) ((const struct sockaddr_in *)s) #ifdef satosin6 #undef satosin6 #endif #define satosin6(s) ((const struct sockaddr_in6 *)s) /* returns 0 on match */ -static int +int key_sockaddrcmp(const struct sockaddr *sa1, const struct sockaddr *sa2, int port) { if (sa1->sa_family != sa2->sa_family || sa1->sa_len != sa2->sa_len) return 1; switch (sa1->sa_family) { #ifdef INET case AF_INET: if (sa1->sa_len != sizeof(struct sockaddr_in)) return 1; if (satosin(sa1)->sin_addr.s_addr != satosin(sa2)->sin_addr.s_addr) { return 1; } if (port && satosin(sa1)->sin_port != satosin(sa2)->sin_port) return 1; break; #endif #ifdef INET6 case AF_INET6: if (sa1->sa_len != sizeof(struct sockaddr_in6)) return 1; /*EINVAL*/ if (satosin6(sa1)->sin6_scope_id != satosin6(sa2)->sin6_scope_id) { return 1; } if (!IN6_ARE_ADDR_EQUAL(&satosin6(sa1)->sin6_addr, &satosin6(sa2)->sin6_addr)) { return 1; } if (port && satosin6(sa1)->sin6_port != satosin6(sa2)->sin6_port) { return 1; } break; #endif default: if (bcmp(sa1, sa2, sa1->sa_len) != 0) return 1; break; } return 0; } /* returns 0 on match */ int key_sockaddrcmp_withmask(const struct sockaddr *sa1, const struct sockaddr *sa2, size_t mask) { if (sa1->sa_family != sa2->sa_family || sa1->sa_len != sa2->sa_len) return (1); switch (sa1->sa_family) { #ifdef INET case AF_INET: return (!key_bbcmp(&satosin(sa1)->sin_addr, &satosin(sa2)->sin_addr, mask)); #endif #ifdef INET6 case AF_INET6: if (satosin6(sa1)->sin6_scope_id != satosin6(sa2)->sin6_scope_id) return (1); return (!key_bbcmp(&satosin6(sa1)->sin6_addr, &satosin6(sa2)->sin6_addr, mask)); #endif } return (1); } #undef satosin #undef satosin6 /* * compare two buffers with mask. * IN: * addr1: source * addr2: object * bits: Number of bits to compare * OUT: * 1 : equal * 0 : not equal */ static int key_bbcmp(const void *a1, const void *a2, u_int bits) { const unsigned char *p1 = a1; const unsigned char *p2 = a2; /* XXX: This could be considerably faster if we compare a word * at a time, but it is complicated on LSB Endian machines */ /* Handle null pointers */ if (p1 == NULL || p2 == NULL) return (p1 == p2); while (bits >= 8) { if (*p1++ != *p2++) return 0; bits -= 8; } if (bits > 0) { u_int8_t mask = ~((1<<(8-bits))-1); if ((*p1 & mask) != (*p2 & mask)) return 0; } return 1; /* Match! */ } static void key_flush_spd(time_t now) { SPTREE_RLOCK_TRACKER; struct secpolicy_list drainq; struct secpolicy *sp, *nextsp; u_int dir; LIST_INIT(&drainq); SPTREE_RLOCK(); for (dir = 0; dir < IPSEC_DIR_MAX; dir++) { TAILQ_FOREACH(sp, &V_sptree[dir], chain) { if (sp->lifetime == 0 && sp->validtime == 0) continue; if ((sp->lifetime && now - sp->created > sp->lifetime) || (sp->validtime && now - sp->lastused > sp->validtime)) { /* Hold extra reference to send SPDEXPIRE */ SP_ADDREF(sp); LIST_INSERT_HEAD(&drainq, sp, drainq); } } } SPTREE_RUNLOCK(); if (LIST_EMPTY(&drainq)) return; SPTREE_WLOCK(); sp = LIST_FIRST(&drainq); while (sp != NULL) { nextsp = LIST_NEXT(sp, drainq); /* Check that SP is still linked */ if (sp->state != IPSEC_SPSTATE_ALIVE) { LIST_REMOVE(sp, drainq); key_freesp(&sp); /* release extra reference */ sp = nextsp; continue; } TAILQ_REMOVE(&V_sptree[sp->spidx.dir], sp, chain); LIST_REMOVE(sp, idhash); sp->state = IPSEC_SPSTATE_DEAD; sp = nextsp; } V_sp_genid++; SPTREE_WUNLOCK(); sp = LIST_FIRST(&drainq); while (sp != NULL) { nextsp = LIST_NEXT(sp, drainq); key_spdexpire(sp); key_freesp(&sp); /* release extra reference */ key_freesp(&sp); /* release last reference */ sp = nextsp; } } static void key_flush_sad(time_t now) { SAHTREE_RLOCK_TRACKER; struct secashead_list emptyq; struct secasvar_list drainq, hexpireq, sexpireq, freeq; struct secashead *sah, *nextsah; struct secasvar *sav, *nextsav; LIST_INIT(&drainq); LIST_INIT(&hexpireq); LIST_INIT(&sexpireq); LIST_INIT(&emptyq); SAHTREE_RLOCK(); TAILQ_FOREACH(sah, &V_sahtree, chain) { /* Check for empty SAH */ if (TAILQ_EMPTY(&sah->savtree_larval) && TAILQ_EMPTY(&sah->savtree_alive)) { SAH_ADDREF(sah); LIST_INSERT_HEAD(&emptyq, sah, drainq); continue; } /* Add all stale LARVAL SAs into drainq */ TAILQ_FOREACH(sav, &sah->savtree_larval, chain) { if (now - sav->created < V_key_larval_lifetime) continue; SAV_ADDREF(sav); LIST_INSERT_HEAD(&drainq, sav, drainq); } TAILQ_FOREACH(sav, &sah->savtree_alive, chain) { /* lifetimes aren't specified */ if (sav->lft_h == NULL) continue; SECASVAR_LOCK(sav); /* * Check again with lock held, because it may * be updated by SADB_UPDATE. */ if (sav->lft_h == NULL) { SECASVAR_UNLOCK(sav); continue; } /* * RFC 2367: * HARD lifetimes MUST take precedence over SOFT * lifetimes, meaning if the HARD and SOFT lifetimes * are the same, the HARD lifetime will appear on the * EXPIRE message. */ /* check HARD lifetime */ if ((sav->lft_h->addtime != 0 && now - sav->created > sav->lft_h->addtime) || (sav->lft_h->usetime != 0 && sav->firstused && now - sav->firstused > sav->lft_h->usetime) || (sav->lft_h->bytes != 0 && counter_u64_fetch( sav->lft_c_bytes) > sav->lft_h->bytes)) { SECASVAR_UNLOCK(sav); SAV_ADDREF(sav); LIST_INSERT_HEAD(&hexpireq, sav, drainq); continue; } /* check SOFT lifetime (only for MATURE SAs) */ if (sav->state == SADB_SASTATE_MATURE && ( (sav->lft_s->addtime != 0 && now - sav->created > sav->lft_s->addtime) || (sav->lft_s->usetime != 0 && sav->firstused && now - sav->firstused > sav->lft_s->usetime) || (sav->lft_s->bytes != 0 && counter_u64_fetch( sav->lft_c_bytes) > sav->lft_s->bytes))) { SECASVAR_UNLOCK(sav); SAV_ADDREF(sav); LIST_INSERT_HEAD(&sexpireq, sav, drainq); continue; } SECASVAR_UNLOCK(sav); } } SAHTREE_RUNLOCK(); if (LIST_EMPTY(&emptyq) && LIST_EMPTY(&drainq) && LIST_EMPTY(&hexpireq) && LIST_EMPTY(&sexpireq)) return; LIST_INIT(&freeq); SAHTREE_WLOCK(); /* Unlink stale LARVAL SAs */ sav = LIST_FIRST(&drainq); while (sav != NULL) { nextsav = LIST_NEXT(sav, drainq); /* Check that SA is still LARVAL */ if (sav->state != SADB_SASTATE_LARVAL) { LIST_REMOVE(sav, drainq); LIST_INSERT_HEAD(&freeq, sav, drainq); sav = nextsav; continue; } TAILQ_REMOVE(&sav->sah->savtree_larval, sav, chain); LIST_REMOVE(sav, spihash); sav->state = SADB_SASTATE_DEAD; sav = nextsav; } /* Unlink all SAs with expired HARD lifetime */ sav = LIST_FIRST(&hexpireq); while (sav != NULL) { nextsav = LIST_NEXT(sav, drainq); /* Check that SA is not unlinked */ if (sav->state == SADB_SASTATE_DEAD) { LIST_REMOVE(sav, drainq); LIST_INSERT_HEAD(&freeq, sav, drainq); sav = nextsav; continue; } TAILQ_REMOVE(&sav->sah->savtree_alive, sav, chain); LIST_REMOVE(sav, spihash); sav->state = SADB_SASTATE_DEAD; sav = nextsav; } /* Mark all SAs with expired SOFT lifetime as DYING */ sav = LIST_FIRST(&sexpireq); while (sav != NULL) { nextsav = LIST_NEXT(sav, drainq); /* Check that SA is not unlinked */ if (sav->state == SADB_SASTATE_DEAD) { LIST_REMOVE(sav, drainq); LIST_INSERT_HEAD(&freeq, sav, drainq); sav = nextsav; continue; } /* * NOTE: this doesn't change SA order in the chain. */ sav->state = SADB_SASTATE_DYING; sav = nextsav; } /* Unlink empty SAHs */ sah = LIST_FIRST(&emptyq); while (sah != NULL) { nextsah = LIST_NEXT(sah, drainq); /* Check that SAH is still empty and not unlinked */ if (sah->state == SADB_SASTATE_DEAD || !TAILQ_EMPTY(&sah->savtree_larval) || !TAILQ_EMPTY(&sah->savtree_alive)) { LIST_REMOVE(sah, drainq); key_freesah(&sah); /* release extra reference */ sah = nextsah; continue; } TAILQ_REMOVE(&V_sahtree, sah, chain); LIST_REMOVE(sah, addrhash); sah->state = SADB_SASTATE_DEAD; sah = nextsah; } SAHTREE_WUNLOCK(); /* Send SPDEXPIRE messages */ sav = LIST_FIRST(&hexpireq); while (sav != NULL) { nextsav = LIST_NEXT(sav, drainq); key_expire(sav, 1); key_freesah(&sav->sah); /* release reference from SAV */ key_freesav(&sav); /* release extra reference */ key_freesav(&sav); /* release last reference */ sav = nextsav; } sav = LIST_FIRST(&sexpireq); while (sav != NULL) { nextsav = LIST_NEXT(sav, drainq); key_expire(sav, 0); key_freesav(&sav); /* release extra reference */ sav = nextsav; } /* Free stale LARVAL SAs */ sav = LIST_FIRST(&drainq); while (sav != NULL) { nextsav = LIST_NEXT(sav, drainq); key_freesah(&sav->sah); /* release reference from SAV */ key_freesav(&sav); /* release extra reference */ key_freesav(&sav); /* release last reference */ sav = nextsav; } /* Free SAs that were unlinked/changed by someone else */ sav = LIST_FIRST(&freeq); while (sav != NULL) { nextsav = LIST_NEXT(sav, drainq); key_freesav(&sav); /* release extra reference */ sav = nextsav; } /* Free empty SAH */ sah = LIST_FIRST(&emptyq); while (sah != NULL) { nextsah = LIST_NEXT(sah, drainq); key_freesah(&sah); /* release extra reference */ key_freesah(&sah); /* release last reference */ sah = nextsah; } } static void key_flush_acq(time_t now) { struct secacq *acq, *nextacq; /* ACQ tree */ ACQ_LOCK(); acq = LIST_FIRST(&V_acqtree); while (acq != NULL) { nextacq = LIST_NEXT(acq, chain); if (now - acq->created > V_key_blockacq_lifetime) { LIST_REMOVE(acq, chain); LIST_REMOVE(acq, addrhash); LIST_REMOVE(acq, seqhash); free(acq, M_IPSEC_SAQ); } acq = nextacq; } ACQ_UNLOCK(); } static void key_flush_spacq(time_t now) { struct secspacq *acq, *nextacq; /* SP ACQ tree */ SPACQ_LOCK(); for (acq = LIST_FIRST(&V_spacqtree); acq != NULL; acq = nextacq) { nextacq = LIST_NEXT(acq, chain); if (now - acq->created > V_key_blockacq_lifetime && __LIST_CHAINED(acq)) { LIST_REMOVE(acq, chain); free(acq, M_IPSEC_SAQ); } } SPACQ_UNLOCK(); } /* * time handler. * scanning SPD and SAD to check status for each entries, * and do to remove or to expire. * XXX: year 2038 problem may remain. */ static void key_timehandler(void *arg) { VNET_ITERATOR_DECL(vnet_iter); time_t now = time_second; VNET_LIST_RLOCK_NOSLEEP(); VNET_FOREACH(vnet_iter) { CURVNET_SET(vnet_iter); key_flush_spd(now); key_flush_sad(now); key_flush_acq(now); key_flush_spacq(now); CURVNET_RESTORE(); } VNET_LIST_RUNLOCK_NOSLEEP(); #ifndef IPSEC_DEBUG2 /* do exchange to tick time !! */ callout_schedule(&key_timer, hz); #endif /* IPSEC_DEBUG2 */ } u_long key_random() { u_long value; key_randomfill(&value, sizeof(value)); return value; } void key_randomfill(void *p, size_t l) { size_t n; u_long v; static int warn = 1; n = 0; n = (size_t)read_random(p, (u_int)l); /* last resort */ while (n < l) { v = random(); bcopy(&v, (u_int8_t *)p + n, l - n < sizeof(v) ? l - n : sizeof(v)); n += sizeof(v); if (warn) { printf("WARNING: pseudo-random number generator " "used for IPsec processing\n"); warn = 0; } } } /* * map SADB_SATYPE_* to IPPROTO_*. * if satype == SADB_SATYPE then satype is mapped to ~0. * OUT: * 0: invalid satype. */ static uint8_t key_satype2proto(uint8_t satype) { switch (satype) { case SADB_SATYPE_UNSPEC: return IPSEC_PROTO_ANY; case SADB_SATYPE_AH: return IPPROTO_AH; case SADB_SATYPE_ESP: return IPPROTO_ESP; case SADB_X_SATYPE_IPCOMP: return IPPROTO_IPCOMP; case SADB_X_SATYPE_TCPSIGNATURE: return IPPROTO_TCP; default: return 0; } /* NOTREACHED */ } /* * map IPPROTO_* to SADB_SATYPE_* * OUT: * 0: invalid protocol type. */ static uint8_t key_proto2satype(uint8_t proto) { switch (proto) { case IPPROTO_AH: return SADB_SATYPE_AH; case IPPROTO_ESP: return SADB_SATYPE_ESP; case IPPROTO_IPCOMP: return SADB_X_SATYPE_IPCOMP; case IPPROTO_TCP: return SADB_X_SATYPE_TCPSIGNATURE; default: return 0; } /* NOTREACHED */ } /* %%% PF_KEY */ /* * SADB_GETSPI processing is to receive * * from the IKMPd, to assign a unique spi value, to hang on the INBOUND * tree with the status of LARVAL, and send * * to the IKMPd. * * IN: mhp: pointer to the pointer to each header. * OUT: NULL if fail. * other if success, return pointer to the message to send. */ static int key_getspi(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp) { struct secasindex saidx; struct sadb_address *src0, *dst0; struct secasvar *sav; uint32_t reqid, spi; int error; uint8_t mode, proto; IPSEC_ASSERT(so != NULL, ("null socket")); IPSEC_ASSERT(m != NULL, ("null mbuf")); IPSEC_ASSERT(mhp != NULL, ("null msghdr")); IPSEC_ASSERT(mhp->msg != NULL, ("null msg")); if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) || SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) #ifdef PFKEY_STRICT_CHECKS || SADB_CHECKHDR(mhp, SADB_EXT_SPIRANGE) #endif ) { ipseclog((LOG_DEBUG, "%s: invalid message: missing required header.\n", __func__)); error = EINVAL; goto fail; } if (SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) || SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST) #ifdef PFKEY_STRICT_CHECKS || SADB_CHECKLEN(mhp, SADB_EXT_SPIRANGE) #endif ) { ipseclog((LOG_DEBUG, "%s: invalid message: wrong header size.\n", __func__)); error = EINVAL; goto fail; } if (SADB_CHECKHDR(mhp, SADB_X_EXT_SA2)) { mode = IPSEC_MODE_ANY; reqid = 0; } else { if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA2)) { ipseclog((LOG_DEBUG, "%s: invalid message: wrong header size.\n", __func__)); error = EINVAL; goto fail; } mode = ((struct sadb_x_sa2 *) mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_mode; reqid = ((struct sadb_x_sa2 *) mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_reqid; } src0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_SRC]); dst0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_DST]); /* map satype to proto */ if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) { ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n", __func__)); error = EINVAL; goto fail; } error = key_checksockaddrs((struct sockaddr *)(src0 + 1), (struct sockaddr *)(dst0 + 1)); if (error != 0) { ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__)); error = EINVAL; goto fail; } KEY_SETSECASIDX(proto, mode, reqid, src0 + 1, dst0 + 1, &saidx); /* * Make sure the port numbers are zero. * In case of NAT-T we will update them later if needed. */ KEY_PORTTOSADDR(&saidx.src, 0); KEY_PORTTOSADDR(&saidx.dst, 0); /* SPI allocation */ spi = key_do_getnewspi( (struct sadb_spirange *)mhp->ext[SADB_EXT_SPIRANGE], &saidx); if (spi == 0) { /* * Requested SPI or SPI range is not available or * already used. */ error = EEXIST; goto fail; } sav = key_newsav(mhp, &saidx, spi, &error); if (sav == NULL) goto fail; if (sav->seq != 0) { /* * RFC2367: * If the SADB_GETSPI message is in response to a * kernel-generated SADB_ACQUIRE, the sadb_msg_seq * MUST be the same as the SADB_ACQUIRE message. * * XXXAE: However it doesn't definethe behaviour how to * check this and what to do if it doesn't match. * Also what we should do if it matches? * * We can compare saidx used in SADB_ACQUIRE with saidx * used in SADB_GETSPI, but this probably can break * existing software. For now just warn if it doesn't match. * * XXXAE: anyway it looks useless. */ key_acqdone(&saidx, sav->seq); } KEYDBG(KEY_STAMP, printf("%s: SA(%p)\n", __func__, sav)); KEYDBG(KEY_DATA, kdebug_secasv(sav)); { struct mbuf *n, *nn; struct sadb_sa *m_sa; struct sadb_msg *newmsg; int off, len; /* create new sadb_msg to reply. */ len = PFKEY_ALIGN8(sizeof(struct sadb_msg)) + PFKEY_ALIGN8(sizeof(struct sadb_sa)); MGETHDR(n, M_NOWAIT, MT_DATA); if (len > MHLEN) { if (!(MCLGET(n, M_NOWAIT))) { m_freem(n); n = NULL; } } if (!n) { error = ENOBUFS; goto fail; } n->m_len = len; n->m_next = NULL; off = 0; m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, caddr_t) + off); off += PFKEY_ALIGN8(sizeof(struct sadb_msg)); m_sa = (struct sadb_sa *)(mtod(n, caddr_t) + off); m_sa->sadb_sa_len = PFKEY_UNIT64(sizeof(struct sadb_sa)); m_sa->sadb_sa_exttype = SADB_EXT_SA; m_sa->sadb_sa_spi = spi; /* SPI is already in network byte order */ off += PFKEY_ALIGN8(sizeof(struct sadb_sa)); IPSEC_ASSERT(off == len, ("length inconsistency (off %u len %u)", off, len)); n->m_next = key_gather_mbuf(m, mhp, 0, 2, SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST); if (!n->m_next) { m_freem(n); error = ENOBUFS; goto fail; } if (n->m_len < sizeof(struct sadb_msg)) { n = m_pullup(n, sizeof(struct sadb_msg)); if (n == NULL) return key_sendup_mbuf(so, m, KEY_SENDUP_ONE); } n->m_pkthdr.len = 0; for (nn = n; nn; nn = nn->m_next) n->m_pkthdr.len += nn->m_len; newmsg = mtod(n, struct sadb_msg *); newmsg->sadb_msg_seq = sav->seq; newmsg->sadb_msg_errno = 0; newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len); m_freem(m); return key_sendup_mbuf(so, n, KEY_SENDUP_ONE); } fail: return (key_senderror(so, m, error)); } /* * allocating new SPI * called by key_getspi(). * OUT: * 0: failure. * others: success, SPI in network byte order. */ static uint32_t key_do_getnewspi(struct sadb_spirange *spirange, struct secasindex *saidx) { uint32_t min, max, newspi, t; int count = V_key_spi_trycnt; /* set spi range to allocate */ if (spirange != NULL) { min = spirange->sadb_spirange_min; max = spirange->sadb_spirange_max; } else { min = V_key_spi_minval; max = V_key_spi_maxval; } /* IPCOMP needs 2-byte SPI */ if (saidx->proto == IPPROTO_IPCOMP) { if (min >= 0x10000) min = 0xffff; if (max >= 0x10000) max = 0xffff; if (min > max) { t = min; min = max; max = t; } } if (min == max) { if (!key_checkspidup(htonl(min))) { ipseclog((LOG_DEBUG, "%s: SPI %u exists already.\n", __func__, min)); return 0; } count--; /* taking one cost. */ newspi = min; } else { /* init SPI */ newspi = 0; /* when requesting to allocate spi ranged */ while (count--) { /* generate pseudo-random SPI value ranged. */ newspi = min + (key_random() % (max - min + 1)); if (!key_checkspidup(htonl(newspi))) break; } if (count == 0 || newspi == 0) { ipseclog((LOG_DEBUG, "%s: failed to allocate SPI.\n", __func__)); return 0; } } /* statistics */ keystat.getspi_count = (keystat.getspi_count + V_key_spi_trycnt - count) / 2; return (htonl(newspi)); } /* * SADB_UPDATE processing * receive * * from the ikmpd, and update a secasvar entry whose status is SADB_SASTATE_LARVAL. * and send * * to the ikmpd. * * m will always be freed. */ static int key_update(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp) { struct secasindex saidx; struct sadb_address *src0, *dst0; struct sadb_sa *sa0; struct secasvar *sav; uint32_t reqid; int error; uint8_t mode, proto; IPSEC_ASSERT(so != NULL, ("null socket")); IPSEC_ASSERT(m != NULL, ("null mbuf")); IPSEC_ASSERT(mhp != NULL, ("null msghdr")); IPSEC_ASSERT(mhp->msg != NULL, ("null msg")); /* map satype to proto */ if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) { ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n", __func__)); return key_senderror(so, m, EINVAL); } if (SADB_CHECKHDR(mhp, SADB_EXT_SA) || SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) || SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) || (mhp->msg->sadb_msg_satype == SADB_SATYPE_ESP && ( SADB_CHECKHDR(mhp, SADB_EXT_KEY_ENCRYPT) || SADB_CHECKLEN(mhp, SADB_EXT_KEY_ENCRYPT))) || (mhp->msg->sadb_msg_satype == SADB_SATYPE_AH && ( SADB_CHECKHDR(mhp, SADB_EXT_KEY_AUTH) || SADB_CHECKLEN(mhp, SADB_EXT_KEY_AUTH))) || (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD) && !SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT)) || (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT) && !SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD))) { ipseclog((LOG_DEBUG, "%s: invalid message: missing required header.\n", __func__)); return key_senderror(so, m, EINVAL); } if (SADB_CHECKLEN(mhp, SADB_EXT_SA) || SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) || SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)) { ipseclog((LOG_DEBUG, "%s: invalid message: wrong header size.\n", __func__)); return key_senderror(so, m, EINVAL); } if (SADB_CHECKHDR(mhp, SADB_X_EXT_SA2)) { mode = IPSEC_MODE_ANY; reqid = 0; } else { if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA2)) { ipseclog((LOG_DEBUG, "%s: invalid message: wrong header size.\n", __func__)); return key_senderror(so, m, EINVAL); } mode = ((struct sadb_x_sa2 *) mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_mode; reqid = ((struct sadb_x_sa2 *) mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_reqid; } sa0 = (struct sadb_sa *)mhp->ext[SADB_EXT_SA]; src0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_SRC]); dst0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_DST]); /* * Only SADB_SASTATE_MATURE SAs may be submitted in an * SADB_UPDATE message. */ if (sa0->sadb_sa_state != SADB_SASTATE_MATURE) { ipseclog((LOG_DEBUG, "%s: invalid state.\n", __func__)); #ifdef PFKEY_STRICT_CHECKS return key_senderror(so, m, EINVAL); #endif } error = key_checksockaddrs((struct sockaddr *)(src0 + 1), (struct sockaddr *)(dst0 + 1)); if (error != 0) { ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__)); return key_senderror(so, m, error); } KEY_SETSECASIDX(proto, mode, reqid, src0 + 1, dst0 + 1, &saidx); /* * Make sure the port numbers are zero. * In case of NAT-T we will update them later if needed. */ KEY_PORTTOSADDR(&saidx.src, 0); KEY_PORTTOSADDR(&saidx.dst, 0); sav = key_getsavbyspi(sa0->sadb_sa_spi); if (sav == NULL) { ipseclog((LOG_DEBUG, "%s: no SA found for SPI %u\n", __func__, ntohl(sa0->sadb_sa_spi))); return key_senderror(so, m, EINVAL); } /* * Check that SADB_UPDATE issued by the same process that did * SADB_GETSPI or SADB_ADD. */ if (sav->pid != mhp->msg->sadb_msg_pid) { ipseclog((LOG_DEBUG, "%s: pid mismatched (SPI %u, pid %u vs. %u)\n", __func__, ntohl(sav->spi), sav->pid, mhp->msg->sadb_msg_pid)); key_freesav(&sav); return key_senderror(so, m, EINVAL); } /* * XXXAE: saidx should match with SA. Use CMP_MODE_REQID since we * didn't set ports for NAT-T yet and exactly match may fail. */ if (key_cmpsaidx(&sav->sah->saidx, &saidx, CMP_MODE_REQID) == 0) { ipseclog((LOG_DEBUG, "%s: saidx mismatched for SPI %u", __func__, ntohl(sav->spi))); key_freesav(&sav); return key_senderror(so, m, ESRCH); } if (sav->state == SADB_SASTATE_LARVAL) { /* * We can set any values except src, dst and SPI. */ error = key_setsaval(sav, mhp); if (error != 0) { key_freesav(&sav); return (key_senderror(so, m, error)); } /* Change SA state to MATURE */ SAHTREE_WLOCK(); if (sav->state != SADB_SASTATE_LARVAL) { /* SA was deleted or another thread made it MATURE. */ SAHTREE_WUNLOCK(); key_freesav(&sav); return (key_senderror(so, m, ESRCH)); } /* * NOTE: we keep SAs in savtree_alive ordered by created * time. When SA's state changed from LARVAL to MATURE, * we update its created time in key_setsaval() and move * it into head of savtree_alive. */ TAILQ_REMOVE(&sav->sah->savtree_larval, sav, chain); TAILQ_INSERT_HEAD(&sav->sah->savtree_alive, sav, chain); sav->state = SADB_SASTATE_MATURE; SAHTREE_WUNLOCK(); } else { /* * For DYING and MATURE SA we can change only state * and lifetimes. Report EINVAL if something else attempted * to change. */ if (!SADB_CHECKHDR(mhp, SADB_EXT_KEY_ENCRYPT) || !SADB_CHECKHDR(mhp, SADB_EXT_KEY_AUTH)) { key_freesav(&sav); return (key_senderror(so, m, EINVAL)); } error = key_updatelifetimes(sav, mhp); if (error != 0) { key_freesav(&sav); return (key_senderror(so, m, error)); } /* Check that SA is still alive */ SAHTREE_WLOCK(); if (sav->state == SADB_SASTATE_DEAD) { /* SA was unlinked */ SAHTREE_WUNLOCK(); key_freesav(&sav); return (key_senderror(so, m, ESRCH)); } /* * NOTE: there is possible state moving from DYING to MATURE, * but this doesn't change created time, so we won't reorder * this SA. */ sav->state = SADB_SASTATE_MATURE; SAHTREE_WUNLOCK(); } KEYDBG(KEY_STAMP, printf("%s: SA(%p)\n", __func__, sav)); KEYDBG(KEY_DATA, kdebug_secasv(sav)); key_freesav(&sav); { struct mbuf *n; /* set msg buf from mhp */ n = key_getmsgbuf_x1(m, mhp); if (n == NULL) { ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__)); return key_senderror(so, m, ENOBUFS); } m_freem(m); return key_sendup_mbuf(so, n, KEY_SENDUP_ALL); } } /* * SADB_ADD processing * add an entry to SA database, when received * * from the ikmpd, * and send * * to the ikmpd. * * IGNORE identity and sensitivity messages. * * m will always be freed. */ static int key_add(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp) { struct secasindex saidx; struct sadb_address *src0, *dst0; struct sadb_sa *sa0; struct secasvar *sav; uint32_t reqid; uint8_t mode, proto; int error; IPSEC_ASSERT(so != NULL, ("null socket")); IPSEC_ASSERT(m != NULL, ("null mbuf")); IPSEC_ASSERT(mhp != NULL, ("null msghdr")); IPSEC_ASSERT(mhp->msg != NULL, ("null msg")); /* map satype to proto */ if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) { ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n", __func__)); return key_senderror(so, m, EINVAL); } if (SADB_CHECKHDR(mhp, SADB_EXT_SA) || SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) || SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) || (mhp->msg->sadb_msg_satype == SADB_SATYPE_ESP && ( SADB_CHECKHDR(mhp, SADB_EXT_KEY_ENCRYPT) || SADB_CHECKLEN(mhp, SADB_EXT_KEY_ENCRYPT))) || (mhp->msg->sadb_msg_satype == SADB_SATYPE_AH && ( SADB_CHECKHDR(mhp, SADB_EXT_KEY_AUTH) || SADB_CHECKLEN(mhp, SADB_EXT_KEY_AUTH))) || (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD) && !SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT)) || (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT) && !SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD))) { ipseclog((LOG_DEBUG, "%s: invalid message: missing required header.\n", __func__)); return key_senderror(so, m, EINVAL); } if (SADB_CHECKLEN(mhp, SADB_EXT_SA) || SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) || SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)) { ipseclog((LOG_DEBUG, "%s: invalid message: wrong header size.\n", __func__)); return key_senderror(so, m, EINVAL); } if (SADB_CHECKHDR(mhp, SADB_X_EXT_SA2)) { mode = IPSEC_MODE_ANY; reqid = 0; } else { if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA2)) { ipseclog((LOG_DEBUG, "%s: invalid message: wrong header size.\n", __func__)); return key_senderror(so, m, EINVAL); } mode = ((struct sadb_x_sa2 *) mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_mode; reqid = ((struct sadb_x_sa2 *) mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_reqid; } sa0 = (struct sadb_sa *)mhp->ext[SADB_EXT_SA]; src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC]; dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST]; /* * Only SADB_SASTATE_MATURE SAs may be submitted in an * SADB_ADD message. */ if (sa0->sadb_sa_state != SADB_SASTATE_MATURE) { ipseclog((LOG_DEBUG, "%s: invalid state.\n", __func__)); #ifdef PFKEY_STRICT_CHECKS return key_senderror(so, m, EINVAL); #endif } error = key_checksockaddrs((struct sockaddr *)(src0 + 1), (struct sockaddr *)(dst0 + 1)); if (error != 0) { ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__)); return key_senderror(so, m, error); } KEY_SETSECASIDX(proto, mode, reqid, src0 + 1, dst0 + 1, &saidx); /* * Make sure the port numbers are zero. * In case of NAT-T we will update them later if needed. */ KEY_PORTTOSADDR(&saidx.src, 0); KEY_PORTTOSADDR(&saidx.dst, 0); /* We can create new SA only if SPI is different. */ sav = key_getsavbyspi(sa0->sadb_sa_spi); if (sav != NULL) { key_freesav(&sav); ipseclog((LOG_DEBUG, "%s: SA already exists.\n", __func__)); return key_senderror(so, m, EEXIST); } sav = key_newsav(mhp, &saidx, sa0->sadb_sa_spi, &error); if (sav == NULL) return key_senderror(so, m, error); KEYDBG(KEY_STAMP, printf("%s: return SA(%p)\n", __func__, sav)); KEYDBG(KEY_DATA, kdebug_secasv(sav)); /* * If SADB_ADD was in response to SADB_ACQUIRE, we need to schedule * ACQ for deletion. */ if (sav->seq != 0) key_acqdone(&saidx, sav->seq); { /* * Don't call key_freesav() on error here, as we would like to * keep the SA in the database. */ struct mbuf *n; /* set msg buf from mhp */ n = key_getmsgbuf_x1(m, mhp); if (n == NULL) { ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__)); return key_senderror(so, m, ENOBUFS); } m_freem(m); return key_sendup_mbuf(so, n, KEY_SENDUP_ALL); } } static int key_setnatt(struct secasvar *sav, const struct sadb_msghdr *mhp) { #ifdef IPSEC_NAT_T struct sadb_x_nat_t_port *sport, *dport; struct sadb_x_nat_t_type *type; if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_TYPE) && !SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_SPORT) && !SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_DPORT)) { if (SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_TYPE) || SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_SPORT) || SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_DPORT)) { ipseclog((LOG_DEBUG, "%s: invalid message: wrong header size.\n", __func__)); return (EINVAL); } type = (struct sadb_x_nat_t_type *) mhp->ext[SADB_X_EXT_NAT_T_TYPE]; sport = (struct sadb_x_nat_t_port *) mhp->ext[SADB_X_EXT_NAT_T_SPORT]; dport = (struct sadb_x_nat_t_port *) mhp->ext[SADB_X_EXT_NAT_T_DPORT]; sav->natt_type = type->sadb_x_nat_t_type_type; KEY_PORTTOSADDR(&sav->sah->saidx.src, sport->sadb_x_nat_t_port_port); KEY_PORTTOSADDR(&sav->sah->saidx.dst, dport->sadb_x_nat_t_port_port); } else return (0); if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_OAI) && !SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_OAR)) { if (SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_OAI) || SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_OAR)) { ipseclog((LOG_DEBUG, "%s: invalid message: wrong header size.\n", __func__)); return (EINVAL); } ipseclog((LOG_DEBUG, "%s: NAT-T OAi/r present\n", __func__)); } if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_FRAG)) { if (SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_FRAG)) { ipseclog((LOG_DEBUG, "%s: invalid message: wrong header size.\n", __func__)); return (EINVAL); } ipseclog((LOG_DEBUG, "%s: NAT-T frag present\n", __func__)); #if 0 struct sadb_x_nat_t_frag *frag; frag = (struct sadb_x_nat_t_frag *) mhp->ext[SADB_X_EXT_NAT_T_FRAG]; /* * In case SADB_X_EXT_NAT_T_FRAG was not given, leave it at 0. * We should actually check for a minimum MTU here, if we * want to support it in ip_output. */ sav->natt_esp_frag_len = frag->sadb_x_nat_t_frag_fraglen; #endif } #endif return (0); } static int key_setident(struct secashead *sah, const struct sadb_msghdr *mhp) { const struct sadb_ident *idsrc, *iddst; int idsrclen, iddstlen; IPSEC_ASSERT(sah != NULL, ("null secashead")); IPSEC_ASSERT(mhp != NULL, ("null msghdr")); IPSEC_ASSERT(mhp->msg != NULL, ("null msg")); /* don't make buffer if not there */ if (SADB_CHECKHDR(mhp, SADB_EXT_IDENTITY_SRC) && SADB_CHECKHDR(mhp, SADB_EXT_IDENTITY_DST)) { sah->idents = NULL; sah->identd = NULL; return (0); } if (SADB_CHECKHDR(mhp, SADB_EXT_IDENTITY_SRC) || SADB_CHECKHDR(mhp, SADB_EXT_IDENTITY_DST)) { ipseclog((LOG_DEBUG, "%s: invalid identity.\n", __func__)); return (EINVAL); } idsrc = (const struct sadb_ident *)mhp->ext[SADB_EXT_IDENTITY_SRC]; iddst = (const struct sadb_ident *)mhp->ext[SADB_EXT_IDENTITY_DST]; idsrclen = mhp->extlen[SADB_EXT_IDENTITY_SRC]; iddstlen = mhp->extlen[SADB_EXT_IDENTITY_DST]; /* validity check */ if (idsrc->sadb_ident_type != iddst->sadb_ident_type) { ipseclog((LOG_DEBUG, "%s: ident type mismatch.\n", __func__)); return EINVAL; } switch (idsrc->sadb_ident_type) { case SADB_IDENTTYPE_PREFIX: case SADB_IDENTTYPE_FQDN: case SADB_IDENTTYPE_USERFQDN: default: /* XXX do nothing */ sah->idents = NULL; sah->identd = NULL; return 0; } /* make structure */ sah->idents = malloc(sizeof(struct secident), M_IPSEC_MISC, M_NOWAIT); if (sah->idents == NULL) { ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__)); return ENOBUFS; } sah->identd = malloc(sizeof(struct secident), M_IPSEC_MISC, M_NOWAIT); if (sah->identd == NULL) { free(sah->idents, M_IPSEC_MISC); sah->idents = NULL; ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__)); return ENOBUFS; } sah->idents->type = idsrc->sadb_ident_type; sah->idents->id = idsrc->sadb_ident_id; sah->identd->type = iddst->sadb_ident_type; sah->identd->id = iddst->sadb_ident_id; return 0; } /* * m will not be freed on return. * it is caller's responsibility to free the result. */ static struct mbuf * key_getmsgbuf_x1(struct mbuf *m, const struct sadb_msghdr *mhp) { struct mbuf *n; IPSEC_ASSERT(m != NULL, ("null mbuf")); IPSEC_ASSERT(mhp != NULL, ("null msghdr")); IPSEC_ASSERT(mhp->msg != NULL, ("null msg")); /* create new sadb_msg to reply. */ n = key_gather_mbuf(m, mhp, 1, 9, SADB_EXT_RESERVED, SADB_EXT_SA, SADB_X_EXT_SA2, SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST, SADB_EXT_LIFETIME_HARD, SADB_EXT_LIFETIME_SOFT, SADB_EXT_IDENTITY_SRC, SADB_EXT_IDENTITY_DST); if (!n) return NULL; if (n->m_len < sizeof(struct sadb_msg)) { n = m_pullup(n, sizeof(struct sadb_msg)); if (n == NULL) return NULL; } mtod(n, struct sadb_msg *)->sadb_msg_errno = 0; mtod(n, struct sadb_msg *)->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len); return n; } /* * SADB_DELETE processing * receive * * from the ikmpd, and set SADB_SASTATE_DEAD, * and send, * * to the ikmpd. * * m will always be freed. */ static int key_delete(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp) { struct secasindex saidx; struct sadb_address *src0, *dst0; struct secasvar *sav; struct sadb_sa *sa0; uint8_t proto; IPSEC_ASSERT(so != NULL, ("null socket")); IPSEC_ASSERT(m != NULL, ("null mbuf")); IPSEC_ASSERT(mhp != NULL, ("null msghdr")); IPSEC_ASSERT(mhp->msg != NULL, ("null msg")); /* map satype to proto */ if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) { ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n", __func__)); return key_senderror(so, m, EINVAL); } if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) || SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) || SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) || SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)) { ipseclog((LOG_DEBUG, "%s: invalid message is passed.\n", __func__)); return key_senderror(so, m, EINVAL); } src0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_SRC]); dst0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_DST]); if (key_checksockaddrs((struct sockaddr *)(src0 + 1), (struct sockaddr *)(dst0 + 1)) != 0) { ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__)); return (key_senderror(so, m, EINVAL)); } KEY_SETSECASIDX(proto, IPSEC_MODE_ANY, 0, src0 + 1, dst0 + 1, &saidx); /* * Make sure the port numbers are zero. * In case of NAT-T we will update them later if needed. */ KEY_PORTTOSADDR(&saidx.src, 0); KEY_PORTTOSADDR(&saidx.dst, 0); if (SADB_CHECKHDR(mhp, SADB_EXT_SA)) { /* * Caller wants us to delete all non-LARVAL SAs * that match the src/dst. This is used during * IKE INITIAL-CONTACT. * XXXAE: this looks like some extension to RFC2367. */ ipseclog((LOG_DEBUG, "%s: doing delete all.\n", __func__)); return (key_delete_all(so, m, mhp, &saidx)); } if (SADB_CHECKLEN(mhp, SADB_EXT_SA)) { ipseclog((LOG_DEBUG, "%s: invalid message: wrong header size.\n", __func__)); return (key_senderror(so, m, EINVAL)); } sa0 = (struct sadb_sa *)mhp->ext[SADB_EXT_SA]; sav = key_getsavbyspi(sa0->sadb_sa_spi); if (sav == NULL) { ipseclog((LOG_DEBUG, "%s: no SA found for SPI %u.\n", __func__, ntohl(sa0->sadb_sa_spi))); return (key_senderror(so, m, ESRCH)); } if (key_cmpsaidx(&sav->sah->saidx, &saidx, CMP_HEAD) == 0) { ipseclog((LOG_DEBUG, "%s: saidx mismatched for SPI %u.\n", __func__, ntohl(sav->spi))); key_freesav(&sav); return (key_senderror(so, m, ESRCH)); } KEYDBG(KEY_STAMP, printf("%s: SA(%p)\n", __func__, sav)); KEYDBG(KEY_DATA, kdebug_secasv(sav)); key_unlinksav(sav); key_freesav(&sav); { struct mbuf *n; struct sadb_msg *newmsg; /* create new sadb_msg to reply. */ /* XXX-BZ NAT-T extensions? */ n = key_gather_mbuf(m, mhp, 1, 4, SADB_EXT_RESERVED, SADB_EXT_SA, SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST); if (!n) return key_senderror(so, m, ENOBUFS); if (n->m_len < sizeof(struct sadb_msg)) { n = m_pullup(n, sizeof(struct sadb_msg)); if (n == NULL) return key_senderror(so, m, ENOBUFS); } newmsg = mtod(n, struct sadb_msg *); newmsg->sadb_msg_errno = 0; newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len); m_freem(m); return key_sendup_mbuf(so, n, KEY_SENDUP_ALL); } } /* * delete all SAs for src/dst. Called from key_delete(). */ static int key_delete_all(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp, struct secasindex *saidx) { struct secasvar_queue drainq; struct secashead *sah; struct secasvar *sav, *nextsav; TAILQ_INIT(&drainq); SAHTREE_WLOCK(); LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) { if (key_cmpsaidx(&sah->saidx, saidx, CMP_HEAD) == 0) continue; /* Move all ALIVE SAs into drainq */ TAILQ_CONCAT(&drainq, &sah->savtree_alive, chain); } /* Unlink all queued SAs from SPI hash */ TAILQ_FOREACH(sav, &drainq, chain) { sav->state = SADB_SASTATE_DEAD; LIST_REMOVE(sav, spihash); } SAHTREE_WUNLOCK(); /* Now we can release reference for all SAs in drainq */ sav = TAILQ_FIRST(&drainq); while (sav != NULL) { KEYDBG(KEY_STAMP, printf("%s: SA(%p)\n", __func__, sav)); KEYDBG(KEY_DATA, kdebug_secasv(sav)); nextsav = TAILQ_NEXT(sav, chain); key_freesah(&sav->sah); /* release reference from SAV */ key_freesav(&sav); /* release last reference */ sav = nextsav; } { struct mbuf *n; struct sadb_msg *newmsg; /* create new sadb_msg to reply. */ /* XXX-BZ NAT-T extensions? */ n = key_gather_mbuf(m, mhp, 1, 3, SADB_EXT_RESERVED, SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST); if (!n) return key_senderror(so, m, ENOBUFS); if (n->m_len < sizeof(struct sadb_msg)) { n = m_pullup(n, sizeof(struct sadb_msg)); if (n == NULL) return key_senderror(so, m, ENOBUFS); } newmsg = mtod(n, struct sadb_msg *); newmsg->sadb_msg_errno = 0; newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len); m_freem(m); return key_sendup_mbuf(so, n, KEY_SENDUP_ALL); } } /* * SADB_GET processing * receive * * from the ikmpd, and get a SP and a SA to respond, * and send, * * to the ikmpd. * * m will always be freed. */ static int key_get(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp) { struct secasindex saidx; struct sadb_address *src0, *dst0; struct sadb_sa *sa0; struct secasvar *sav; uint8_t proto; IPSEC_ASSERT(so != NULL, ("null socket")); IPSEC_ASSERT(m != NULL, ("null mbuf")); IPSEC_ASSERT(mhp != NULL, ("null msghdr")); IPSEC_ASSERT(mhp->msg != NULL, ("null msg")); /* map satype to proto */ if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) { ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n", __func__)); return key_senderror(so, m, EINVAL); } if (SADB_CHECKHDR(mhp, SADB_EXT_SA) || SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) || SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST)) { ipseclog((LOG_DEBUG, "%s: invalid message: missing required header.\n", __func__)); return key_senderror(so, m, EINVAL); } if (SADB_CHECKLEN(mhp, SADB_EXT_SA) || SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) || SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)) { ipseclog((LOG_DEBUG, "%s: invalid message: wrong header size.\n", __func__)); return key_senderror(so, m, EINVAL); } sa0 = (struct sadb_sa *)mhp->ext[SADB_EXT_SA]; src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC]; dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST]; if (key_checksockaddrs((struct sockaddr *)(src0 + 1), (struct sockaddr *)(dst0 + 1)) != 0) { ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__)); return key_senderror(so, m, EINVAL); } KEY_SETSECASIDX(proto, IPSEC_MODE_ANY, 0, src0 + 1, dst0 + 1, &saidx); /* * Make sure the port numbers are zero. * In case of NAT-T we will update them later if needed. */ KEY_PORTTOSADDR(&saidx.src, 0); KEY_PORTTOSADDR(&saidx.dst, 0); sav = key_getsavbyspi(sa0->sadb_sa_spi); if (sav == NULL) { ipseclog((LOG_DEBUG, "%s: no SA found.\n", __func__)); return key_senderror(so, m, ESRCH); } if (key_cmpsaidx(&sav->sah->saidx, &saidx, CMP_HEAD) == 0) { ipseclog((LOG_DEBUG, "%s: saidx mismatched for SPI %u.\n", __func__, ntohl(sav->spi))); key_freesav(&sav); return (key_senderror(so, m, ESRCH)); } { struct mbuf *n; uint8_t satype; /* map proto to satype */ if ((satype = key_proto2satype(sav->sah->saidx.proto)) == 0) { ipseclog((LOG_DEBUG, "%s: there was invalid proto in SAD.\n", __func__)); key_freesav(&sav); return key_senderror(so, m, EINVAL); } /* create new sadb_msg to reply. */ n = key_setdumpsa(sav, SADB_GET, satype, mhp->msg->sadb_msg_seq, mhp->msg->sadb_msg_pid); key_freesav(&sav); if (!n) return key_senderror(so, m, ENOBUFS); m_freem(m); return key_sendup_mbuf(so, n, KEY_SENDUP_ONE); } } /* XXX make it sysctl-configurable? */ static void key_getcomb_setlifetime(struct sadb_comb *comb) { comb->sadb_comb_soft_allocations = 1; comb->sadb_comb_hard_allocations = 1; comb->sadb_comb_soft_bytes = 0; comb->sadb_comb_hard_bytes = 0; comb->sadb_comb_hard_addtime = 86400; /* 1 day */ comb->sadb_comb_soft_addtime = comb->sadb_comb_soft_addtime * 80 / 100; comb->sadb_comb_soft_usetime = 28800; /* 8 hours */ comb->sadb_comb_hard_usetime = comb->sadb_comb_hard_usetime * 80 / 100; } /* * XXX reorder combinations by preference * XXX no idea if the user wants ESP authentication or not */ static struct mbuf * key_getcomb_esp() { struct sadb_comb *comb; struct enc_xform *algo; struct mbuf *result = NULL, *m, *n; int encmin; int i, off, o; int totlen; const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb)); m = NULL; for (i = 1; i <= SADB_EALG_MAX; i++) { algo = esp_algorithm_lookup(i); if (algo == NULL) continue; /* discard algorithms with key size smaller than system min */ if (_BITS(algo->maxkey) < V_ipsec_esp_keymin) continue; if (_BITS(algo->minkey) < V_ipsec_esp_keymin) encmin = V_ipsec_esp_keymin; else encmin = _BITS(algo->minkey); if (V_ipsec_esp_auth) m = key_getcomb_ah(); else { IPSEC_ASSERT(l <= MLEN, ("l=%u > MLEN=%lu", l, (u_long) MLEN)); MGET(m, M_NOWAIT, MT_DATA); if (m) { M_ALIGN(m, l); m->m_len = l; m->m_next = NULL; bzero(mtod(m, caddr_t), m->m_len); } } if (!m) goto fail; totlen = 0; for (n = m; n; n = n->m_next) totlen += n->m_len; IPSEC_ASSERT((totlen % l) == 0, ("totlen=%u, l=%u", totlen, l)); for (off = 0; off < totlen; off += l) { n = m_pulldown(m, off, l, &o); if (!n) { /* m is already freed */ goto fail; } comb = (struct sadb_comb *)(mtod(n, caddr_t) + o); bzero(comb, sizeof(*comb)); key_getcomb_setlifetime(comb); comb->sadb_comb_encrypt = i; comb->sadb_comb_encrypt_minbits = encmin; comb->sadb_comb_encrypt_maxbits = _BITS(algo->maxkey); } if (!result) result = m; else m_cat(result, m); } return result; fail: if (result) m_freem(result); return NULL; } static void key_getsizes_ah(const struct auth_hash *ah, int alg, u_int16_t* min, u_int16_t* max) { *min = *max = ah->keysize; if (ah->keysize == 0) { /* * Transform takes arbitrary key size but algorithm * key size is restricted. Enforce this here. */ switch (alg) { case SADB_X_AALG_MD5: *min = *max = 16; break; case SADB_X_AALG_SHA: *min = *max = 20; break; case SADB_X_AALG_NULL: *min = 1; *max = 256; break; case SADB_X_AALG_SHA2_256: *min = *max = 32; break; case SADB_X_AALG_SHA2_384: *min = *max = 48; break; case SADB_X_AALG_SHA2_512: *min = *max = 64; break; default: DPRINTF(("%s: unknown AH algorithm %u\n", __func__, alg)); break; } } } /* * XXX reorder combinations by preference */ static struct mbuf * key_getcomb_ah() { struct sadb_comb *comb; struct auth_hash *algo; struct mbuf *m; u_int16_t minkeysize, maxkeysize; int i; const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb)); m = NULL; for (i = 1; i <= SADB_AALG_MAX; i++) { #if 1 /* we prefer HMAC algorithms, not old algorithms */ if (i != SADB_AALG_SHA1HMAC && i != SADB_AALG_MD5HMAC && i != SADB_X_AALG_SHA2_256 && i != SADB_X_AALG_SHA2_384 && i != SADB_X_AALG_SHA2_512) continue; #endif algo = ah_algorithm_lookup(i); if (!algo) continue; key_getsizes_ah(algo, i, &minkeysize, &maxkeysize); /* discard algorithms with key size smaller than system min */ if (_BITS(minkeysize) < V_ipsec_ah_keymin) continue; if (!m) { IPSEC_ASSERT(l <= MLEN, ("l=%u > MLEN=%lu", l, (u_long) MLEN)); MGET(m, M_NOWAIT, MT_DATA); if (m) { M_ALIGN(m, l); m->m_len = l; m->m_next = NULL; } } else M_PREPEND(m, l, M_NOWAIT); if (!m) return NULL; comb = mtod(m, struct sadb_comb *); bzero(comb, sizeof(*comb)); key_getcomb_setlifetime(comb); comb->sadb_comb_auth = i; comb->sadb_comb_auth_minbits = _BITS(minkeysize); comb->sadb_comb_auth_maxbits = _BITS(maxkeysize); } return m; } /* * not really an official behavior. discussed in pf_key@inner.net in Sep2000. * XXX reorder combinations by preference */ static struct mbuf * key_getcomb_ipcomp() { struct sadb_comb *comb; struct comp_algo *algo; struct mbuf *m; int i; const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb)); m = NULL; for (i = 1; i <= SADB_X_CALG_MAX; i++) { algo = ipcomp_algorithm_lookup(i); if (!algo) continue; if (!m) { IPSEC_ASSERT(l <= MLEN, ("l=%u > MLEN=%lu", l, (u_long) MLEN)); MGET(m, M_NOWAIT, MT_DATA); if (m) { M_ALIGN(m, l); m->m_len = l; m->m_next = NULL; } } else M_PREPEND(m, l, M_NOWAIT); if (!m) return NULL; comb = mtod(m, struct sadb_comb *); bzero(comb, sizeof(*comb)); key_getcomb_setlifetime(comb); comb->sadb_comb_encrypt = i; /* what should we set into sadb_comb_*_{min,max}bits? */ } return m; } /* * XXX no way to pass mode (transport/tunnel) to userland * XXX replay checking? * XXX sysctl interface to ipsec_{ah,esp}_keymin */ static struct mbuf * key_getprop(const struct secasindex *saidx) { struct sadb_prop *prop; struct mbuf *m, *n; const int l = PFKEY_ALIGN8(sizeof(struct sadb_prop)); int totlen; switch (saidx->proto) { case IPPROTO_ESP: m = key_getcomb_esp(); break; case IPPROTO_AH: m = key_getcomb_ah(); break; case IPPROTO_IPCOMP: m = key_getcomb_ipcomp(); break; default: return NULL; } if (!m) return NULL; M_PREPEND(m, l, M_NOWAIT); if (!m) return NULL; totlen = 0; for (n = m; n; n = n->m_next) totlen += n->m_len; prop = mtod(m, struct sadb_prop *); bzero(prop, sizeof(*prop)); prop->sadb_prop_len = PFKEY_UNIT64(totlen); prop->sadb_prop_exttype = SADB_EXT_PROPOSAL; prop->sadb_prop_replay = 32; /* XXX */ return m; } /* * SADB_ACQUIRE processing called by key_checkrequest() and key_acquire2(). * send * * to KMD, and expect to receive * with SADB_ACQUIRE if error occurred, * or * with SADB_GETSPI * from KMD by PF_KEY. * * XXX x_policy is outside of RFC2367 (KAME extension). * XXX sensitivity is not supported. * XXX for ipcomp, RFC2367 does not define how to fill in proposal. * see comment for key_getcomb_ipcomp(). * * OUT: * 0 : succeed * others: error number */ static int key_acquire(const struct secasindex *saidx, struct secpolicy *sp) { union sockaddr_union addr; struct mbuf *result, *m; uint32_t seq; int error; uint16_t ul_proto; uint8_t mask, satype; IPSEC_ASSERT(saidx != NULL, ("null saidx")); satype = key_proto2satype(saidx->proto); IPSEC_ASSERT(satype != 0, ("null satype, protocol %u", saidx->proto)); error = -1; result = NULL; ul_proto = IPSEC_ULPROTO_ANY; /* Get seq number to check whether sending message or not. */ seq = key_getacq(saidx, &error); if (seq == 0) return (error); m = key_setsadbmsg(SADB_ACQUIRE, 0, satype, seq, 0, 0); if (!m) { error = ENOBUFS; goto fail; } result = m; /* * No SADB_X_EXT_NAT_T_* here: we do not know * anything related to NAT-T at this time. */ /* * set sadb_address for saidx's. * * Note that if sp is supplied, then we're being called from * key_allocsa_policy() and should supply port and protocol * information. * XXXAE: why only TCP and UDP? ICMP and SCTP looks applicable too. * XXXAE: probably we can handle this in the ipsec[46]_allocsa(). * XXXAE: it looks like we should save this info in the ACQ entry. */ if (sp != NULL && (sp->spidx.ul_proto == IPPROTO_TCP || sp->spidx.ul_proto == IPPROTO_UDP)) ul_proto = sp->spidx.ul_proto; addr = saidx->src; mask = FULLMASK; if (ul_proto != IPSEC_ULPROTO_ANY) { switch (sp->spidx.src.sa.sa_family) { case AF_INET: if (sp->spidx.src.sin.sin_port != IPSEC_PORT_ANY) { addr.sin.sin_port = sp->spidx.src.sin.sin_port; mask = sp->spidx.prefs; } break; case AF_INET6: if (sp->spidx.src.sin6.sin6_port != IPSEC_PORT_ANY) { addr.sin6.sin6_port = sp->spidx.src.sin6.sin6_port; mask = sp->spidx.prefs; } break; default: break; } } m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC, &addr.sa, mask, ul_proto); if (!m) { error = ENOBUFS; goto fail; } m_cat(result, m); addr = saidx->dst; mask = FULLMASK; if (ul_proto != IPSEC_ULPROTO_ANY) { switch (sp->spidx.dst.sa.sa_family) { case AF_INET: if (sp->spidx.dst.sin.sin_port != IPSEC_PORT_ANY) { addr.sin.sin_port = sp->spidx.dst.sin.sin_port; mask = sp->spidx.prefd; } break; case AF_INET6: if (sp->spidx.dst.sin6.sin6_port != IPSEC_PORT_ANY) { addr.sin6.sin6_port = sp->spidx.dst.sin6.sin6_port; mask = sp->spidx.prefd; } break; default: break; } } m = key_setsadbaddr(SADB_EXT_ADDRESS_DST, &addr.sa, mask, ul_proto); if (!m) { error = ENOBUFS; goto fail; } m_cat(result, m); /* XXX proxy address (optional) */ /* set sadb_x_policy */ if (sp != NULL) { m = key_setsadbxpolicy(sp->policy, sp->spidx.dir, sp->id, sp->priority); if (!m) { error = ENOBUFS; goto fail; } m_cat(result, m); } /* XXX identity (optional) */ #if 0 if (idexttype && fqdn) { /* create identity extension (FQDN) */ struct sadb_ident *id; int fqdnlen; fqdnlen = strlen(fqdn) + 1; /* +1 for terminating-NUL */ id = (struct sadb_ident *)p; bzero(id, sizeof(*id) + PFKEY_ALIGN8(fqdnlen)); id->sadb_ident_len = PFKEY_UNIT64(sizeof(*id) + PFKEY_ALIGN8(fqdnlen)); id->sadb_ident_exttype = idexttype; id->sadb_ident_type = SADB_IDENTTYPE_FQDN; bcopy(fqdn, id + 1, fqdnlen); p += sizeof(struct sadb_ident) + PFKEY_ALIGN8(fqdnlen); } if (idexttype) { /* create identity extension (USERFQDN) */ struct sadb_ident *id; int userfqdnlen; if (userfqdn) { /* +1 for terminating-NUL */ userfqdnlen = strlen(userfqdn) + 1; } else userfqdnlen = 0; id = (struct sadb_ident *)p; bzero(id, sizeof(*id) + PFKEY_ALIGN8(userfqdnlen)); id->sadb_ident_len = PFKEY_UNIT64(sizeof(*id) + PFKEY_ALIGN8(userfqdnlen)); id->sadb_ident_exttype = idexttype; id->sadb_ident_type = SADB_IDENTTYPE_USERFQDN; /* XXX is it correct? */ if (curproc && curproc->p_cred) id->sadb_ident_id = curproc->p_cred->p_ruid; if (userfqdn && userfqdnlen) bcopy(userfqdn, id + 1, userfqdnlen); p += sizeof(struct sadb_ident) + PFKEY_ALIGN8(userfqdnlen); } #endif /* XXX sensitivity (optional) */ /* create proposal/combination extension */ m = key_getprop(saidx); #if 0 /* * spec conformant: always attach proposal/combination extension, * the problem is that we have no way to attach it for ipcomp, * due to the way sadb_comb is declared in RFC2367. */ if (!m) { error = ENOBUFS; goto fail; } m_cat(result, m); #else /* * outside of spec; make proposal/combination extension optional. */ if (m) m_cat(result, m); #endif if ((result->m_flags & M_PKTHDR) == 0) { error = EINVAL; goto fail; } if (result->m_len < sizeof(struct sadb_msg)) { result = m_pullup(result, sizeof(struct sadb_msg)); if (result == NULL) { error = ENOBUFS; goto fail; } } result->m_pkthdr.len = 0; for (m = result; m; m = m->m_next) result->m_pkthdr.len += m->m_len; mtod(result, struct sadb_msg *)->sadb_msg_len = PFKEY_UNIT64(result->m_pkthdr.len); KEYDBG(KEY_STAMP, printf("%s: SP(%p)\n", __func__, sp)); KEYDBG(KEY_DATA, kdebug_secasindex(saidx, NULL)); return key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED); fail: if (result) m_freem(result); return error; } -static struct secacq * -key_newacq(const struct secasindex *saidx) -{ - struct secacq *newacq; - - /* get new entry */ - newacq = malloc(sizeof(struct secacq), M_IPSEC_SAQ, M_NOWAIT|M_ZERO); - if (newacq == NULL) { - ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__)); - return NULL; - } - - /* copy secindex */ - bcopy(saidx, &newacq->saidx, sizeof(newacq->saidx)); - newacq->seq = (V_acq_seq == ~0 ? 1 : ++V_acq_seq); - newacq->created = time_second; - newacq->count = 0; - - /* add to acqtree */ - ACQ_LOCK(); - LIST_INSERT_HEAD(&V_acqtree, newacq, chain); - ACQ_UNLOCK(); - - return newacq; -} - static uint32_t key_newacq(const struct secasindex *saidx, int *perror) { struct secacq *acq; uint32_t seq; acq = malloc(sizeof(*acq), M_IPSEC_SAQ, M_NOWAIT | M_ZERO); if (acq == NULL) { ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__)); *perror = ENOBUFS; return (0); } /* copy secindex */ bcopy(saidx, &acq->saidx, sizeof(acq->saidx)); acq->created = time_second; acq->count = 0; /* add to acqtree */ ACQ_LOCK(); seq = acq->seq = (V_acq_seq == ~0 ? 1 : ++V_acq_seq); LIST_INSERT_HEAD(&V_acqtree, acq, chain); LIST_INSERT_HEAD(ACQADDRHASH_HASH(saidx), acq, addrhash); LIST_INSERT_HEAD(ACQSEQHASH_HASH(seq), acq, seqhash); ACQ_UNLOCK(); *perror = 0; return (seq); } static uint32_t key_getacq(const struct secasindex *saidx, int *perror) { struct secacq *acq; uint32_t seq; ACQ_LOCK(); LIST_FOREACH(acq, ACQADDRHASH_HASH(saidx), addrhash) { if (key_cmpsaidx(&acq->saidx, saidx, CMP_EXACTLY)) { if (acq->count > V_key_blockacq_count) { /* * Reset counter and send message. * Also reset created time to keep ACQ for * this saidx. */ acq->created = time_second; acq->count = 0; seq = acq->seq; } else { /* * Increment counter and do nothing. * We send SADB_ACQUIRE message only * for each V_key_blockacq_count packet. */ acq->count++; seq = 0; } break; } } ACQ_UNLOCK(); if (acq != NULL) { *perror = 0; return (seq); } /* allocate new entry */ return (key_newacq(saidx, perror)); } static int key_acqreset(uint32_t seq) { struct secacq *acq; ACQ_LOCK(); LIST_FOREACH(acq, ACQSEQHASH_HASH(seq), seqhash) { if (acq->seq == seq) { acq->count = 0; acq->created = time_second; break; } } ACQ_UNLOCK(); if (acq == NULL) return (ESRCH); return (0); } /* * Mark ACQ entry as stale to remove it in key_flush_acq(). * Called after successful SADB_GETSPI message. */ static int key_acqdone(const struct secasindex *saidx, uint32_t seq) { struct secacq *acq; ACQ_LOCK(); LIST_FOREACH(acq, ACQSEQHASH_HASH(seq), seqhash) { if (acq->seq == seq) break; } if (acq != NULL) { if (key_cmpsaidx(&acq->saidx, saidx, CMP_EXACTLY) == 0) { ipseclog((LOG_DEBUG, "%s: Mismatched saidx for ACQ %u", __func__, seq)); acq = NULL; } else { acq->created = 0; } } else { ipseclog((LOG_DEBUG, "%s: ACQ %u is not found.", __func__, seq)); } ACQ_UNLOCK(); if (acq == NULL) return (ESRCH); return (0); } static struct secspacq * key_newspacq(struct secpolicyindex *spidx) { struct secspacq *acq; /* get new entry */ acq = malloc(sizeof(struct secspacq), M_IPSEC_SAQ, M_NOWAIT|M_ZERO); if (acq == NULL) { ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__)); return NULL; } /* copy secindex */ bcopy(spidx, &acq->spidx, sizeof(acq->spidx)); acq->created = time_second; acq->count = 0; /* add to spacqtree */ SPACQ_LOCK(); LIST_INSERT_HEAD(&V_spacqtree, acq, chain); SPACQ_UNLOCK(); return acq; } static struct secspacq * key_getspacq(struct secpolicyindex *spidx) { struct secspacq *acq; SPACQ_LOCK(); LIST_FOREACH(acq, &V_spacqtree, chain) { if (key_cmpspidx_exactly(spidx, &acq->spidx)) { /* NB: return holding spacq_lock */ return acq; } } SPACQ_UNLOCK(); return NULL; } /* * SADB_ACQUIRE processing, * in first situation, is receiving * * from the ikmpd, and clear sequence of its secasvar entry. * * In second situation, is receiving * * from a user land process, and return * * to the socket. * * m will always be freed. */ static int key_acquire2(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp) { SAHTREE_RLOCK_TRACKER; struct sadb_address *src0, *dst0; struct secasindex saidx; struct secashead *sah; uint32_t reqid; int error; uint8_t mode, proto; IPSEC_ASSERT(so != NULL, ("null socket")); IPSEC_ASSERT(m != NULL, ("null mbuf")); IPSEC_ASSERT(mhp != NULL, ("null msghdr")); IPSEC_ASSERT(mhp->msg != NULL, ("null msg")); /* * Error message from KMd. * We assume that if error was occurred in IKEd, the length of PFKEY * message is equal to the size of sadb_msg structure. * We do not raise error even if error occurred in this function. */ if (mhp->msg->sadb_msg_len == PFKEY_UNIT64(sizeof(struct sadb_msg))) { /* check sequence number */ if (mhp->msg->sadb_msg_seq == 0 || mhp->msg->sadb_msg_errno == 0) { ipseclog((LOG_DEBUG, "%s: must specify sequence " "number and errno.\n", __func__)); } else { /* * IKEd reported that error occurred. * XXXAE: what it expects from the kernel? * Probably we should send SADB_ACQUIRE again? * If so, reset ACQ's state. * XXXAE: it looks useless. */ key_acqreset(mhp->msg->sadb_msg_seq); } m_freem(m); return (0); } /* * This message is from user land. */ /* map satype to proto */ if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) { ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n", __func__)); return key_senderror(so, m, EINVAL); } if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) || SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) || SADB_CHECKHDR(mhp, SADB_EXT_PROPOSAL)) { ipseclog((LOG_DEBUG, "%s: invalid message: missing required header.\n", __func__)); return key_senderror(so, m, EINVAL); } if (SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) || SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST) || SADB_CHECKLEN(mhp, SADB_EXT_PROPOSAL)) { ipseclog((LOG_DEBUG, "%s: invalid message: wrong header size.\n", __func__)); return key_senderror(so, m, EINVAL); } if (SADB_CHECKHDR(mhp, SADB_X_EXT_SA2)) { mode = IPSEC_MODE_ANY; reqid = 0; } else { if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA2)) { ipseclog((LOG_DEBUG, "%s: invalid message: wrong header size.\n", __func__)); return key_senderror(so, m, EINVAL); } mode = ((struct sadb_x_sa2 *) mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_mode; reqid = ((struct sadb_x_sa2 *) mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_reqid; } src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC]; dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST]; error = key_checksockaddrs((struct sockaddr *)(src0 + 1), (struct sockaddr *)(dst0 + 1)); if (error != 0) { ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__)); return key_senderror(so, m, EINVAL); } KEY_SETSECASIDX(proto, mode, reqid, src0 + 1, dst0 + 1, &saidx); /* * Make sure the port numbers are zero. * In case of NAT-T we will update them later if needed. */ KEY_PORTTOSADDR(&saidx.src, 0); KEY_PORTTOSADDR(&saidx.dst, 0); /* get a SA index */ SAHTREE_RLOCK(); LIST_FOREACH(sah, SAHADDRHASH_HASH(&saidx), addrhash) { if (key_cmpsaidx(&sah->saidx, &saidx, CMP_MODE_REQID)) break; } SAHTREE_RUNLOCK(); if (sah != NULL) { ipseclog((LOG_DEBUG, "%s: a SA exists already.\n", __func__)); return key_senderror(so, m, EEXIST); } error = key_acquire(&saidx, NULL); if (error != 0) { ipseclog((LOG_DEBUG, "%s: error %d returned from key_acquire()\n", __func__, error)); return key_senderror(so, m, error); } m_freem(m); return (0); } /* * SADB_REGISTER processing. * If SATYPE_UNSPEC has been passed as satype, only return sabd_supported. * receive * * from the ikmpd, and register a socket to send PF_KEY messages, * and send * * to KMD by PF_KEY. * If socket is detached, must free from regnode. * * m will always be freed. */ static int key_register(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp) { struct secreg *reg, *newreg = NULL; IPSEC_ASSERT(so != NULL, ("null socket")); IPSEC_ASSERT(m != NULL, ("null mbuf")); IPSEC_ASSERT(mhp != NULL, ("null msghdr")); IPSEC_ASSERT(mhp->msg != NULL, ("null msg")); /* check for invalid register message */ if (mhp->msg->sadb_msg_satype >= sizeof(V_regtree)/sizeof(V_regtree[0])) return key_senderror(so, m, EINVAL); /* When SATYPE_UNSPEC is specified, only return sabd_supported. */ if (mhp->msg->sadb_msg_satype == SADB_SATYPE_UNSPEC) goto setmsg; /* check whether existing or not */ REGTREE_LOCK(); LIST_FOREACH(reg, &V_regtree[mhp->msg->sadb_msg_satype], chain) { if (reg->so == so) { REGTREE_UNLOCK(); ipseclog((LOG_DEBUG, "%s: socket exists already.\n", __func__)); return key_senderror(so, m, EEXIST); } } /* create regnode */ newreg = malloc(sizeof(struct secreg), M_IPSEC_SAR, M_NOWAIT|M_ZERO); if (newreg == NULL) { REGTREE_UNLOCK(); ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__)); return key_senderror(so, m, ENOBUFS); } newreg->so = so; ((struct keycb *)sotorawcb(so))->kp_registered++; /* add regnode to regtree. */ LIST_INSERT_HEAD(&V_regtree[mhp->msg->sadb_msg_satype], newreg, chain); REGTREE_UNLOCK(); setmsg: { struct mbuf *n; struct sadb_msg *newmsg; struct sadb_supported *sup; u_int len, alen, elen; int off; int i; struct sadb_alg *alg; /* create new sadb_msg to reply. */ alen = 0; for (i = 1; i <= SADB_AALG_MAX; i++) { if (ah_algorithm_lookup(i)) alen += sizeof(struct sadb_alg); } if (alen) alen += sizeof(struct sadb_supported); elen = 0; for (i = 1; i <= SADB_EALG_MAX; i++) { if (esp_algorithm_lookup(i)) elen += sizeof(struct sadb_alg); } if (elen) elen += sizeof(struct sadb_supported); len = sizeof(struct sadb_msg) + alen + elen; if (len > MCLBYTES) return key_senderror(so, m, ENOBUFS); MGETHDR(n, M_NOWAIT, MT_DATA); if (len > MHLEN) { if (!(MCLGET(n, M_NOWAIT))) { m_freem(n); n = NULL; } } if (!n) return key_senderror(so, m, ENOBUFS); n->m_pkthdr.len = n->m_len = len; n->m_next = NULL; off = 0; m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, caddr_t) + off); newmsg = mtod(n, struct sadb_msg *); newmsg->sadb_msg_errno = 0; newmsg->sadb_msg_len = PFKEY_UNIT64(len); off += PFKEY_ALIGN8(sizeof(struct sadb_msg)); /* for authentication algorithm */ if (alen) { sup = (struct sadb_supported *)(mtod(n, caddr_t) + off); sup->sadb_supported_len = PFKEY_UNIT64(alen); sup->sadb_supported_exttype = SADB_EXT_SUPPORTED_AUTH; off += PFKEY_ALIGN8(sizeof(*sup)); for (i = 1; i <= SADB_AALG_MAX; i++) { struct auth_hash *aalgo; u_int16_t minkeysize, maxkeysize; aalgo = ah_algorithm_lookup(i); if (!aalgo) continue; alg = (struct sadb_alg *)(mtod(n, caddr_t) + off); alg->sadb_alg_id = i; alg->sadb_alg_ivlen = 0; key_getsizes_ah(aalgo, i, &minkeysize, &maxkeysize); alg->sadb_alg_minbits = _BITS(minkeysize); alg->sadb_alg_maxbits = _BITS(maxkeysize); off += PFKEY_ALIGN8(sizeof(*alg)); } } /* for encryption algorithm */ if (elen) { sup = (struct sadb_supported *)(mtod(n, caddr_t) + off); sup->sadb_supported_len = PFKEY_UNIT64(elen); sup->sadb_supported_exttype = SADB_EXT_SUPPORTED_ENCRYPT; off += PFKEY_ALIGN8(sizeof(*sup)); for (i = 1; i <= SADB_EALG_MAX; i++) { struct enc_xform *ealgo; ealgo = esp_algorithm_lookup(i); if (!ealgo) continue; alg = (struct sadb_alg *)(mtod(n, caddr_t) + off); alg->sadb_alg_id = i; alg->sadb_alg_ivlen = ealgo->ivsize; alg->sadb_alg_minbits = _BITS(ealgo->minkey); alg->sadb_alg_maxbits = _BITS(ealgo->maxkey); off += PFKEY_ALIGN8(sizeof(struct sadb_alg)); } } IPSEC_ASSERT(off == len, ("length assumption failed (off %u len %u)", off, len)); m_freem(m); return key_sendup_mbuf(so, n, KEY_SENDUP_REGISTERED); } } /* * free secreg entry registered. * XXX: I want to do free a socket marked done SADB_RESIGER to socket. */ void key_freereg(struct socket *so) { struct secreg *reg; int i; IPSEC_ASSERT(so != NULL, ("NULL so")); /* * check whether existing or not. * check all type of SA, because there is a potential that * one socket is registered to multiple type of SA. */ REGTREE_LOCK(); for (i = 0; i <= SADB_SATYPE_MAX; i++) { LIST_FOREACH(reg, &V_regtree[i], chain) { if (reg->so == so && __LIST_CHAINED(reg)) { LIST_REMOVE(reg, chain); free(reg, M_IPSEC_SAR); break; } } } REGTREE_UNLOCK(); } /* * SADB_EXPIRE processing * send * * to KMD by PF_KEY. * NOTE: We send only soft lifetime extension. * * OUT: 0 : succeed * others : error number */ static int key_expire(struct secasvar *sav, int hard) { struct mbuf *result = NULL, *m; struct sadb_lifetime *lt; int error, len; uint8_t satype; IPSEC_ASSERT (sav != NULL, ("null sav")); IPSEC_ASSERT (sav->sah != NULL, ("null sa header")); KEYDBG(KEY_STAMP, printf("%s: SA(%p) expired %s lifetime\n", __func__, sav, hard ? "hard": "soft")); KEYDBG(KEY_DATA, kdebug_secasv(sav)); /* set msg header */ satype = key_proto2satype(sav->sah->saidx.proto); IPSEC_ASSERT(satype != 0, ("invalid proto, satype %u", satype)); m = key_setsadbmsg(SADB_EXPIRE, 0, satype, sav->seq, 0, sav->refcnt); if (!m) { error = ENOBUFS; goto fail; } result = m; /* create SA extension */ m = key_setsadbsa(sav); if (!m) { error = ENOBUFS; goto fail; } m_cat(result, m); /* create SA extension */ m = key_setsadbxsa2(sav->sah->saidx.mode, sav->replay ? sav->replay->count : 0, sav->sah->saidx.reqid); if (!m) { error = ENOBUFS; goto fail; } m_cat(result, m); /* create lifetime extension (current and soft) */ len = PFKEY_ALIGN8(sizeof(*lt)) * 2; m = m_get2(len, M_NOWAIT, MT_DATA, 0); if (m == NULL) { error = ENOBUFS; goto fail; } m_align(m, len); m->m_len = len; bzero(mtod(m, caddr_t), len); lt = mtod(m, struct sadb_lifetime *); lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime)); lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_CURRENT; lt->sadb_lifetime_allocations = (uint32_t)counter_u64_fetch(sav->lft_c_allocations); lt->sadb_lifetime_bytes = counter_u64_fetch(sav->lft_c_bytes); lt->sadb_lifetime_addtime = sav->created; lt->sadb_lifetime_usetime = sav->firstused; lt = (struct sadb_lifetime *)(mtod(m, caddr_t) + len / 2); lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime)); if (hard) { lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_HARD; lt->sadb_lifetime_allocations = sav->lft_h->allocations; lt->sadb_lifetime_bytes = sav->lft_h->bytes; lt->sadb_lifetime_addtime = sav->lft_h->addtime; lt->sadb_lifetime_usetime = sav->lft_h->usetime; } else { lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_SOFT; lt->sadb_lifetime_allocations = sav->lft_s->allocations; lt->sadb_lifetime_bytes = sav->lft_s->bytes; lt->sadb_lifetime_addtime = sav->lft_s->addtime; lt->sadb_lifetime_usetime = sav->lft_s->usetime; } m_cat(result, m); /* set sadb_address for source */ m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC, &sav->sah->saidx.src.sa, FULLMASK, IPSEC_ULPROTO_ANY); if (!m) { error = ENOBUFS; goto fail; } m_cat(result, m); /* set sadb_address for destination */ m = key_setsadbaddr(SADB_EXT_ADDRESS_DST, &sav->sah->saidx.dst.sa, FULLMASK, IPSEC_ULPROTO_ANY); if (!m) { error = ENOBUFS; goto fail; } m_cat(result, m); /* * XXX-BZ Handle NAT-T extensions here. */ if ((result->m_flags & M_PKTHDR) == 0) { error = EINVAL; goto fail; } if (result->m_len < sizeof(struct sadb_msg)) { result = m_pullup(result, sizeof(struct sadb_msg)); if (result == NULL) { error = ENOBUFS; goto fail; } } result->m_pkthdr.len = 0; for (m = result; m; m = m->m_next) result->m_pkthdr.len += m->m_len; mtod(result, struct sadb_msg *)->sadb_msg_len = PFKEY_UNIT64(result->m_pkthdr.len); return key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED); fail: if (result) m_freem(result); return error; } static void key_freesah_flushed(struct secashead_queue *flushq) { struct secashead *sah, *nextsah; struct secasvar *sav, *nextsav; sah = TAILQ_FIRST(flushq); while (sah != NULL) { sav = TAILQ_FIRST(&sah->savtree_larval); while (sav != NULL) { nextsav = TAILQ_NEXT(sav, chain); TAILQ_REMOVE(&sah->savtree_larval, sav, chain); key_freesav(&sav); /* release last reference */ key_freesah(&sah); /* release reference from SAV */ sav = nextsav; } sav = TAILQ_FIRST(&sah->savtree_alive); while (sav != NULL) { nextsav = TAILQ_NEXT(sav, chain); TAILQ_REMOVE(&sah->savtree_alive, sav, chain); key_freesav(&sav); /* release last reference */ key_freesah(&sah); /* release reference from SAV */ sav = nextsav; } nextsah = TAILQ_NEXT(sah, chain); key_freesah(&sah); /* release last reference */ sah = nextsah; } } /* * SADB_FLUSH processing * receive * * from the ikmpd, and free all entries in secastree. * and send, * * to the ikmpd. * NOTE: to do is only marking SADB_SASTATE_DEAD. * * m will always be freed. */ static int key_flush(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp) { struct secashead_queue flushq; struct sadb_msg *newmsg; struct secashead *sah, *nextsah; struct secasvar *sav; uint8_t proto; int i; IPSEC_ASSERT(so != NULL, ("null socket")); IPSEC_ASSERT(mhp != NULL, ("null msghdr")); IPSEC_ASSERT(mhp->msg != NULL, ("null msg")); /* map satype to proto */ if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) { ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n", __func__)); return key_senderror(so, m, EINVAL); } KEYDBG(KEY_STAMP, printf("%s: proto %u\n", __func__, proto)); TAILQ_INIT(&flushq); if (proto == IPSEC_PROTO_ANY) { /* no SATYPE specified, i.e. flushing all SA. */ SAHTREE_WLOCK(); /* Move all SAHs into flushq */ TAILQ_CONCAT(&flushq, &V_sahtree, chain); /* Flush all buckets in SPI hash */ for (i = 0; i < V_savhash_mask + 1; i++) LIST_INIT(&V_savhashtbl[i]); /* Flush all buckets in SAHADDRHASH */ for (i = 0; i < V_sahaddrhash_mask + 1; i++) LIST_INIT(&V_sahaddrhashtbl[i]); /* Mark all SAHs as unlinked */ TAILQ_FOREACH(sah, &flushq, chain) { sah->state = SADB_SASTATE_DEAD; /* * Callout handler makes its job using * RLOCK and drain queues. In case, when this * function will be called just before it * acquires WLOCK, we need to mark SAs as * unlinked to prevent second unlink. */ TAILQ_FOREACH(sav, &sah->savtree_larval, chain) { sav->state = SADB_SASTATE_DEAD; } TAILQ_FOREACH(sav, &sah->savtree_alive, chain) { sav->state = SADB_SASTATE_DEAD; } } SAHTREE_WUNLOCK(); } else { SAHTREE_WLOCK(); sah = TAILQ_FIRST(&V_sahtree); while (sah != NULL) { IPSEC_ASSERT(sah->state != SADB_SASTATE_DEAD, ("DEAD SAH %p in SADB_FLUSH", sah)); nextsah = TAILQ_NEXT(sah, chain); if (sah->saidx.proto != proto) { sah = nextsah; continue; } sah->state = SADB_SASTATE_DEAD; TAILQ_REMOVE(&V_sahtree, sah, chain); LIST_REMOVE(sah, addrhash); /* Unlink all SAs from SPI hash */ TAILQ_FOREACH(sav, &sah->savtree_larval, chain) { LIST_REMOVE(sav, spihash); sav->state = SADB_SASTATE_DEAD; } TAILQ_FOREACH(sav, &sah->savtree_alive, chain) { LIST_REMOVE(sav, spihash); sav->state = SADB_SASTATE_DEAD; } /* Add SAH into flushq */ TAILQ_INSERT_HEAD(&flushq, sah, chain); sah = nextsah; } SAHTREE_WUNLOCK(); } key_freesah_flushed(&flushq); /* Free all queued SAs and SAHs */ if (m->m_len < sizeof(struct sadb_msg) || sizeof(struct sadb_msg) > m->m_len + M_TRAILINGSPACE(m)) { ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__)); return key_senderror(so, m, ENOBUFS); } if (m->m_next) m_freem(m->m_next); m->m_next = NULL; m->m_pkthdr.len = m->m_len = sizeof(struct sadb_msg); newmsg = mtod(m, struct sadb_msg *); newmsg->sadb_msg_errno = 0; newmsg->sadb_msg_len = PFKEY_UNIT64(m->m_pkthdr.len); return key_sendup_mbuf(so, m, KEY_SENDUP_ALL); } /* * SADB_DUMP processing * dump all entries including status of DEAD in SAD. * receive * * from the ikmpd, and dump all secasvar leaves * and send, * ..... * to the ikmpd. * * m will always be freed. */ static int key_dump(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp) { SAHTREE_RLOCK_TRACKER; struct secashead *sah; struct secasvar *sav; struct sadb_msg *newmsg; struct mbuf *n; uint32_t cnt; uint8_t proto, satype; IPSEC_ASSERT(so != NULL, ("null socket")); IPSEC_ASSERT(m != NULL, ("null mbuf")); IPSEC_ASSERT(mhp != NULL, ("null msghdr")); IPSEC_ASSERT(mhp->msg != NULL, ("null msg")); /* map satype to proto */ if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) { ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n", __func__)); return key_senderror(so, m, EINVAL); } /* count sav entries to be sent to the userland. */ cnt = 0; SAHTREE_RLOCK(); TAILQ_FOREACH(sah, &V_sahtree, chain) { if (mhp->msg->sadb_msg_satype != SADB_SATYPE_UNSPEC && proto != sah->saidx.proto) continue; TAILQ_FOREACH(sav, &sah->savtree_larval, chain) cnt++; TAILQ_FOREACH(sav, &sah->savtree_alive, chain) cnt++; } if (cnt == 0) { SAHTREE_RUNLOCK(); return key_senderror(so, m, ENOENT); } /* send this to the userland, one at a time. */ newmsg = NULL; TAILQ_FOREACH(sah, &V_sahtree, chain) { if (mhp->msg->sadb_msg_satype != SADB_SATYPE_UNSPEC && proto != sah->saidx.proto) continue; /* map proto to satype */ if ((satype = key_proto2satype(sah->saidx.proto)) == 0) { SAHTREE_RUNLOCK(); ipseclog((LOG_DEBUG, "%s: there was invalid proto in " "SAD.\n", __func__)); return key_senderror(so, m, EINVAL); } TAILQ_FOREACH(sav, &sah->savtree_larval, chain) { n = key_setdumpsa(sav, SADB_DUMP, satype, --cnt, mhp->msg->sadb_msg_pid); if (n == NULL) { SAHTREE_RUNLOCK(); return key_senderror(so, m, ENOBUFS); } key_sendup_mbuf(so, n, KEY_SENDUP_ONE); } TAILQ_FOREACH(sav, &sah->savtree_alive, chain) { n = key_setdumpsa(sav, SADB_DUMP, satype, --cnt, mhp->msg->sadb_msg_pid); if (n == NULL) { SAHTREE_RUNLOCK(); return key_senderror(so, m, ENOBUFS); } key_sendup_mbuf(so, n, KEY_SENDUP_ONE); } } SAHTREE_RUNLOCK(); m_freem(m); return (0); } /* * SADB_X_PROMISC processing * * m will always be freed. */ static int key_promisc(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp) { int olen; IPSEC_ASSERT(so != NULL, ("null socket")); IPSEC_ASSERT(m != NULL, ("null mbuf")); IPSEC_ASSERT(mhp != NULL, ("null msghdr")); IPSEC_ASSERT(mhp->msg != NULL, ("null msg")); olen = PFKEY_UNUNIT64(mhp->msg->sadb_msg_len); if (olen < sizeof(struct sadb_msg)) { #if 1 return key_senderror(so, m, EINVAL); #else m_freem(m); return 0; #endif } else if (olen == sizeof(struct sadb_msg)) { /* enable/disable promisc mode */ struct keycb *kp; if ((kp = (struct keycb *)sotorawcb(so)) == NULL) return key_senderror(so, m, EINVAL); mhp->msg->sadb_msg_errno = 0; switch (mhp->msg->sadb_msg_satype) { case 0: case 1: kp->kp_promisc = mhp->msg->sadb_msg_satype; break; default: return key_senderror(so, m, EINVAL); } /* send the original message back to everyone */ mhp->msg->sadb_msg_errno = 0; return key_sendup_mbuf(so, m, KEY_SENDUP_ALL); } else { /* send packet as is */ m_adj(m, PFKEY_ALIGN8(sizeof(struct sadb_msg))); /* TODO: if sadb_msg_seq is specified, send to specific pid */ return key_sendup_mbuf(so, m, KEY_SENDUP_ALL); } } static int (*key_typesw[])(struct socket *, struct mbuf *, const struct sadb_msghdr *) = { NULL, /* SADB_RESERVED */ key_getspi, /* SADB_GETSPI */ key_update, /* SADB_UPDATE */ key_add, /* SADB_ADD */ key_delete, /* SADB_DELETE */ key_get, /* SADB_GET */ key_acquire2, /* SADB_ACQUIRE */ key_register, /* SADB_REGISTER */ NULL, /* SADB_EXPIRE */ key_flush, /* SADB_FLUSH */ key_dump, /* SADB_DUMP */ key_promisc, /* SADB_X_PROMISC */ NULL, /* SADB_X_PCHANGE */ key_spdadd, /* SADB_X_SPDUPDATE */ key_spdadd, /* SADB_X_SPDADD */ key_spddelete, /* SADB_X_SPDDELETE */ key_spdget, /* SADB_X_SPDGET */ NULL, /* SADB_X_SPDACQUIRE */ key_spddump, /* SADB_X_SPDDUMP */ key_spdflush, /* SADB_X_SPDFLUSH */ key_spdadd, /* SADB_X_SPDSETIDX */ NULL, /* SADB_X_SPDEXPIRE */ key_spddelete2, /* SADB_X_SPDDELETE2 */ }; /* * parse sadb_msg buffer to process PFKEYv2, * and create a data to response if needed. * I think to be dealed with mbuf directly. * IN: * msgp : pointer to pointer to a received buffer pulluped. * This is rewrited to response. * so : pointer to socket. * OUT: * length for buffer to send to user process. */ int key_parse(struct mbuf *m, struct socket *so) { struct sadb_msg *msg; struct sadb_msghdr mh; u_int orglen; int error; int target; IPSEC_ASSERT(so != NULL, ("null socket")); IPSEC_ASSERT(m != NULL, ("null mbuf")); #if 0 /*kdebug_sadb assumes msg in linear buffer*/ KEYDEBUG(KEYDEBUG_KEY_DUMP, ipseclog((LOG_DEBUG, "%s: passed sadb_msg\n", __func__)); kdebug_sadb(msg)); #endif if (m->m_len < sizeof(struct sadb_msg)) { m = m_pullup(m, sizeof(struct sadb_msg)); if (!m) return ENOBUFS; } msg = mtod(m, struct sadb_msg *); orglen = PFKEY_UNUNIT64(msg->sadb_msg_len); target = KEY_SENDUP_ONE; if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len != orglen) { ipseclog((LOG_DEBUG, "%s: invalid message length.\n",__func__)); PFKEYSTAT_INC(out_invlen); error = EINVAL; goto senderror; } if (msg->sadb_msg_version != PF_KEY_V2) { ipseclog((LOG_DEBUG, "%s: PF_KEY version %u is mismatched.\n", __func__, msg->sadb_msg_version)); PFKEYSTAT_INC(out_invver); error = EINVAL; goto senderror; } if (msg->sadb_msg_type > SADB_MAX) { ipseclog((LOG_DEBUG, "%s: invalid type %u is passed.\n", __func__, msg->sadb_msg_type)); PFKEYSTAT_INC(out_invmsgtype); error = EINVAL; goto senderror; } /* for old-fashioned code - should be nuked */ if (m->m_pkthdr.len > MCLBYTES) { m_freem(m); return ENOBUFS; } if (m->m_next) { struct mbuf *n; MGETHDR(n, M_NOWAIT, MT_DATA); if (n && m->m_pkthdr.len > MHLEN) { if (!(MCLGET(n, M_NOWAIT))) { m_free(n); n = NULL; } } if (!n) { m_freem(m); return ENOBUFS; } m_copydata(m, 0, m->m_pkthdr.len, mtod(n, caddr_t)); n->m_pkthdr.len = n->m_len = m->m_pkthdr.len; n->m_next = NULL; m_freem(m); m = n; } /* align the mbuf chain so that extensions are in contiguous region. */ error = key_align(m, &mh); if (error) return error; msg = mh.msg; /* check SA type */ switch (msg->sadb_msg_satype) { case SADB_SATYPE_UNSPEC: switch (msg->sadb_msg_type) { case SADB_GETSPI: case SADB_UPDATE: case SADB_ADD: case SADB_DELETE: case SADB_GET: case SADB_ACQUIRE: case SADB_EXPIRE: ipseclog((LOG_DEBUG, "%s: must specify satype " "when msg type=%u.\n", __func__, msg->sadb_msg_type)); PFKEYSTAT_INC(out_invsatype); error = EINVAL; goto senderror; } break; case SADB_SATYPE_AH: case SADB_SATYPE_ESP: case SADB_X_SATYPE_IPCOMP: case SADB_X_SATYPE_TCPSIGNATURE: switch (msg->sadb_msg_type) { case SADB_X_SPDADD: case SADB_X_SPDDELETE: case SADB_X_SPDGET: case SADB_X_SPDDUMP: case SADB_X_SPDFLUSH: case SADB_X_SPDSETIDX: case SADB_X_SPDUPDATE: case SADB_X_SPDDELETE2: ipseclog((LOG_DEBUG, "%s: illegal satype=%u\n", __func__, msg->sadb_msg_type)); PFKEYSTAT_INC(out_invsatype); error = EINVAL; goto senderror; } break; case SADB_SATYPE_RSVP: case SADB_SATYPE_OSPFV2: case SADB_SATYPE_RIPV2: case SADB_SATYPE_MIP: ipseclog((LOG_DEBUG, "%s: type %u isn't supported.\n", __func__, msg->sadb_msg_satype)); PFKEYSTAT_INC(out_invsatype); error = EOPNOTSUPP; goto senderror; case 1: /* XXX: What does it do? */ if (msg->sadb_msg_type == SADB_X_PROMISC) break; /*FALLTHROUGH*/ default: ipseclog((LOG_DEBUG, "%s: invalid type %u is passed.\n", __func__, msg->sadb_msg_satype)); PFKEYSTAT_INC(out_invsatype); error = EINVAL; goto senderror; } /* check field of upper layer protocol and address family */ if (mh.ext[SADB_EXT_ADDRESS_SRC] != NULL && mh.ext[SADB_EXT_ADDRESS_DST] != NULL) { struct sadb_address *src0, *dst0; u_int plen; src0 = (struct sadb_address *)(mh.ext[SADB_EXT_ADDRESS_SRC]); dst0 = (struct sadb_address *)(mh.ext[SADB_EXT_ADDRESS_DST]); /* check upper layer protocol */ if (src0->sadb_address_proto != dst0->sadb_address_proto) { ipseclog((LOG_DEBUG, "%s: upper layer protocol " "mismatched.\n", __func__)); PFKEYSTAT_INC(out_invaddr); error = EINVAL; goto senderror; } /* check family */ if (PFKEY_ADDR_SADDR(src0)->sa_family != PFKEY_ADDR_SADDR(dst0)->sa_family) { ipseclog((LOG_DEBUG, "%s: address family mismatched.\n", __func__)); PFKEYSTAT_INC(out_invaddr); error = EINVAL; goto senderror; } if (PFKEY_ADDR_SADDR(src0)->sa_len != PFKEY_ADDR_SADDR(dst0)->sa_len) { ipseclog((LOG_DEBUG, "%s: address struct size " "mismatched.\n", __func__)); PFKEYSTAT_INC(out_invaddr); error = EINVAL; goto senderror; } switch (PFKEY_ADDR_SADDR(src0)->sa_family) { case AF_INET: if (PFKEY_ADDR_SADDR(src0)->sa_len != sizeof(struct sockaddr_in)) { PFKEYSTAT_INC(out_invaddr); error = EINVAL; goto senderror; } break; case AF_INET6: if (PFKEY_ADDR_SADDR(src0)->sa_len != sizeof(struct sockaddr_in6)) { PFKEYSTAT_INC(out_invaddr); error = EINVAL; goto senderror; } break; default: ipseclog((LOG_DEBUG, "%s: unsupported address family\n", __func__)); PFKEYSTAT_INC(out_invaddr); error = EAFNOSUPPORT; goto senderror; } switch (PFKEY_ADDR_SADDR(src0)->sa_family) { case AF_INET: plen = sizeof(struct in_addr) << 3; break; case AF_INET6: plen = sizeof(struct in6_addr) << 3; break; default: plen = 0; /*fool gcc*/ break; } /* check max prefix length */ if (src0->sadb_address_prefixlen > plen || dst0->sadb_address_prefixlen > plen) { ipseclog((LOG_DEBUG, "%s: illegal prefixlen.\n", __func__)); PFKEYSTAT_INC(out_invaddr); error = EINVAL; goto senderror; } /* * prefixlen == 0 is valid because there can be a case when * all addresses are matched. */ } if (msg->sadb_msg_type >= nitems(key_typesw) || key_typesw[msg->sadb_msg_type] == NULL) { PFKEYSTAT_INC(out_invmsgtype); error = EINVAL; goto senderror; } return (*key_typesw[msg->sadb_msg_type])(so, m, &mh); senderror: msg->sadb_msg_errno = error; return key_sendup_mbuf(so, m, target); } static int key_senderror(struct socket *so, struct mbuf *m, int code) { struct sadb_msg *msg; IPSEC_ASSERT(m->m_len >= sizeof(struct sadb_msg), ("mbuf too small, len %u", m->m_len)); msg = mtod(m, struct sadb_msg *); msg->sadb_msg_errno = code; return key_sendup_mbuf(so, m, KEY_SENDUP_ONE); } /* * set the pointer to each header into message buffer. * m will be freed on error. * XXX larger-than-MCLBYTES extension? */ static int key_align(struct mbuf *m, struct sadb_msghdr *mhp) { struct mbuf *n; struct sadb_ext *ext; size_t off, end; int extlen; int toff; IPSEC_ASSERT(m != NULL, ("null mbuf")); IPSEC_ASSERT(mhp != NULL, ("null msghdr")); IPSEC_ASSERT(m->m_len >= sizeof(struct sadb_msg), ("mbuf too small, len %u", m->m_len)); /* initialize */ bzero(mhp, sizeof(*mhp)); mhp->msg = mtod(m, struct sadb_msg *); mhp->ext[0] = (struct sadb_ext *)mhp->msg; /*XXX backward compat */ end = PFKEY_UNUNIT64(mhp->msg->sadb_msg_len); extlen = end; /*just in case extlen is not updated*/ for (off = sizeof(struct sadb_msg); off < end; off += extlen) { n = m_pulldown(m, off, sizeof(struct sadb_ext), &toff); if (!n) { /* m is already freed */ return ENOBUFS; } ext = (struct sadb_ext *)(mtod(n, caddr_t) + toff); /* set pointer */ switch (ext->sadb_ext_type) { case SADB_EXT_SA: case SADB_EXT_ADDRESS_SRC: case SADB_EXT_ADDRESS_DST: case SADB_EXT_ADDRESS_PROXY: case SADB_EXT_LIFETIME_CURRENT: case SADB_EXT_LIFETIME_HARD: case SADB_EXT_LIFETIME_SOFT: case SADB_EXT_KEY_AUTH: case SADB_EXT_KEY_ENCRYPT: case SADB_EXT_IDENTITY_SRC: case SADB_EXT_IDENTITY_DST: case SADB_EXT_SENSITIVITY: case SADB_EXT_PROPOSAL: case SADB_EXT_SUPPORTED_AUTH: case SADB_EXT_SUPPORTED_ENCRYPT: case SADB_EXT_SPIRANGE: case SADB_X_EXT_POLICY: case SADB_X_EXT_SA2: #ifdef IPSEC_NAT_T case SADB_X_EXT_NAT_T_TYPE: case SADB_X_EXT_NAT_T_SPORT: case SADB_X_EXT_NAT_T_DPORT: case SADB_X_EXT_NAT_T_OAI: case SADB_X_EXT_NAT_T_OAR: case SADB_X_EXT_NAT_T_FRAG: #endif /* duplicate check */ /* * XXX Are there duplication payloads of either * KEY_AUTH or KEY_ENCRYPT ? */ if (mhp->ext[ext->sadb_ext_type] != NULL) { ipseclog((LOG_DEBUG, "%s: duplicate ext_type " "%u\n", __func__, ext->sadb_ext_type)); m_freem(m); PFKEYSTAT_INC(out_dupext); return EINVAL; } break; default: ipseclog((LOG_DEBUG, "%s: invalid ext_type %u\n", __func__, ext->sadb_ext_type)); m_freem(m); PFKEYSTAT_INC(out_invexttype); return EINVAL; } extlen = PFKEY_UNUNIT64(ext->sadb_ext_len); if (key_validate_ext(ext, extlen)) { m_freem(m); PFKEYSTAT_INC(out_invlen); return EINVAL; } n = m_pulldown(m, off, extlen, &toff); if (!n) { /* m is already freed */ return ENOBUFS; } ext = (struct sadb_ext *)(mtod(n, caddr_t) + toff); mhp->ext[ext->sadb_ext_type] = ext; mhp->extoff[ext->sadb_ext_type] = off; mhp->extlen[ext->sadb_ext_type] = extlen; } if (off != end) { m_freem(m); PFKEYSTAT_INC(out_invlen); return EINVAL; } return 0; } static int key_validate_ext(const struct sadb_ext *ext, int len) { const struct sockaddr *sa; enum { NONE, ADDR } checktype = NONE; int baselen = 0; const int sal = offsetof(struct sockaddr, sa_len) + sizeof(sa->sa_len); if (len != PFKEY_UNUNIT64(ext->sadb_ext_len)) return EINVAL; /* if it does not match minimum/maximum length, bail */ if (ext->sadb_ext_type >= nitems(minsize) || ext->sadb_ext_type >= nitems(maxsize)) return EINVAL; if (!minsize[ext->sadb_ext_type] || len < minsize[ext->sadb_ext_type]) return EINVAL; if (maxsize[ext->sadb_ext_type] && len > maxsize[ext->sadb_ext_type]) return EINVAL; /* more checks based on sadb_ext_type XXX need more */ switch (ext->sadb_ext_type) { case SADB_EXT_ADDRESS_SRC: case SADB_EXT_ADDRESS_DST: case SADB_EXT_ADDRESS_PROXY: baselen = PFKEY_ALIGN8(sizeof(struct sadb_address)); checktype = ADDR; break; case SADB_EXT_IDENTITY_SRC: case SADB_EXT_IDENTITY_DST: if (((const struct sadb_ident *)ext)->sadb_ident_type == SADB_X_IDENTTYPE_ADDR) { baselen = PFKEY_ALIGN8(sizeof(struct sadb_ident)); checktype = ADDR; } else checktype = NONE; break; default: checktype = NONE; break; } switch (checktype) { case NONE: break; case ADDR: sa = (const struct sockaddr *)(((const u_int8_t*)ext)+baselen); if (len < baselen + sal) return EINVAL; if (baselen + PFKEY_ALIGN8(sa->sa_len) != len) return EINVAL; break; } return 0; } void key_init(void) { int i; for (i = 0; i < IPSEC_DIR_MAX; i++) TAILQ_INIT(&V_sptree[i]); V_key_lft_zone = uma_zcreate("IPsec SA lft_c", sizeof(uint64_t) * 2, NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_PCPU); TAILQ_INIT(&V_sahtree); V_sphashtbl = hashinit(SPHASH_NHASH, M_IPSEC_SP, &V_sphash_mask); V_savhashtbl = hashinit(SAVHASH_NHASH, M_IPSEC_SA, &V_savhash_mask); V_sahaddrhashtbl = hashinit(SAHHASH_NHASH, M_IPSEC_SAH, &V_sahaddrhash_mask); V_acqaddrhashtbl = hashinit(ACQHASH_NHASH, M_IPSEC_SAQ, &V_acqaddrhash_mask); V_acqseqhashtbl = hashinit(ACQHASH_NHASH, M_IPSEC_SAQ, &V_acqseqhash_mask); for (i = 0; i <= SADB_SATYPE_MAX; i++) LIST_INIT(&V_regtree[i]); LIST_INIT(&V_acqtree); LIST_INIT(&V_spacqtree); if (!IS_DEFAULT_VNET(curvnet)) return; SPTREE_LOCK_INIT(); REGTREE_LOCK_INIT(); SAHTREE_LOCK_INIT(); ACQ_LOCK_INIT(); SPACQ_LOCK_INIT(); #ifndef IPSEC_DEBUG2 callout_init(&key_timer, 1); callout_reset(&key_timer, hz, key_timehandler, NULL); #endif /*IPSEC_DEBUG2*/ /* initialize key statistics */ keystat.getspi_count = 1; if (bootverbose) printf("IPsec: Initialized Security Association Processing.\n"); } #ifdef VIMAGE void key_destroy(void) { struct secashead_queue sahdrainq; struct secpolicy_queue drainq; struct secpolicy *sp, *nextsp; struct secacq *acq, *nextacq; struct secspacq *spacq, *nextspacq; struct secashead *sah, *nextsah; struct secreg *reg; int i; /* * XXX: can we just call free() for each object without * walking through safe way with releasing references? */ TAILQ_INIT(&drainq); SPTREE_WLOCK(); for (i = 0; i < IPSEC_DIR_MAX; i++) { TAILQ_CONCAT(&drainq, &V_sptree[i], chain); } SPTREE_WUNLOCK(); sp = TAILQ_FIRST(&drainq); while (sp != NULL) { nextsp = TAILQ_NEXT(sp, chain); - KEY_FREESP(&sp); + key_freesp(&sp); sp = nextsp; } TAILQ_INIT(&sahdrainq); SAHTREE_WLOCK(); TAILQ_CONCAT(&sahdrainq, &V_sahtree, chain); for (i = 0; i < V_savhash_mask + 1; i++) LIST_INIT(&V_savhashtbl[i]); for (i = 0; i < V_sahaddrhash_mask + 1; i++) LIST_INIT(&V_sahaddrhashtbl[i]); TAILQ_FOREACH(sah, &sahdrainq, chain) { sah->state = SADB_SASTATE_DEAD; TAILQ_FOREACH(sav, &sah->savtree_larval, chain) { sav->state = SADB_SASTATE_DEAD; } TAILQ_FOREACH(sav, &sah->savtree_alive, chain) { sav->state = SADB_SASTATE_DEAD; } } SAHTREE_WUNLOCK(); key_freesah_flushed(&sahdrainq); hashdestroy(V_sphashtbl, M_IPSEC_SP, V_sphash_mask); hashdestroy(V_savhashtbl, M_IPSEC_SA, V_savhash_mask); hashdestroy(V_sahaddrhashtbl, M_IPSEC_SAH, V_sahaddrhash_mask); REGTREE_LOCK(); for (i = 0; i <= SADB_SATYPE_MAX; i++) { LIST_FOREACH(reg, &V_regtree[i], chain) { if (__LIST_CHAINED(reg)) { LIST_REMOVE(reg, chain); free(reg, M_IPSEC_SAR); break; } } } REGTREE_UNLOCK(); ACQ_LOCK(); acq = LIST_FIRST(&V_acqtree); while (acq != NULL) { nextacq = LIST_NEXT(acq, chain); LIST_REMOVE(acq, chain); free(acq, M_IPSEC_SAQ); acq = nextacq; } ACQ_UNLOCK(); SPACQ_LOCK(); for (spacq = LIST_FIRST(&V_spacqtree); spacq != NULL; spacq = nextspacq) { nextspacq = LIST_NEXT(spacq, chain); if (__LIST_CHAINED(spacq)) { LIST_REMOVE(spacq, chain); free(spacq, M_IPSEC_SAQ); } } SPACQ_UNLOCK(); hashdestroy(V_acqaddrhashtbl, M_IPSEC_SAQ, V_acqaddrhash_mask); hashdestroy(V_acqseqhashtbl, M_IPSEC_SAQ, V_acqseqhash_mask); uma_zdestroy(V_key_lft_zone); } #endif /* record data transfer on SA, and update timestamps */ void key_sa_recordxfer(struct secasvar *sav, struct mbuf *m) { IPSEC_ASSERT(sav != NULL, ("Null secasvar")); IPSEC_ASSERT(m != NULL, ("Null mbuf")); /* * XXX Currently, there is a difference of bytes size * between inbound and outbound processing. */ counter_u64_add(sav->lft_c_bytes, m->m_pkthdr.len); /* * We use the number of packets as the unit of * allocations. We increment the variable * whenever {esp,ah}_{in,out}put is called. */ counter_u64_add(sav->lft_c_allocations, 1); /* * NOTE: We record CURRENT usetime by using wall clock, * in seconds. HARD and SOFT lifetime are measured by the time * difference (again in seconds) from usetime. * * usetime * v expire expire * -----+-----+--------+---> t * <--------------> HARD * <-----> SOFT */ if (sav->firstused == 0) sav->firstused = time_second; } /* * Take one of the kernel's security keys and convert it into a PF_KEY * structure within an mbuf, suitable for sending up to a waiting * application in user land. * * IN: * src: A pointer to a kernel security key. * exttype: Which type of key this is. Refer to the PF_KEY data structures. * OUT: * a valid mbuf or NULL indicating an error * */ static struct mbuf * key_setkey(struct seckey *src, u_int16_t exttype) { struct mbuf *m; struct sadb_key *p; int len; if (src == NULL) return NULL; len = PFKEY_ALIGN8(sizeof(struct sadb_key) + _KEYLEN(src)); m = m_get2(len, M_NOWAIT, MT_DATA, 0); if (m == NULL) return NULL; m_align(m, len); m->m_len = len; p = mtod(m, struct sadb_key *); bzero(p, len); p->sadb_key_len = PFKEY_UNIT64(len); p->sadb_key_exttype = exttype; p->sadb_key_bits = src->bits; bcopy(src->key_data, _KEYBUF(p), _KEYLEN(src)); return m; } /* * Take one of the kernel's lifetime data structures and convert it * into a PF_KEY structure within an mbuf, suitable for sending up to * a waiting application in user land. * * IN: * src: A pointer to a kernel lifetime structure. * exttype: Which type of lifetime this is. Refer to the PF_KEY * data structures for more information. * OUT: * a valid mbuf or NULL indicating an error * */ static struct mbuf * key_setlifetime(struct seclifetime *src, u_int16_t exttype) { struct mbuf *m = NULL; struct sadb_lifetime *p; int len = PFKEY_ALIGN8(sizeof(struct sadb_lifetime)); if (src == NULL) return NULL; m = m_get2(len, M_NOWAIT, MT_DATA, 0); if (m == NULL) return m; m_align(m, len); m->m_len = len; p = mtod(m, struct sadb_lifetime *); bzero(p, len); p->sadb_lifetime_len = PFKEY_UNIT64(len); p->sadb_lifetime_exttype = exttype; p->sadb_lifetime_allocations = src->allocations; p->sadb_lifetime_bytes = src->bytes; p->sadb_lifetime_addtime = src->addtime; p->sadb_lifetime_usetime = src->usetime; return m; } Index: projects/ipsec/sys/netipsec/key.h =================================================================== --- projects/ipsec/sys/netipsec/key.h (revision 309109) +++ projects/ipsec/sys/netipsec/key.h (revision 309110) @@ -1,134 +1,95 @@ /* $FreeBSD$ */ /* $KAME: key.h,v 1.21 2001/07/27 03:51:30 itojun Exp $ */ /*- * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. * 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. Neither the name of the project 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 PROJECT 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 PROJECT OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #ifndef _NETIPSEC_KEY_H_ #define _NETIPSEC_KEY_H_ #ifdef _KERNEL struct secpolicy; struct secpolicyindex; -struct ipsecrequest; struct secasvar; struct sockaddr; struct socket; struct sadb_msg; struct sadb_x_policy; struct secasindex; union sockaddr_union; +struct secpolicy *key_newsp(void); +struct secpolicy *key_allocsp(struct secpolicyindex *, u_int); +struct secpolicy *key_msg2sp(struct sadb_x_policy *, size_t, int *); +int key_sp2msg(struct secpolicy *, void *, size_t *); +void key_addref(struct secpolicy *); +void key_freesp(struct secpolicy **); +int key_spdacquire(struct secpolicy *); +int key_havesp(u_int); uint32_t key_getspgen(void); +uint32_t key_newreqid(void); -extern void key_addref(struct secpolicy *sp); -extern int key_havesp(u_int dir); -extern struct secpolicy *key_allocsp(struct secpolicyindex *, u_int, - const char*, int); -extern struct secpolicy *key_allocsp2(u_int32_t spi, union sockaddr_union *dst, - u_int8_t proto, u_int dir, const char*, int); -extern struct secpolicy *key_newsp(const char*, int); -#if 0 -extern struct secpolicy *key_gettunnel(const struct sockaddr *, - const struct sockaddr *, const struct sockaddr *, - const struct sockaddr *, const char*, int); -#endif -/* NB: prepend with _ for KAME IPv6 compatbility */ -extern void _key_freesp(struct secpolicy **, const char*, int); +struct secasvar *key_allocsa(union sockaddr_union *, uint8_t, uint32_t); +struct secasvar *key_allocsa_tunnel(union sockaddr_union *, + union sockaddr_union *, uint8_t); +struct secasvar *key_allocsa_policy(struct secpolicy *, + const struct secasindex *, int *); +void key_freesav(struct secasvar **); -#define KEY_ALLOCSP(spidx, dir) \ - key_allocsp(spidx, dir, __FILE__, __LINE__) -#define KEY_ALLOCSP2(spi, dst, proto, dir) \ - key_allocsp2(spi, dst, proto, dir, __FILE__, __LINE__) -#define KEY_NEWSP() \ - key_newsp(__FILE__, __LINE__) -#if 0 -#define KEY_GETTUNNEL(osrc, odst, isrc, idst) \ - key_gettunnel(osrc, odst, isrc, idst, __FILE__, __LINE__) -#endif -#define KEY_FREESP(spp) \ - _key_freesp(spp, __FILE__, __LINE__) +int key_sockaddrcmp(const struct sockaddr *, const struct sockaddr *, int); +int key_sockaddrcmp_withmask(const struct sockaddr *, const struct sockaddr *, + size_t); -extern struct secasvar *key_allocsa(union sockaddr_union *, u_int, u_int32_t, - const char*, int); -extern struct secasvar *key_allocsa_tunnel(union sockaddr_union *, - union sockaddr_union *, u_int, const char*, int); -extern void key_addrefsa(struct secasvar *, const char*, int); -extern void key_freesav(struct secasvar **, const char*, int); - -#define KEY_ALLOCSA(dst, proto, spi) \ - key_allocsa(dst, proto, spi, __FILE__, __LINE__) -#define KEY_ALLOCSA_TUNNEL(src, dst, proto) \ - key_allocsa_tunnel(src, dst, proto, __FILE__, __LINE__) -#define KEY_ADDREFSA(sav) \ - key_addrefsa(sav, __FILE__, __LINE__) -#define KEY_FREESAV(psav) \ - key_freesav(psav, __FILE__, __LINE__) - -extern void key_freeso(struct socket *); -extern int key_checktunnelsanity(struct secasvar *, u_int, - caddr_t, caddr_t); -extern int key_checkrequest(struct ipsecrequest *isr, - const struct secasindex *); -extern struct secpolicy *key_msg2sp(struct sadb_x_policy *, - size_t, int *); - -int key_sp2msg(struct secpolicy *, void *request, size_t *len); - -extern int key_ismyaddr(struct sockaddr *); -extern int key_spdacquire(struct secpolicy *); extern u_long key_random(void); extern void key_randomfill(void *, size_t); extern void key_freereg(struct socket *); extern int key_parse(struct mbuf *, struct socket *); extern void key_init(void); #ifdef VIMAGE extern void key_destroy(void); #endif extern void key_sa_recordxfer(struct secasvar *, struct mbuf *); #ifdef IPSEC_NAT_T -u_int16_t key_portfromsaddr(struct sockaddr *); -#define KEY_PORTFROMSADDR(saddr) \ - key_portfromsaddr((struct sockaddr *)(saddr)) +uint16_t key_portfromsaddr(struct sockaddr *); #endif #ifdef MALLOC_DECLARE MALLOC_DECLARE(M_IPSEC_SA); MALLOC_DECLARE(M_IPSEC_SAH); MALLOC_DECLARE(M_IPSEC_SP); MALLOC_DECLARE(M_IPSEC_SR); MALLOC_DECLARE(M_IPSEC_MISC); MALLOC_DECLARE(M_IPSEC_SAQ); MALLOC_DECLARE(M_IPSEC_SAR); MALLOC_DECLARE(M_IPSEC_INPCB); #endif /* MALLOC_DECLARE */ #endif /* defined(_KERNEL) */ #endif /* _NETIPSEC_KEY_H_ */ Index: projects/ipsec/sys/netipsec/xform_ah.c =================================================================== --- projects/ipsec/sys/netipsec/xform_ah.c (revision 309109) +++ projects/ipsec/sys/netipsec/xform_ah.c (revision 309110) @@ -1,1180 +1,1180 @@ /* $FreeBSD$ */ /* $OpenBSD: ip_ah.c,v 1.63 2001/06/26 06:18:58 angelos Exp $ */ /*- * The authors of this code are John Ioannidis (ji@tla.org), * Angelos D. Keromytis (kermit@csd.uch.gr) and * Niels Provos (provos@physnet.uni-hamburg.de). * * The original version of this code was written by John Ioannidis * for BSD/OS in Athens, Greece, in November 1995. * * Ported to OpenBSD and NetBSD, with additional transforms, in December 1996, * by Angelos D. Keromytis. * * Additional transforms and features in 1997 and 1998 by Angelos D. Keromytis * and Niels Provos. * * Additional features in 1999 by Angelos D. Keromytis and Niklas Hallqvist. * * Copyright (c) 1995, 1996, 1997, 1998, 1999 by John Ioannidis, * Angelos D. Keromytis and Niels Provos. * Copyright (c) 1999 Niklas Hallqvist. * Copyright (c) 2001 Angelos D. Keromytis. * * Permission to use, copy, and modify this software with or without fee * is hereby granted, provided that this entire notice is included in * all copies of any software which is or includes a copy or * modification of this software. * You may use this code under the GNU public license if you so wish. Please * contribute changes back to the authors under this freer than GPL license * so that we may further the use of strong encryption without limitations to * all. * * THIS SOFTWARE IS BEING PROVIDED "AS IS", WITHOUT ANY EXPRESS OR * IMPLIED WARRANTY. IN PARTICULAR, NONE OF THE AUTHORS MAKES ANY * REPRESENTATION OR WARRANTY OF ANY KIND CONCERNING THE * MERCHANTABILITY OF THIS SOFTWARE OR ITS FITNESS FOR ANY PARTICULAR * PURPOSE. */ #include "opt_inet.h" #include "opt_inet6.h" #include #include #include #include #include #include #include -#include +#include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef INET6 #include #include #include #endif #include #include #include /* * Return header size in bytes. The old protocol did not support * the replay counter; the new protocol always includes the counter. */ #define HDRSIZE(sav) \ (((sav)->flags & SADB_X_EXT_OLD) ? \ sizeof (struct ah) : sizeof (struct ah) + sizeof (u_int32_t)) /* * Return authenticator size in bytes, based on a field in the * algorithm descriptor. */ #define AUTHSIZE(sav) ((sav->flags & SADB_X_EXT_OLD) ? 16 : \ xform_ah_authsize((sav)->tdb_authalgxform)) VNET_DEFINE(int, ah_enable) = 1; /* control flow of packets with AH */ VNET_DEFINE(int, ah_cleartos) = 1; /* clear ip_tos when doing AH calc */ VNET_PCPUSTAT_DEFINE(struct ahstat, ahstat); VNET_PCPUSTAT_SYSINIT(ahstat); #ifdef VIMAGE VNET_PCPUSTAT_SYSUNINIT(ahstat); #endif /* VIMAGE */ #ifdef INET SYSCTL_DECL(_net_inet_ah); SYSCTL_INT(_net_inet_ah, OID_AUTO, ah_enable, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ah_enable), 0, ""); SYSCTL_INT(_net_inet_ah, OID_AUTO, ah_cleartos, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ah_cleartos), 0, ""); SYSCTL_VNET_PCPUSTAT(_net_inet_ah, IPSECCTL_STATS, stats, struct ahstat, ahstat, "AH statistics (struct ahstat, netipsec/ah_var.h)"); #endif static unsigned char ipseczeroes[256]; /* larger than an ip6 extension hdr */ static int ah_input_cb(struct cryptop*); static int ah_output_cb(struct cryptop*); int xform_ah_authsize(struct auth_hash *esph) { int alen; if (esph == NULL) return 0; switch (esph->type) { case CRYPTO_SHA2_256_HMAC: case CRYPTO_SHA2_384_HMAC: case CRYPTO_SHA2_512_HMAC: alen = esph->hashsize / 2; /* RFC4868 2.3 */ break; case CRYPTO_AES_128_NIST_GMAC: case CRYPTO_AES_192_NIST_GMAC: case CRYPTO_AES_256_NIST_GMAC: alen = esph->hashsize; break; default: alen = AH_HMAC_HASHLEN; break; } return alen; } /* * NB: this is public for use by the PF_KEY support. */ struct auth_hash * ah_algorithm_lookup(int alg) { if (alg > SADB_AALG_MAX) return NULL; switch (alg) { case SADB_X_AALG_NULL: return &auth_hash_null; case SADB_AALG_MD5HMAC: return &auth_hash_hmac_md5; case SADB_AALG_SHA1HMAC: return &auth_hash_hmac_sha1; case SADB_X_AALG_RIPEMD160HMAC: return &auth_hash_hmac_ripemd_160; case SADB_X_AALG_MD5: return &auth_hash_key_md5; case SADB_X_AALG_SHA: return &auth_hash_key_sha1; case SADB_X_AALG_SHA2_256: return &auth_hash_hmac_sha2_256; case SADB_X_AALG_SHA2_384: return &auth_hash_hmac_sha2_384; case SADB_X_AALG_SHA2_512: return &auth_hash_hmac_sha2_512; case SADB_X_AALG_AES128GMAC: return &auth_hash_nist_gmac_aes_128; case SADB_X_AALG_AES192GMAC: return &auth_hash_nist_gmac_aes_192; case SADB_X_AALG_AES256GMAC: return &auth_hash_nist_gmac_aes_256; } return NULL; } size_t ah_hdrsiz(struct secasvar *sav) { size_t size; if (sav != NULL) { int authsize; IPSEC_ASSERT(sav->tdb_authalgxform != NULL, ("null xform")); /*XXX not right for null algorithm--does it matter??*/ authsize = AUTHSIZE(sav); size = roundup(authsize, sizeof (u_int32_t)) + HDRSIZE(sav); } else { /* default guess */ size = sizeof (struct ah) + sizeof (u_int32_t) + 16; } return size; } /* * NB: public for use by esp_init. */ int ah_init0(struct secasvar *sav, struct xformsw *xsp, struct cryptoini *cria) { struct auth_hash *thash; int keylen; thash = ah_algorithm_lookup(sav->alg_auth); if (thash == NULL) { DPRINTF(("%s: unsupported authentication algorithm %u\n", __func__, sav->alg_auth)); return EINVAL; } /* * Verify the replay state block allocation is consistent with * the protocol type. We check here so we can make assumptions * later during protocol processing. */ /* NB: replay state is setup elsewhere (sigh) */ if (((sav->flags&SADB_X_EXT_OLD) == 0) ^ (sav->replay != NULL)) { DPRINTF(("%s: replay state block inconsistency, " "%s algorithm %s replay state\n", __func__, (sav->flags & SADB_X_EXT_OLD) ? "old" : "new", sav->replay == NULL ? "without" : "with")); return EINVAL; } if (sav->key_auth == NULL) { DPRINTF(("%s: no authentication key for %s algorithm\n", __func__, thash->name)); return EINVAL; } keylen = _KEYLEN(sav->key_auth); if (keylen != thash->keysize && thash->keysize != 0) { DPRINTF(("%s: invalid keylength %d, algorithm %s requires " "keysize %d\n", __func__, keylen, thash->name, thash->keysize)); return EINVAL; } sav->tdb_xform = xsp; sav->tdb_authalgxform = thash; /* Initialize crypto session. */ bzero(cria, sizeof (*cria)); cria->cri_alg = sav->tdb_authalgxform->type; cria->cri_klen = _KEYBITS(sav->key_auth); cria->cri_key = sav->key_auth->key_data; cria->cri_mlen = AUTHSIZE(sav); return 0; } /* * ah_init() is called when an SPI is being set up. */ static int ah_init(struct secasvar *sav, struct xformsw *xsp) { struct cryptoini cria; int error; error = ah_init0(sav, xsp, &cria); return error ? error : crypto_newsession(&sav->tdb_cryptoid, &cria, V_crypto_support); } /* * Paranoia. * * NB: public for use by esp_zeroize (XXX). */ int ah_zeroize(struct secasvar *sav) { int err; if (sav->key_auth) bzero(sav->key_auth->key_data, _KEYLEN(sav->key_auth)); err = crypto_freesession(sav->tdb_cryptoid); sav->tdb_cryptoid = 0; sav->tdb_authalgxform = NULL; sav->tdb_xform = NULL; return err; } /* * Massage IPv4/IPv6 headers for AH processing. */ static int ah_massage_headers(struct mbuf **m0, int proto, int skip, int alg, int out) { struct mbuf *m = *m0; unsigned char *ptr; int off, count; #ifdef INET struct ip *ip; #endif /* INET */ #ifdef INET6 struct ip6_ext *ip6e; struct ip6_hdr ip6; int alloc, len, ad; #endif /* INET6 */ switch (proto) { #ifdef INET case AF_INET: /* * This is the least painful way of dealing with IPv4 header * and option processing -- just make sure they're in * contiguous memory. */ *m0 = m = m_pullup(m, skip); if (m == NULL) { DPRINTF(("%s: m_pullup failed\n", __func__)); return ENOBUFS; } /* Fix the IP header */ ip = mtod(m, struct ip *); if (V_ah_cleartos) ip->ip_tos = 0; ip->ip_ttl = 0; ip->ip_sum = 0; if (alg == CRYPTO_MD5_KPDK || alg == CRYPTO_SHA1_KPDK) ip->ip_off &= htons(IP_DF); else ip->ip_off = htons(0); ptr = mtod(m, unsigned char *) + sizeof(struct ip); /* IPv4 option processing */ for (off = sizeof(struct ip); off < skip;) { if (ptr[off] == IPOPT_EOL || ptr[off] == IPOPT_NOP || off + 1 < skip) ; else { DPRINTF(("%s: illegal IPv4 option length for " "option %d\n", __func__, ptr[off])); m_freem(m); return EINVAL; } switch (ptr[off]) { case IPOPT_EOL: off = skip; /* End the loop. */ break; case IPOPT_NOP: off++; break; case IPOPT_SECURITY: /* 0x82 */ case 0x85: /* Extended security. */ case 0x86: /* Commercial security. */ case 0x94: /* Router alert */ case 0x95: /* RFC1770 */ /* Sanity check for option length. */ if (ptr[off + 1] < 2) { DPRINTF(("%s: illegal IPv4 option " "length for option %d\n", __func__, ptr[off])); m_freem(m); return EINVAL; } off += ptr[off + 1]; break; case IPOPT_LSRR: case IPOPT_SSRR: /* Sanity check for option length. */ if (ptr[off + 1] < 2) { DPRINTF(("%s: illegal IPv4 option " "length for option %d\n", __func__, ptr[off])); m_freem(m); return EINVAL; } /* * On output, if we have either of the * source routing options, we should * swap the destination address of the * IP header with the last address * specified in the option, as that is * what the destination's IP header * will look like. */ if (out) bcopy(ptr + off + ptr[off + 1] - sizeof(struct in_addr), &(ip->ip_dst), sizeof(struct in_addr)); /* Fall through */ default: /* Sanity check for option length. */ if (ptr[off + 1] < 2) { DPRINTF(("%s: illegal IPv4 option " "length for option %d\n", __func__, ptr[off])); m_freem(m); return EINVAL; } /* Zeroize all other options. */ count = ptr[off + 1]; bcopy(ipseczeroes, ptr, count); off += count; break; } /* Sanity check. */ if (off > skip) { DPRINTF(("%s: malformed IPv4 options header\n", __func__)); m_freem(m); return EINVAL; } } break; #endif /* INET */ #ifdef INET6 case AF_INET6: /* Ugly... */ /* Copy and "cook" the IPv6 header. */ m_copydata(m, 0, sizeof(ip6), (caddr_t) &ip6); /* We don't do IPv6 Jumbograms. */ if (ip6.ip6_plen == 0) { DPRINTF(("%s: unsupported IPv6 jumbogram\n", __func__)); m_freem(m); return EMSGSIZE; } ip6.ip6_flow = 0; ip6.ip6_hlim = 0; ip6.ip6_vfc &= ~IPV6_VERSION_MASK; ip6.ip6_vfc |= IPV6_VERSION; /* Scoped address handling. */ if (IN6_IS_SCOPE_LINKLOCAL(&ip6.ip6_src)) ip6.ip6_src.s6_addr16[1] = 0; if (IN6_IS_SCOPE_LINKLOCAL(&ip6.ip6_dst)) ip6.ip6_dst.s6_addr16[1] = 0; /* Done with IPv6 header. */ m_copyback(m, 0, sizeof(struct ip6_hdr), (caddr_t) &ip6); /* Let's deal with the remaining headers (if any). */ if (skip - sizeof(struct ip6_hdr) > 0) { if (m->m_len <= skip) { ptr = (unsigned char *) malloc( skip - sizeof(struct ip6_hdr), M_XDATA, M_NOWAIT); if (ptr == NULL) { DPRINTF(("%s: failed to allocate memory" "for IPv6 headers\n",__func__)); m_freem(m); return ENOBUFS; } /* * Copy all the protocol headers after * the IPv6 header. */ m_copydata(m, sizeof(struct ip6_hdr), skip - sizeof(struct ip6_hdr), ptr); alloc = 1; } else { /* No need to allocate memory. */ ptr = mtod(m, unsigned char *) + sizeof(struct ip6_hdr); alloc = 0; } } else break; off = ip6.ip6_nxt & 0xff; /* Next header type. */ for (len = 0; len < skip - sizeof(struct ip6_hdr);) switch (off) { case IPPROTO_HOPOPTS: case IPPROTO_DSTOPTS: ip6e = (struct ip6_ext *) (ptr + len); /* * Process the mutable/immutable * options -- borrows heavily from the * KAME code. */ for (count = len + sizeof(struct ip6_ext); count < len + ((ip6e->ip6e_len + 1) << 3);) { if (ptr[count] == IP6OPT_PAD1) { count++; continue; /* Skip padding. */ } /* Sanity check. */ if (count > len + ((ip6e->ip6e_len + 1) << 3)) { m_freem(m); /* Free, if we allocated. */ if (alloc) free(ptr, M_XDATA); return EINVAL; } ad = ptr[count + 1]; /* If mutable option, zeroize. */ if (ptr[count] & IP6OPT_MUTABLE) bcopy(ipseczeroes, ptr + count, ptr[count + 1]); count += ad; /* Sanity check. */ if (count > skip - sizeof(struct ip6_hdr)) { m_freem(m); /* Free, if we allocated. */ if (alloc) free(ptr, M_XDATA); return EINVAL; } } /* Advance. */ len += ((ip6e->ip6e_len + 1) << 3); off = ip6e->ip6e_nxt; break; case IPPROTO_ROUTING: /* * Always include routing headers in * computation. */ ip6e = (struct ip6_ext *) (ptr + len); len += ((ip6e->ip6e_len + 1) << 3); off = ip6e->ip6e_nxt; break; default: DPRINTF(("%s: unexpected IPv6 header type %d", __func__, off)); if (alloc) free(ptr, M_XDATA); m_freem(m); return EINVAL; } /* Copyback and free, if we allocated. */ if (alloc) { m_copyback(m, sizeof(struct ip6_hdr), skip - sizeof(struct ip6_hdr), ptr); free(ptr, M_XDATA); } break; #endif /* INET6 */ } return 0; } /* * ah_input() gets called to verify that an input packet * passes authentication. */ static int ah_input(struct mbuf *m, struct secasvar *sav, int skip, int protoff) { char buf[128]; struct cryptodesc *crda; struct cryptop *crp; struct auth_hash *ahx; struct xform_data *xd; struct newah *ah; uint64_t cryptoid; int hl, rplen, authsize, error; IPSEC_ASSERT(sav != NULL, ("null SA")); IPSEC_ASSERT(sav->key_auth != NULL, ("null authentication key")); IPSEC_ASSERT(sav->tdb_authalgxform != NULL, ("null authentication xform")); /* Figure out header size. */ rplen = HDRSIZE(sav); /* XXX don't pullup, just copy header */ IP6_EXTHDR_GET(ah, struct newah *, m, skip, rplen); if (ah == NULL) { DPRINTF(("ah_input: cannot pullup header\n")); AHSTAT_INC(ahs_hdrops); /*XXX*/ m_freem(m); return ENOBUFS; } /* Check replay window, if applicable. */ SECASVAR_LOCK(sav); if (sav->replay != NULL && sav->replay->wsize != 0 && ipsec_chkreplay(ntohl(ah->ah_seq), sav) == 0) { SECASVAR_UNLOCK(sav); AHSTAT_INC(ahs_replay); DPRINTF(("%s: packet replay failure: %s\n", __func__, ipsec_logsastr(sav, buf, sizeof(buf)))); m_freem(m); return (EACCES); } cryptoid = sav->tdb_cryptoid; SECASVAR_UNLOCK(sav); /* Verify AH header length. */ hl = ah->ah_len * sizeof (u_int32_t); ahx = sav->tdb_authalgxform; authsize = AUTHSIZE(sav); if (hl != authsize + rplen - sizeof (struct ah)) { DPRINTF(("%s: bad authenticator length %u (expecting %lu)" " for packet in SA %s/%08lx\n", __func__, hl, (u_long) (authsize + rplen - sizeof (struct ah)), ipsec_address(&sav->sah->saidx.dst, buf, sizeof(buf)), (u_long) ntohl(sav->spi))); AHSTAT_INC(ahs_badauthl); m_freem(m); return EACCES; } AHSTAT_ADD(ahs_ibytes, m->m_pkthdr.len - skip - hl); /* Get crypto descriptors. */ crp = crypto_getreq(1); if (crp == NULL) { DPRINTF(("%s: failed to acquire crypto descriptor\n", __func__)); AHSTAT_INC(ahs_crypto); m_freem(m); return ENOBUFS; } crda = crp->crp_desc; IPSEC_ASSERT(crda != NULL, ("null crypto descriptor")); crda->crd_skip = 0; crda->crd_len = m->m_pkthdr.len; crda->crd_inject = skip + rplen; /* Authentication operation. */ crda->crd_alg = ahx->type; crda->crd_klen = _KEYBITS(sav->key_auth); crda->crd_key = sav->key_auth->key_data; /* Allocate IPsec-specific opaque crypto info. */ xd = malloc(sizeof(*xd) + skip + rplen + authsize, M_XDATA, M_NOWAIT | M_ZERO); if (xd == NULL) { DPRINTF(("%s: failed to allocate xform_data\n", __func__)); AHSTAT_INC(ahs_crypto); crypto_freereq(crp); m_freem(m); return ENOBUFS; } /* * Save the authenticator, the skipped portion of the packet, * and the AH header. */ m_copydata(m, 0, skip + rplen + authsize, (caddr_t)(xd + 1)); /* Zeroize the authenticator on the packet. */ m_copyback(m, skip + rplen, authsize, ipseczeroes); /* "Massage" the packet headers for crypto processing. */ error = ah_massage_headers(&m, sav->sah->saidx.dst.sa.sa_family, skip, ahx->type, 0); if (error != 0) { /* NB: mbuf is free'd by ah_massage_headers */ AHSTAT_INC(ahs_hdrops); free(xd, M_XDATA); crypto_freereq(crp); return (error); } /* Crypto operation descriptor. */ crp->crp_ilen = m->m_pkthdr.len; /* Total input length. */ crp->crp_flags = CRYPTO_F_IMBUF | CRYPTO_F_CBIFSYNC; crp->crp_buf = (caddr_t) m; crp->crp_callback = ah_input_cb; crp->crp_sid = cryptoid; crp->crp_opaque = (caddr_t) xd; /* These are passed as-is to the callback. */ xd->sav = sav; xd->nxt = ah->ah_nxt; xd->protoff = protoff; xd->skip = skip; xd->cryptoid = cryptoid; return (crypto_dispatch(crp)); } /* * AH input callback from the crypto driver. */ static int ah_input_cb(struct cryptop *crp) { char buf[IPSEC_ADDRSTRLEN]; unsigned char calc[AH_ALEN_MAX]; struct mbuf *m; struct cryptodesc *crd; struct auth_hash *ahx; struct xform_data *xd; struct secasvar *sav; struct secasindex *saidx; caddr_t ptr; uint64_t cryptoid; int authsize, rplen, error, skip, protoff; uint8_t nxt; crd = crp->crp_desc; m = (struct mbuf *) crp->crp_buf; xd = (struct xform_data *) crp->crp_opaque; sav = xd->sav; skip = xd->skip; nxt = xd->nxt; protoff = xd->protoff; cryptoid = xd->cryptoid; saidx = &sav->sah->saidx; IPSEC_ASSERT(saidx->dst.sa.sa_family == AF_INET || saidx->dst.sa.sa_family == AF_INET6, ("unexpected protocol family %u", saidx->dst.sa.sa_family)); ahx = (struct auth_hash *) sav->tdb_authalgxform; /* Check for crypto errors. */ if (crp->crp_etype) { if (crp->crp_etype == EAGAIN) { /* Reset the session ID */ if (ipsec_updateid(sav, &crp->crp_sid, &cryptoid) != 0) crypto_freesession(cryptoid); xd->cryptoid = crp->crp_sid; return (crypto_dispatch(crp)); } AHSTAT_INC(ahs_noxform); DPRINTF(("%s: crypto error %d\n", __func__, crp->crp_etype)); error = crp->crp_etype; goto bad; } else { AHSTAT_INC(ahs_hist[sav->alg_auth]); crypto_freereq(crp); /* No longer needed. */ crp = NULL; } /* Shouldn't happen... */ if (m == NULL) { AHSTAT_INC(ahs_crypto); DPRINTF(("%s: bogus returned buffer from crypto\n", __func__)); error = EINVAL; goto bad; } /* Figure out header size. */ rplen = HDRSIZE(sav); authsize = AUTHSIZE(sav); /* Copy authenticator off the packet. */ m_copydata(m, skip + rplen, authsize, calc); /* Verify authenticator. */ ptr = (caddr_t) (xd + 1); if (timingsafe_bcmp(ptr + skip + rplen, calc, authsize)) { DPRINTF(("%s: authentication hash mismatch for packet " "in SA %s/%08lx\n", __func__, ipsec_address(&saidx->dst, buf, sizeof(buf)), (u_long) ntohl(sav->spi))); AHSTAT_INC(ahs_badauth); error = EACCES; goto bad; } /* Fix the Next Protocol field. */ ((uint8_t *) ptr)[protoff] = nxt; /* Copyback the saved (uncooked) network headers. */ m_copyback(m, 0, skip, ptr); free(xd, M_XDATA), xd = NULL; /* No longer needed */ /* * Header is now authenticated. */ m->m_flags |= M_AUTHIPHDR|M_AUTHIPDGM; /* * Update replay sequence number, if appropriate. */ if (sav->replay) { u_int32_t seq; m_copydata(m, skip + offsetof(struct newah, ah_seq), sizeof (seq), (caddr_t) &seq); SECASVAR_LOCK(sav); if (ipsec_updatereplay(ntohl(seq), sav)) { SECASVAR_UNLOCK(sav); AHSTAT_INC(ahs_replay); error = EACCES; goto bad; } SECASVAR_UNLOCK(sav); } /* * Remove the AH header and authenticator from the mbuf. */ error = m_striphdr(m, skip, rplen + authsize); if (error) { DPRINTF(("%s: mangled mbuf chain for SA %s/%08lx\n", __func__, ipsec_address(&saidx->dst, buf, sizeof(buf)), (u_long) ntohl(sav->spi))); AHSTAT_INC(ahs_hdrops); goto bad; } switch (saidx->dst.sa.sa_family) { #ifdef INET6 case AF_INET6: error = ipsec6_common_input_cb(m, sav, skip, protoff); break; #endif #ifdef INET case AF_INET: error = ipsec4_common_input_cb(m, sav, skip, protoff); break; #endif default: panic("%s: Unexpected address family: %d saidx=%p", __func__, saidx->dst.sa.sa_family, saidx); } return error; bad: if (sav) key_freesav(&sav); if (m != NULL) m_freem(m); if (xd != NULL) free(xd, M_XDATA); if (crp != NULL) crypto_freereq(crp); return error; } /* * AH output routine, called by ipsec[46]_perform_request(). */ static int ah_output(struct mbuf *m, struct secpolicy *sp, struct secasvar *sav, u_int idx, int skip, int protoff) { char buf[IPSEC_ADDRSTRLEN]; struct auth_hash *ahx; struct cryptodesc *crda; struct xform_data *xd; struct mbuf *mi; struct cryptop *crp; struct newah *ah; uint64_t cryptoid; uint16_t iplen; int error, rplen, authsize, maxpacketsize, roff; uint8_t prot; IPSEC_ASSERT(sav != NULL, ("null SA")); ahx = sav->tdb_authalgxform; IPSEC_ASSERT(ahx != NULL, ("null authentication xform")); AHSTAT_INC(ahs_output); /* Figure out header size. */ rplen = HDRSIZE(sav); /* Check for maximum packet size violations. */ switch (sav->sah->saidx.dst.sa.sa_family) { #ifdef INET case AF_INET: maxpacketsize = IP_MAXPACKET; break; #endif /* INET */ #ifdef INET6 case AF_INET6: maxpacketsize = IPV6_MAXPACKET; break; #endif /* INET6 */ default: DPRINTF(("%s: unknown/unsupported protocol family %u, " "SA %s/%08lx\n", __func__, sav->sah->saidx.dst.sa.sa_family, ipsec_address(&sav->sah->saidx.dst, buf, sizeof(buf)), (u_long) ntohl(sav->spi))); AHSTAT_INC(ahs_nopf); error = EPFNOSUPPORT; goto bad; } authsize = AUTHSIZE(sav); if (rplen + authsize + m->m_pkthdr.len > maxpacketsize) { DPRINTF(("%s: packet in SA %s/%08lx got too big " "(len %u, max len %u)\n", __func__, ipsec_address(&sav->sah->saidx.dst, buf, sizeof(buf)), (u_long) ntohl(sav->spi), rplen + authsize + m->m_pkthdr.len, maxpacketsize)); AHSTAT_INC(ahs_toobig); error = EMSGSIZE; goto bad; } /* Update the counters. */ AHSTAT_ADD(ahs_obytes, m->m_pkthdr.len - skip); m = m_unshare(m, M_NOWAIT); if (m == NULL) { DPRINTF(("%s: cannot clone mbuf chain, SA %s/%08lx\n", __func__, ipsec_address(&sav->sah->saidx.dst, buf, sizeof(buf)), (u_long) ntohl(sav->spi))); AHSTAT_INC(ahs_hdrops); error = ENOBUFS; goto bad; } /* Inject AH header. */ mi = m_makespace(m, skip, rplen + authsize, &roff); if (mi == NULL) { DPRINTF(("%s: failed to inject %u byte AH header for SA " "%s/%08lx\n", __func__, rplen + authsize, ipsec_address(&sav->sah->saidx.dst, buf, sizeof(buf)), (u_long) ntohl(sav->spi))); AHSTAT_INC(ahs_hdrops); /*XXX differs from openbsd */ error = ENOBUFS; goto bad; } /* * The AH header is guaranteed by m_makespace() to be in * contiguous memory, at roff bytes offset into the returned mbuf. */ ah = (struct newah *)(mtod(mi, caddr_t) + roff); /* Initialize the AH header. */ m_copydata(m, protoff, sizeof(u_int8_t), (caddr_t) &ah->ah_nxt); ah->ah_len = (rplen + authsize - sizeof(struct ah)) / sizeof(u_int32_t); ah->ah_reserve = 0; ah->ah_spi = sav->spi; /* Zeroize authenticator. */ m_copyback(m, skip + rplen, authsize, ipseczeroes); /* Insert packet replay counter, as requested. */ SECASVAR_LOCK(sav); if (sav->replay) { if (sav->replay->count == ~0 && (sav->flags & SADB_X_EXT_CYCSEQ) == 0) { SECASVAR_UNLOCK(sav); DPRINTF(("%s: replay counter wrapped for SA %s/%08lx\n", __func__, ipsec_address(&sav->sah->saidx.dst, buf, sizeof(buf)), (u_long) ntohl(sav->spi))); AHSTAT_INC(ahs_wrap); error = EACCES; goto bad; } #ifdef REGRESSION /* Emulate replay attack when ipsec_replay is TRUE. */ if (!V_ipsec_replay) #endif sav->replay->count++; ah->ah_seq = htonl(sav->replay->count); } cryptoid = sav->tdb_cryptoid; SECASVAR_UNLOCK(sav); /* Get crypto descriptors. */ crp = crypto_getreq(1); if (crp == NULL) { DPRINTF(("%s: failed to acquire crypto descriptors\n", __func__)); AHSTAT_INC(ahs_crypto); error = ENOBUFS; goto bad; } crda = crp->crp_desc; crda->crd_skip = 0; crda->crd_inject = skip + rplen; crda->crd_len = m->m_pkthdr.len; /* Authentication operation. */ crda->crd_alg = ahx->type; crda->crd_key = sav->key_auth->key_data; crda->crd_klen = _KEYBITS(sav->key_auth); /* Allocate IPsec-specific opaque crypto info. */ xd = malloc(sizeof(struct xform_data) + skip, M_XDATA, M_NOWAIT | M_ZERO); if (xd == NULL) { crypto_freereq(crp); DPRINTF(("%s: failed to allocate xform_data\n", __func__)); AHSTAT_INC(ahs_crypto); error = ENOBUFS; goto bad; } /* Save the skipped portion of the packet. */ m_copydata(m, 0, skip, (caddr_t) (xd + 1)); /* * Fix IP header length on the header used for * authentication. We don't need to fix the original * header length as it will be fixed by our caller. */ switch (sav->sah->saidx.dst.sa.sa_family) { #ifdef INET case AF_INET: bcopy(((caddr_t)(xd + 1)) + offsetof(struct ip, ip_len), (caddr_t) &iplen, sizeof(u_int16_t)); iplen = htons(ntohs(iplen) + rplen + authsize); m_copyback(m, offsetof(struct ip, ip_len), sizeof(u_int16_t), (caddr_t) &iplen); break; #endif /* INET */ #ifdef INET6 case AF_INET6: bcopy(((caddr_t)(xd + 1)) + offsetof(struct ip6_hdr, ip6_plen), (caddr_t) &iplen, sizeof(uint16_t)); iplen = htons(ntohs(iplen) + rplen + authsize); m_copyback(m, offsetof(struct ip6_hdr, ip6_plen), sizeof(uint16_t), (caddr_t) &iplen); break; #endif /* INET6 */ } /* Fix the Next Header field in saved header. */ ((uint8_t *) (xd + 1))[protoff] = IPPROTO_AH; /* Update the Next Protocol field in the IP header. */ prot = IPPROTO_AH; m_copyback(m, protoff, sizeof(uint8_t), (caddr_t) &prot); /* "Massage" the packet headers for crypto processing. */ error = ah_massage_headers(&m, sav->sah->saidx.dst.sa.sa_family, skip, ahx->type, 1); if (error != 0) { m = NULL; /* mbuf was free'd by ah_massage_headers. */ free(xd, M_XDATA); crypto_freereq(crp); goto bad; } /* Crypto operation descriptor. */ crp->crp_ilen = m->m_pkthdr.len; /* Total input length. */ crp->crp_flags = CRYPTO_F_IMBUF | CRYPTO_F_CBIFSYNC; crp->crp_buf = (caddr_t) m; crp->crp_callback = ah_output_cb; crp->crp_sid = cryptoid; crp->crp_opaque = (caddr_t) xd; /* These are passed as-is to the callback. */ xd->sp = sp; xd->sav = sav; xd->skip = skip; xd->idx = idx; xd->cryptoid = cryptoid; return crypto_dispatch(crp); bad: if (m) m_freem(m); return (error); } /* * AH output callback from the crypto driver. */ static int ah_output_cb(struct cryptop *crp) { struct xform_data *xd; struct secpolicy *sp; struct secasvar *sav; struct mbuf *m; uint64_t cryptoid; caddr_t ptr; u_int idx; int skip, error; m = (struct mbuf *) crp->crp_buf; xd = (struct xform_data *) crp->crp_opaque; sp = xd->sp; sav = xd->sav; skip = xd->skip; idx = xd->idx; cryptoid = xd->cryptoid; ptr = (caddr_t) (xd + 1); /* Check for crypto errors. */ if (crp->crp_etype) { if (crp->crp_etype == EAGAIN) { /* Reset the session ID */ if (ipsec_updateid(sav, &crp->crp_sid, &cryptoid) != 0) crypto_freesession(cryptoid); xd->cryptoid = crp->crp_sid; return (crypto_dispatch(crp)); } AHSTAT_INC(ahs_noxform); DPRINTF(("%s: crypto error %d\n", __func__, crp->crp_etype)); error = crp->crp_etype; m_freem(m); goto bad; } /* Shouldn't happen... */ if (m == NULL) { AHSTAT_INC(ahs_crypto); DPRINTF(("%s: bogus returned buffer from crypto\n", __func__)); error = EINVAL; goto bad; } /* * Copy original headers (with the new protocol number) back * in place. */ m_copyback(m, 0, skip, ptr); free(xd, M_XDATA); crypto_freereq(crp); AHSTAT_INC(ahs_hist[sav->alg_auth]); #ifdef REGRESSION /* Emulate man-in-the-middle attack when ipsec_integrity is TRUE. */ if (V_ipsec_integrity) { int alen; /* * Corrupt HMAC if we want to test integrity verification of * the other side. */ alen = AUTHSIZE(sav); m_copyback(m, m->m_pkthdr.len - alen, alen, ipseczeroes); } #endif /* NB: m is reclaimed by ipsec_process_done. */ error = ipsec_process_done(m, sp, sav, idx); return (error); bad: free(xd, M_XDATA); crypto_freereq(crp); key_freesav(&sav); key_freesp(&sp); return (error); } static struct xformsw ah_xformsw = { XF_AH, XFT_AUTH, "IPsec AH", ah_init, ah_zeroize, ah_input, ah_output, }; static void ah_attach(void) { xform_register(&ah_xformsw); } SYSINIT(ah_xform_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ah_attach, NULL); Index: projects/ipsec/sys/netipsec/xform_esp.c =================================================================== --- projects/ipsec/sys/netipsec/xform_esp.c (revision 309109) +++ projects/ipsec/sys/netipsec/xform_esp.c (revision 309110) @@ -1,990 +1,990 @@ /* $FreeBSD$ */ /* $OpenBSD: ip_esp.c,v 1.69 2001/06/26 06:18:59 angelos Exp $ */ /*- * The authors of this code are John Ioannidis (ji@tla.org), * Angelos D. Keromytis (kermit@csd.uch.gr) and * Niels Provos (provos@physnet.uni-hamburg.de). * * The original version of this code was written by John Ioannidis * for BSD/OS in Athens, Greece, in November 1995. * * Ported to OpenBSD and NetBSD, with additional transforms, in December 1996, * by Angelos D. Keromytis. * * Additional transforms and features in 1997 and 1998 by Angelos D. Keromytis * and Niels Provos. * * Additional features in 1999 by Angelos D. Keromytis. * * Copyright (C) 1995, 1996, 1997, 1998, 1999 by John Ioannidis, * Angelos D. Keromytis and Niels Provos. * Copyright (c) 2001 Angelos D. Keromytis. * * Permission to use, copy, and modify this software with or without fee * is hereby granted, provided that this entire notice is included in * all copies of any software which is or includes a copy or * modification of this software. * You may use this code under the GNU public license if you so wish. Please * contribute changes back to the authors under this freer than GPL license * so that we may further the use of strong encryption without limitations to * all. * * THIS SOFTWARE IS BEING PROVIDED "AS IS", WITHOUT ANY EXPRESS OR * IMPLIED WARRANTY. IN PARTICULAR, NONE OF THE AUTHORS MAKES ANY * REPRESENTATION OR WARRANTY OF ANY KIND CONCERNING THE * MERCHANTABILITY OF THIS SOFTWARE OR ITS FITNESS FOR ANY PARTICULAR * PURPOSE. */ #include "opt_inet.h" #include "opt_inet6.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 INET6 #include #include #include #endif #include #include #include #include VNET_DEFINE(int, esp_enable) = 1; VNET_PCPUSTAT_DEFINE(struct espstat, espstat); VNET_PCPUSTAT_SYSINIT(espstat); #ifdef VIMAGE VNET_PCPUSTAT_SYSUNINIT(espstat); #endif /* VIMAGE */ SYSCTL_DECL(_net_inet_esp); SYSCTL_INT(_net_inet_esp, OID_AUTO, esp_enable, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(esp_enable), 0, ""); SYSCTL_VNET_PCPUSTAT(_net_inet_esp, IPSECCTL_STATS, stats, struct espstat, espstat, "ESP statistics (struct espstat, netipsec/esp_var.h"); static int esp_input_cb(struct cryptop *op); static int esp_output_cb(struct cryptop *crp); /* * NB: this is public for use by the PF_KEY support. * NB: if you add support here; be sure to add code to esp_attach below! */ struct enc_xform * esp_algorithm_lookup(int alg) { if (alg >= ESP_ALG_MAX) return NULL; switch (alg) { case SADB_EALG_DESCBC: return &enc_xform_des; case SADB_EALG_3DESCBC: return &enc_xform_3des; case SADB_X_EALG_AES: return &enc_xform_rijndael128; case SADB_X_EALG_BLOWFISHCBC: return &enc_xform_blf; case SADB_X_EALG_CAST128CBC: return &enc_xform_cast5; case SADB_EALG_NULL: return &enc_xform_null; case SADB_X_EALG_CAMELLIACBC: return &enc_xform_camellia; case SADB_X_EALG_AESCTR: return &enc_xform_aes_icm; case SADB_X_EALG_AESGCM16: return &enc_xform_aes_nist_gcm; case SADB_X_EALG_AESGMAC: return &enc_xform_aes_nist_gmac; } return NULL; } size_t esp_hdrsiz(struct secasvar *sav) { size_t size; if (sav != NULL) { /*XXX not right for null algorithm--does it matter??*/ IPSEC_ASSERT(sav->tdb_encalgxform != NULL, ("SA with null xform")); if (sav->flags & SADB_X_EXT_OLD) size = sizeof (struct esp); else size = sizeof (struct newesp); size += sav->tdb_encalgxform->blocksize + 9; /*XXX need alg check???*/ if (sav->tdb_authalgxform != NULL && sav->replay) size += ah_hdrsiz(sav); } else { /* * base header size * + max iv length for CBC mode * + max pad length * + sizeof (pad length field) * + sizeof (next header field) * + max icv supported. */ size = sizeof (struct newesp) + EALG_MAX_BLOCK_LEN + 9 + 16; } return size; } /* * esp_init() is called when an SPI is being set up. */ static int esp_init(struct secasvar *sav, struct xformsw *xsp) { struct enc_xform *txform; struct cryptoini cria, crie; int keylen; int error; txform = esp_algorithm_lookup(sav->alg_enc); if (txform == NULL) { DPRINTF(("%s: unsupported encryption algorithm %d\n", __func__, sav->alg_enc)); return EINVAL; } if (sav->key_enc == NULL) { DPRINTF(("%s: no encoding key for %s algorithm\n", __func__, txform->name)); return EINVAL; } if ((sav->flags & (SADB_X_EXT_OLD | SADB_X_EXT_IV4B)) == SADB_X_EXT_IV4B) { DPRINTF(("%s: 4-byte IV not supported with protocol\n", __func__)); return EINVAL; } /* subtract off the salt, RFC4106, 8.1 and RFC3686, 5.1 */ keylen = _KEYLEN(sav->key_enc) - SAV_ISCTRORGCM(sav) * 4; if (txform->minkey > keylen || keylen > txform->maxkey) { DPRINTF(("%s: invalid key length %u, must be in the range " "[%u..%u] for algorithm %s\n", __func__, keylen, txform->minkey, txform->maxkey, txform->name)); return EINVAL; } if (SAV_ISCTRORGCM(sav)) sav->ivlen = 8; /* RFC4106 3.1 and RFC3686 3.1 */ else sav->ivlen = txform->ivsize; /* * Setup AH-related state. */ if (sav->alg_auth != 0) { error = ah_init0(sav, xsp, &cria); if (error) return error; } /* NB: override anything set in ah_init0 */ sav->tdb_xform = xsp; sav->tdb_encalgxform = txform; /* * Whenever AES-GCM is used for encryption, one * of the AES authentication algorithms is chosen * as well, based on the key size. */ if (sav->alg_enc == SADB_X_EALG_AESGCM16) { switch (keylen) { case AES_128_GMAC_KEY_LEN: sav->alg_auth = SADB_X_AALG_AES128GMAC; sav->tdb_authalgxform = &auth_hash_nist_gmac_aes_128; break; case AES_192_GMAC_KEY_LEN: sav->alg_auth = SADB_X_AALG_AES192GMAC; sav->tdb_authalgxform = &auth_hash_nist_gmac_aes_192; break; case AES_256_GMAC_KEY_LEN: sav->alg_auth = SADB_X_AALG_AES256GMAC; sav->tdb_authalgxform = &auth_hash_nist_gmac_aes_256; break; default: DPRINTF(("%s: invalid key length %u" "for algorithm %s\n", __func__, keylen, txform->name)); return EINVAL; } bzero(&cria, sizeof(cria)); cria.cri_alg = sav->tdb_authalgxform->type; cria.cri_key = sav->key_enc->key_data; cria.cri_klen = _KEYBITS(sav->key_enc) - SAV_ISGCM(sav) * 32; } /* Initialize crypto session. */ bzero(&crie, sizeof(crie)); crie.cri_alg = sav->tdb_encalgxform->type; crie.cri_key = sav->key_enc->key_data; crie.cri_klen = _KEYBITS(sav->key_enc) - SAV_ISCTRORGCM(sav) * 32; if (sav->tdb_authalgxform && sav->tdb_encalgxform) { /* init both auth & enc */ crie.cri_next = &cria; error = crypto_newsession(&sav->tdb_cryptoid, &crie, V_crypto_support); } else if (sav->tdb_encalgxform) { error = crypto_newsession(&sav->tdb_cryptoid, &crie, V_crypto_support); } else if (sav->tdb_authalgxform) { error = crypto_newsession(&sav->tdb_cryptoid, &cria, V_crypto_support); } else { /* XXX cannot happen? */ DPRINTF(("%s: no encoding OR authentication xform!\n", __func__)); error = EINVAL; } return error; } /* * Paranoia. */ static int esp_zeroize(struct secasvar *sav) { /* NB: ah_zerorize free's the crypto session state */ int error = ah_zeroize(sav); if (sav->key_enc) bzero(sav->key_enc->key_data, _KEYLEN(sav->key_enc)); sav->tdb_encalgxform = NULL; sav->tdb_xform = NULL; return error; } /* * ESP input processing, called (eventually) through the protocol switch. */ static int esp_input(struct mbuf *m, struct secasvar *sav, int skip, int protoff) { char buf[128]; struct auth_hash *esph; struct enc_xform *espx; struct xform_data *xd; struct cryptodesc *crde; struct cryptop *crp; struct newesp *esp; uint8_t *ivp; uint64_t cryptoid; int plen, alen, hlen; IPSEC_ASSERT(sav != NULL, ("null SA")); IPSEC_ASSERT(sav->tdb_encalgxform != NULL, ("null encoding xform")); /* Valid IP Packet length ? */ if ( (skip&3) || (m->m_pkthdr.len&3) ){ DPRINTF(("%s: misaligned packet, skip %u pkt len %u", __func__, skip, m->m_pkthdr.len)); ESPSTAT_INC(esps_badilen); m_freem(m); return EINVAL; } /* XXX don't pullup, just copy header */ IP6_EXTHDR_GET(esp, struct newesp *, m, skip, sizeof (struct newesp)); esph = sav->tdb_authalgxform; espx = sav->tdb_encalgxform; /* Determine the ESP header and auth length */ if (sav->flags & SADB_X_EXT_OLD) hlen = sizeof (struct esp) + sav->ivlen; else hlen = sizeof (struct newesp) + sav->ivlen; alen = xform_ah_authsize(esph); /* * Verify payload length is multiple of encryption algorithm * block size. * * NB: This works for the null algorithm because the blocksize * is 4 and all packets must be 4-byte aligned regardless * of the algorithm. */ plen = m->m_pkthdr.len - (skip + hlen + alen); if ((plen & (espx->blocksize - 1)) || (plen <= 0)) { DPRINTF(("%s: payload of %d octets not a multiple of %d octets," " SA %s/%08lx\n", __func__, plen, espx->blocksize, ipsec_address(&sav->sah->saidx.dst, buf, sizeof(buf)), (u_long)ntohl(sav->spi))); ESPSTAT_INC(esps_badilen); m_freem(m); return EINVAL; } /* * Check sequence number. */ SECASVAR_LOCK(sav); if (esph != NULL && sav->replay != NULL && sav->replay->wsize != 0) { if (ipsec_chkreplay(ntohl(esp->esp_seq), sav) == 0) { SECASVAR_UNLOCK(sav); DPRINTF(("%s: packet replay check for %s\n", __func__, ipsec_logsastr(sav, buf, sizeof(buf)))); ESPSTAT_INC(esps_replay); m_freem(m); return (EACCES); } } cryptoid = sav->tdb_cryptoid; SECASVAR_UNLOCK(sav); /* Update the counters */ ESPSTAT_ADD(esps_ibytes, m->m_pkthdr.len - (skip + hlen + alen)); /* Get crypto descriptors */ crp = crypto_getreq(esph && espx ? 2 : 1); if (crp == NULL) { DPRINTF(("%s: failed to acquire crypto descriptors\n", __func__)); ESPSTAT_INC(esps_crypto); m_freem(m); return ENOBUFS; } /* Get IPsec-specific opaque pointer */ xd = malloc(sizeof(*xd) + alen, M_XDATA, M_NOWAIT | M_ZERO); if (xd == NULL) { DPRINTF(("%s: failed to allocate xform_data\n", __func__)); ESPSTAT_INC(esps_crypto); crypto_freereq(crp); m_freem(m); return ENOBUFS; } if (esph != NULL) { struct cryptodesc *crda = crp->crp_desc; IPSEC_ASSERT(crda != NULL, ("null ah crypto descriptor")); /* Authentication descriptor */ crda->crd_skip = skip; if (SAV_ISGCM(sav)) crda->crd_len = 8; /* RFC4106 5, SPI + SN */ else crda->crd_len = m->m_pkthdr.len - (skip + alen); crda->crd_inject = m->m_pkthdr.len - alen; crda->crd_alg = esph->type; /* Copy the authenticator */ m_copydata(m, m->m_pkthdr.len - alen, alen, (caddr_t) (xd + 1)); /* Chain authentication request */ crde = crda->crd_next; } else { crde = crp->crp_desc; } /* Crypto operation descriptor */ crp->crp_ilen = m->m_pkthdr.len; /* Total input length */ crp->crp_flags = CRYPTO_F_IMBUF | CRYPTO_F_CBIFSYNC; crp->crp_buf = (caddr_t) m; crp->crp_callback = esp_input_cb; crp->crp_sid = cryptoid; crp->crp_opaque = (caddr_t) xd; /* These are passed as-is to the callback */ xd->sav = sav; xd->protoff = protoff; xd->skip = skip; xd->cryptoid = cryptoid; /* Decryption descriptor */ IPSEC_ASSERT(crde != NULL, ("null esp crypto descriptor")); crde->crd_skip = skip + hlen; crde->crd_len = m->m_pkthdr.len - (skip + hlen + alen); crde->crd_inject = skip + hlen - sav->ivlen; if (SAV_ISCTRORGCM(sav)) { ivp = &crde->crd_iv[0]; /* GCM IV Format: RFC4106 4 */ /* CTR IV Format: RFC3686 4 */ /* Salt is last four bytes of key, RFC4106 8.1 */ /* Nonce is last four bytes of key, RFC3686 5.1 */ memcpy(ivp, sav->key_enc->key_data + _KEYLEN(sav->key_enc) - 4, 4); if (SAV_ISCTR(sav)) { /* Initial block counter is 1, RFC3686 4 */ be32enc(&ivp[sav->ivlen + 4], 1); } m_copydata(m, skip + hlen - sav->ivlen, sav->ivlen, &ivp[4]); crde->crd_flags |= CRD_F_IV_EXPLICIT; } crde->crd_alg = espx->type; return (crypto_dispatch(crp)); } /* * ESP input callback from the crypto driver. */ static int esp_input_cb(struct cryptop *crp) { char buf[128]; u_int8_t lastthree[3], aalg[AH_HMAC_MAXHASHLEN]; int hlen, skip, protoff, error, alen; struct mbuf *m; struct cryptodesc *crd; struct auth_hash *esph; struct enc_xform *espx; struct xform_data *xd; struct secasvar *sav; struct secasindex *saidx; caddr_t ptr; uint64_t cryptoid; crd = crp->crp_desc; IPSEC_ASSERT(crd != NULL, ("null crypto descriptor!")); m = (struct mbuf *) crp->crp_buf; xd = (struct xform_data *) crp->crp_opaque; sav = xd->sav; skip = xd->skip; protoff = xd->protoff; cryptoid = xd->cryptoid; saidx = &sav->sah->saidx; esph = sav->tdb_authalgxform; espx = sav->tdb_encalgxform; /* Check for crypto errors */ if (crp->crp_etype) { if (crp->crp_etype == EAGAIN) { /* Reset the session ID */ if (ipsec_updateid(sav, &crp->crp_sid, &cryptoid) != 0) crypto_freesession(cryptoid); xd->cryptoid = crp->crp_sid; return (crypto_dispatch(crp)); } ESPSTAT_INC(esps_noxform); DPRINTF(("%s: crypto error %d\n", __func__, crp->crp_etype)); error = crp->crp_etype; goto bad; } /* Shouldn't happen... */ if (m == NULL) { ESPSTAT_INC(esps_crypto); DPRINTF(("%s: bogus returned buffer from crypto\n", __func__)); error = EINVAL; goto bad; } ESPSTAT_INC(esps_hist[sav->alg_enc]); /* If authentication was performed, check now. */ if (esph != NULL) { alen = xform_ah_authsize(esph); AHSTAT_INC(ahs_hist[sav->alg_auth]); /* Copy the authenticator from the packet */ m_copydata(m, m->m_pkthdr.len - alen, alen, aalg); ptr = (caddr_t) (xd + 1); /* Verify authenticator */ if (timingsafe_bcmp(ptr, aalg, alen) != 0) { DPRINTF(("%s: authentication hash mismatch for " "packet in SA %s/%08lx\n", __func__, ipsec_address(&saidx->dst, buf, sizeof(buf)), (u_long) ntohl(sav->spi))); ESPSTAT_INC(esps_badauth); error = EACCES; goto bad; } m->m_flags |= M_AUTHIPDGM; /* Remove trailing authenticator */ m_adj(m, -alen); } /* Release the crypto descriptors */ free(xd, M_XDATA), xd = NULL; crypto_freereq(crp), crp = NULL; /* * Packet is now decrypted. */ m->m_flags |= M_DECRYPTED; /* * Update replay sequence number, if appropriate. */ if (sav->replay) { u_int32_t seq; m_copydata(m, skip + offsetof(struct newesp, esp_seq), sizeof (seq), (caddr_t) &seq); SECASVAR_LOCK(sav); if (ipsec_updatereplay(ntohl(seq), sav)) { SECASVAR_UNLOCK(sav); DPRINTF(("%s: packet replay check for %s\n", __func__, ipsec_logsastr(sav, buf, sizeof(buf)))); ESPSTAT_INC(esps_replay); error = EACCES; goto bad; } SECASVAR_UNLOCK(sav); } /* Determine the ESP header length */ if (sav->flags & SADB_X_EXT_OLD) hlen = sizeof (struct esp) + sav->ivlen; else hlen = sizeof (struct newesp) + sav->ivlen; /* Remove the ESP header and IV from the mbuf. */ error = m_striphdr(m, skip, hlen); if (error) { ESPSTAT_INC(esps_hdrops); DPRINTF(("%s: bad mbuf chain, SA %s/%08lx\n", __func__, ipsec_address(&sav->sah->saidx.dst, buf, sizeof(buf)), (u_long) ntohl(sav->spi))); goto bad; } /* Save the last three bytes of decrypted data */ m_copydata(m, m->m_pkthdr.len - 3, 3, lastthree); /* Verify pad length */ if (lastthree[1] + 2 > m->m_pkthdr.len - skip) { ESPSTAT_INC(esps_badilen); DPRINTF(("%s: invalid padding length %d for %u byte packet " "in SA %s/%08lx\n", __func__, lastthree[1], m->m_pkthdr.len - skip, ipsec_address(&sav->sah->saidx.dst, buf, sizeof(buf)), (u_long) ntohl(sav->spi))); error = EINVAL; goto bad; } /* Verify correct decryption by checking the last padding bytes */ if ((sav->flags & SADB_X_EXT_PMASK) != SADB_X_EXT_PRAND) { if (lastthree[1] != lastthree[0] && lastthree[1] != 0) { ESPSTAT_INC(esps_badenc); DPRINTF(("%s: decryption failed for packet in " "SA %s/%08lx\n", __func__, ipsec_address( &sav->sah->saidx.dst, buf, sizeof(buf)), (u_long) ntohl(sav->spi))); error = EINVAL; goto bad; } } /* Trim the mbuf chain to remove trailing authenticator and padding */ m_adj(m, -(lastthree[1] + 2)); /* Restore the Next Protocol field */ m_copyback(m, protoff, sizeof (u_int8_t), lastthree + 2); switch (saidx->dst.sa.sa_family) { #ifdef INET6 case AF_INET6: error = ipsec6_common_input_cb(m, sav, skip, protoff); break; #endif #ifdef INET case AF_INET: error = ipsec4_common_input_cb(m, sav, skip, protoff); break; #endif default: panic("%s: Unexpected address family: %d saidx=%p", __func__, saidx->dst.sa.sa_family, saidx); } return error; bad: if (sav != NULL) key_freesav(&sav); if (m != NULL) m_freem(m); if (xd != NULL) free(xd, M_XDATA); if (crp != NULL) crypto_freereq(crp); return error; } /* * ESP output routine, called by ipsec[46]_perform_request(). */ static int esp_output(struct mbuf *m, struct secpolicy *sp, struct secasvar *sav, u_int idx, int skip, int protoff) { char buf[IPSEC_ADDRSTRLEN]; struct cryptodesc *crde = NULL, *crda = NULL; struct cryptop *crp; struct enc_xform *espx; struct auth_hash *esph; struct mbuf *mo = NULL; struct xform_data *xd; struct secasindex *saidx; unsigned char *pad; uint8_t *ivp; uint64_t cntr, cryptoid; int hlen, rlen, padding, blks, alen, i, roff; int error, maxpacketsize; uint8_t prot; IPSEC_ASSERT(sav != NULL, ("null SA")); esph = sav->tdb_authalgxform; espx = sav->tdb_encalgxform; IPSEC_ASSERT(espx != NULL, ("null encoding xform")); if (sav->flags & SADB_X_EXT_OLD) hlen = sizeof (struct esp) + sav->ivlen; else hlen = sizeof (struct newesp) + sav->ivlen; rlen = m->m_pkthdr.len - skip; /* Raw payload length. */ /* * RFC4303 2.4 Requires 4 byte alignment. */ blks = MAX(4, espx->blocksize); /* Cipher blocksize */ /* XXX clamp padding length a la KAME??? */ padding = ((blks - ((rlen + 2) % blks)) % blks) + 2; alen = xform_ah_authsize(esph); ESPSTAT_INC(esps_output); saidx = &sav->sah->saidx; /* Check for maximum packet size violations. */ switch (saidx->dst.sa.sa_family) { #ifdef INET case AF_INET: maxpacketsize = IP_MAXPACKET; break; #endif /* INET */ #ifdef INET6 case AF_INET6: maxpacketsize = IPV6_MAXPACKET; break; #endif /* INET6 */ default: DPRINTF(("%s: unknown/unsupported protocol " "family %d, SA %s/%08lx\n", __func__, saidx->dst.sa.sa_family, ipsec_address(&saidx->dst, buf, sizeof(buf)), (u_long) ntohl(sav->spi))); ESPSTAT_INC(esps_nopf); error = EPFNOSUPPORT; goto bad; } /* DPRINTF(("%s: skip %d hlen %d rlen %d padding %d alen %d blksd %d\n", __func__, skip, hlen, rlen, padding, alen, blks)); */ if (skip + hlen + rlen + padding + alen > maxpacketsize) { DPRINTF(("%s: packet in SA %s/%08lx got too big " "(len %u, max len %u)\n", __func__, ipsec_address(&saidx->dst, buf, sizeof(buf)), (u_long) ntohl(sav->spi), skip + hlen + rlen + padding + alen, maxpacketsize)); ESPSTAT_INC(esps_toobig); error = EMSGSIZE; goto bad; } /* Update the counters. */ ESPSTAT_ADD(esps_obytes, m->m_pkthdr.len - skip); m = m_unshare(m, M_NOWAIT); if (m == NULL) { DPRINTF(("%s: cannot clone mbuf chain, SA %s/%08lx\n", __func__, ipsec_address(&saidx->dst, buf, sizeof(buf)), (u_long) ntohl(sav->spi))); ESPSTAT_INC(esps_hdrops); error = ENOBUFS; goto bad; } /* Inject ESP header. */ mo = m_makespace(m, skip, hlen, &roff); if (mo == NULL) { DPRINTF(("%s: %u byte ESP hdr inject failed for SA %s/%08lx\n", __func__, hlen, ipsec_address(&saidx->dst, buf, sizeof(buf)), (u_long) ntohl(sav->spi))); ESPSTAT_INC(esps_hdrops); /* XXX diffs from openbsd */ error = ENOBUFS; goto bad; } /* Initialize ESP header. */ bcopy((caddr_t) &sav->spi, mtod(mo, caddr_t) + roff, sizeof(uint32_t)); SECASVAR_LOCK(sav); if (sav->replay) { uint32_t replay; #ifdef REGRESSION /* Emulate replay attack when ipsec_replay is TRUE. */ if (!V_ipsec_replay) #endif sav->replay->count++; replay = htonl(sav->replay->count); bcopy((caddr_t) &replay, mtod(mo, caddr_t) + roff + sizeof(uint32_t), sizeof(uint32_t)); } cryptoid = sav->tdb_cryptoid; if (SAV_ISCTRORGCM(sav)) cntr = sav->cntr++; SECASVAR_UNLOCK(sav); /* * Add padding -- better to do it ourselves than use the crypto engine, * although if/when we support compression, we'd have to do that. */ pad = (u_char *) m_pad(m, padding + alen); if (pad == NULL) { DPRINTF(("%s: m_pad failed for SA %s/%08lx\n", __func__, ipsec_address(&saidx->dst, buf, sizeof(buf)), (u_long) ntohl(sav->spi))); m = NULL; /* NB: free'd by m_pad */ error = ENOBUFS; goto bad; } /* * Add padding: random, zero, or self-describing. * XXX catch unexpected setting */ switch (sav->flags & SADB_X_EXT_PMASK) { case SADB_X_EXT_PRAND: (void) read_random(pad, padding - 2); break; case SADB_X_EXT_PZERO: bzero(pad, padding - 2); break; case SADB_X_EXT_PSEQ: for (i = 0; i < padding - 2; i++) pad[i] = i+1; break; } /* Fix padding length and Next Protocol in padding itself. */ pad[padding - 2] = padding - 2; m_copydata(m, protoff, sizeof(u_int8_t), pad + padding - 1); /* Fix Next Protocol in IPv4/IPv6 header. */ prot = IPPROTO_ESP; m_copyback(m, protoff, sizeof(u_int8_t), (u_char *) &prot); /* Get crypto descriptors. */ crp = crypto_getreq(esph != NULL ? 2 : 1); if (crp == NULL) { DPRINTF(("%s: failed to acquire crypto descriptors\n", __func__)); ESPSTAT_INC(esps_crypto); error = ENOBUFS; goto bad; } /* IPsec-specific opaque crypto info. */ xd = malloc(sizeof(struct xform_data), M_XDATA, M_NOWAIT | M_ZERO); if (xd == NULL) { crypto_freereq(crp); DPRINTF(("%s: failed to allocate xform_data\n", __func__)); ESPSTAT_INC(esps_crypto); error = ENOBUFS; goto bad; } crde = crp->crp_desc; crda = crde->crd_next; /* Encryption descriptor. */ crde->crd_skip = skip + hlen; crde->crd_len = m->m_pkthdr.len - (skip + hlen + alen); crde->crd_flags = CRD_F_ENCRYPT; crde->crd_inject = skip + hlen - sav->ivlen; /* Encryption operation. */ crde->crd_alg = espx->type; if (SAV_ISCTRORGCM(sav)) { ivp = &crde->crd_iv[0]; /* GCM IV Format: RFC4106 4 */ /* CTR IV Format: RFC3686 4 */ /* Salt is last four bytes of key, RFC4106 8.1 */ /* Nonce is last four bytes of key, RFC3686 5.1 */ memcpy(ivp, sav->key_enc->key_data + _KEYLEN(sav->key_enc) - 4, 4); be64enc(&ivp[4], cntr); if (SAV_ISCTR(sav)) { /* Initial block counter is 1, RFC3686 4 */ /* XXXAE: should we use this only for first packet? */ be32enc(&ivp[sav->ivlen + 4], 1); } m_copyback(m, skip + hlen - sav->ivlen, sav->ivlen, &ivp[4]); crde->crd_flags |= CRD_F_IV_EXPLICIT|CRD_F_IV_PRESENT; } /* Callback parameters */ xd->sp = sp; xd->sav = sav; xd->idx = idx; xd->cryptoid = cryptoid; /* Crypto operation descriptor. */ crp->crp_ilen = m->m_pkthdr.len; /* Total input length. */ crp->crp_flags = CRYPTO_F_IMBUF | CRYPTO_F_CBIFSYNC; crp->crp_buf = (caddr_t) m; crp->crp_callback = esp_output_cb; crp->crp_opaque = (caddr_t) xd; crp->crp_sid = cryptoid; if (esph) { /* Authentication descriptor. */ crda->crd_alg = esph->type; crda->crd_skip = skip; if (SAV_ISGCM(sav)) crda->crd_len = 8; /* RFC4106 5, SPI + SN */ else crda->crd_len = m->m_pkthdr.len - (skip + alen); crda->crd_inject = m->m_pkthdr.len - alen; } return crypto_dispatch(crp); bad: if (m) m_freem(m); return (error); } /* * ESP output callback from the crypto driver. */ static int esp_output_cb(struct cryptop *crp) { struct xform_data *xd; struct secpolicy *sp; struct secasvar *sav; struct mbuf *m; uint64_t cryptoid; u_int idx; int error; xd = (struct xform_data *) crp->crp_opaque; m = (struct mbuf *) crp->crp_buf; sp = xd->sp; sav = xd->sav; idx = xd->idx; cryptoid = xd->cryptoid; /* Check for crypto errors. */ if (crp->crp_etype) { if (crp->crp_etype == EAGAIN) { /* Reset the session ID */ if (ipsec_updateid(sav, &crp->crp_sid, &cryptoid) != 0) crypto_freesession(cryptoid); xd->cryptoid = crp->crp_sid; return (crypto_dispatch(crp)); } ESPSTAT_INC(esps_noxform); DPRINTF(("%s: crypto error %d\n", __func__, crp->crp_etype)); error = crp->crp_etype; m_freem(m); goto bad; } /* Shouldn't happen... */ if (m == NULL) { ESPSTAT_INC(esps_crypto); DPRINTF(("%s: bogus returned buffer from crypto\n", __func__)); error = EINVAL; goto bad; } free(xd, M_XDATA); crypto_freereq(crp); ESPSTAT_INC(esps_hist[sav->alg_enc]); if (sav->tdb_authalgxform != NULL) AHSTAT_INC(ahs_hist[sav->alg_auth]); #ifdef REGRESSION /* Emulate man-in-the-middle attack when ipsec_integrity is TRUE. */ if (V_ipsec_integrity) { static unsigned char ipseczeroes[AH_HMAC_MAXHASHLEN]; struct auth_hash *esph; /* * Corrupt HMAC if we want to test integrity verification of * the other side. */ esph = sav->tdb_authalgxform; if (esph != NULL) { int alen; alen = xform_ah_authsize(esph); m_copyback(m, m->m_pkthdr.len - alen, alen, ipseczeroes); } } #endif /* NB: m is reclaimed by ipsec_process_done. */ error = ipsec_process_done(m, sp, sav, idx); return (error); bad: free(xd, M_XDATA); crypto_freereq(crp); key_freesav(&sav); key_freesp(&sp); return (error); } static struct xformsw esp_xformsw = { XF_ESP, XFT_CONF|XFT_AUTH, "IPsec ESP", esp_init, esp_zeroize, esp_input, esp_output }; static void esp_attach(void) { xform_register(&esp_xformsw); } SYSINIT(esp_xform_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, esp_attach, NULL); Index: projects/ipsec/sys/netipsec/xform_ipcomp.c =================================================================== --- projects/ipsec/sys/netipsec/xform_ipcomp.c (revision 309109) +++ projects/ipsec/sys/netipsec/xform_ipcomp.c (revision 309110) @@ -1,766 +1,765 @@ /* $FreeBSD$ */ /* $OpenBSD: ip_ipcomp.c,v 1.1 2001/07/05 12:08:52 jjbg Exp $ */ /*- * Copyright (c) 2001 Jean-Jacques Bernard-Gundol (jj@wabbitt.org) * * 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 ``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 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 payload compression protocol (IPComp), see RFC 2393 */ #include "opt_inet.h" #include "opt_inet6.h" #include #include #include #include #include -#include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef INET6 #include #include #include #endif #include #include #include #include #include #include VNET_DEFINE(int, ipcomp_enable) = 1; VNET_PCPUSTAT_DEFINE(struct ipcompstat, ipcompstat); VNET_PCPUSTAT_SYSINIT(ipcompstat); #ifdef VIMAGE VNET_PCPUSTAT_SYSUNINIT(ipcompstat); #endif /* VIMAGE */ SYSCTL_DECL(_net_inet_ipcomp); SYSCTL_INT(_net_inet_ipcomp, OID_AUTO, ipcomp_enable, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipcomp_enable), 0, ""); SYSCTL_VNET_PCPUSTAT(_net_inet_ipcomp, IPSECCTL_STATS, stats, struct ipcompstat, ipcompstat, "IPCOMP statistics (struct ipcompstat, netipsec/ipcomp_var.h"); static int ipcomp_input_cb(struct cryptop *crp); static int ipcomp_output_cb(struct cryptop *crp); struct comp_algo * ipcomp_algorithm_lookup(int alg) { if (alg >= IPCOMP_ALG_MAX) return NULL; switch (alg) { case SADB_X_CALG_DEFLATE: return &comp_algo_deflate; } return NULL; } /* * RFC 3173 p 2.2. Non-Expansion Policy: * If the total size of a compressed payload and the IPComp header, as * defined in section 3, is not smaller than the size of the original * payload, the IP datagram MUST be sent in the original non-compressed * form. * * When we use IPComp in tunnel mode, for small packets we will receive * encapsulated IP-IP datagrams without any compression and without IPComp * header. */ static int ipcomp_encapcheck(union sockaddr_union *src, union sockaddr_union *dst) { struct secasvar *sav; sav = key_allocsa_tunnel(src, dst, IPPROTO_IPCOMP); if (sav == NULL) return (0); key_freesav(&sav); if (src->sa.sa_family == AF_INET) return (sizeof(struct in_addr) << 4); else return (sizeof(struct in6_addr) << 4); } static int ipcomp_nonexp_input(struct mbuf **mp, int *offp, int proto) { int isr; switch (proto) { #ifdef INET case IPPROTO_IPV4: isr = NETISR_IP; break; #endif #ifdef INET6 case IPPROTO_IPV6: isr = NETISR_IPV6; break; #endif default: IPCOMPSTAT_INC(ipcomps_nopf); m_freem(*mp); return (IPPROTO_DONE); } m_adj(*mp, *offp); IPCOMPSTAT_ADD(ipcomps_ibytes, (*mp)->m_pkthdr.len); IPCOMPSTAT_INC(ipcomps_input); netisr_dispatch(isr, *mp); return (IPPROTO_DONE); } /* * ipcomp_init() is called when an CPI is being set up. */ static int ipcomp_init(struct secasvar *sav, struct xformsw *xsp) { struct comp_algo *tcomp; struct cryptoini cric; /* NB: algorithm really comes in alg_enc and not alg_comp! */ tcomp = ipcomp_algorithm_lookup(sav->alg_enc); if (tcomp == NULL) { DPRINTF(("%s: unsupported compression algorithm %d\n", __func__, sav->alg_comp)); return EINVAL; } sav->alg_comp = sav->alg_enc; /* set for doing histogram */ sav->tdb_xform = xsp; sav->tdb_compalgxform = tcomp; /* Initialize crypto session */ bzero(&cric, sizeof (cric)); cric.cri_alg = sav->tdb_compalgxform->type; return crypto_newsession(&sav->tdb_cryptoid, &cric, V_crypto_support); } /* * ipcomp_zeroize() used when IPCA is deleted */ static int ipcomp_zeroize(struct secasvar *sav) { int err; err = crypto_freesession(sav->tdb_cryptoid); sav->tdb_cryptoid = 0; return err; } /* * ipcomp_input() gets called to uncompress an input packet */ static int ipcomp_input(struct mbuf *m, struct secasvar *sav, int skip, int protoff) { struct xform_data *xd; struct cryptodesc *crdc; struct cryptop *crp; struct ipcomp *ipcomp; caddr_t addr; int hlen = IPCOMP_HLENGTH; /* * Check that the next header of the IPComp is not IPComp again, before * doing any real work. Given it is not possible to do double * compression it means someone is playing tricks on us. */ if (m->m_len < skip + hlen && (m = m_pullup(m, skip + hlen)) == NULL) { IPCOMPSTAT_INC(ipcomps_hdrops); /*XXX*/ DPRINTF(("%s: m_pullup failed\n", __func__)); return (ENOBUFS); } addr = (caddr_t) mtod(m, struct ip *) + skip; ipcomp = (struct ipcomp *)addr; if (ipcomp->comp_nxt == IPPROTO_IPCOMP) { m_freem(m); IPCOMPSTAT_INC(ipcomps_pdrops); /* XXX have our own stats? */ DPRINTF(("%s: recursive compression detected\n", __func__)); return (EINVAL); } /* Get crypto descriptors */ crp = crypto_getreq(1); if (crp == NULL) { m_freem(m); DPRINTF(("%s: no crypto descriptors\n", __func__)); IPCOMPSTAT_INC(ipcomps_crypto); return ENOBUFS; } /* Get IPsec-specific opaque pointer */ xd = malloc(sizeof(*xd), M_XDATA, M_NOWAIT | M_ZERO); if (xd == NULL) { DPRINTF(("%s: cannot allocate xform_data\n", __func__)); IPCOMPSTAT_INC(ipcomps_crypto); crypto_freereq(crp); m_freem(m); return ENOBUFS; } crdc = crp->crp_desc; crdc->crd_skip = skip + hlen; crdc->crd_len = m->m_pkthdr.len - (skip + hlen); crdc->crd_inject = skip; /* Decompression operation */ crdc->crd_alg = sav->tdb_compalgxform->type; /* Crypto operation descriptor */ crp->crp_ilen = m->m_pkthdr.len - (skip + hlen); crp->crp_flags = CRYPTO_F_IMBUF | CRYPTO_F_CBIFSYNC; crp->crp_buf = (caddr_t) m; crp->crp_callback = ipcomp_input_cb; crp->crp_opaque = (caddr_t) xd; /* These are passed as-is to the callback */ xd->sav = sav; xd->protoff = protoff; xd->skip = skip; SECASVAR_LOCK(sav); crp->crp_sid = xd->cryptoid = sav->tdb_cryptoid; SECASVAR_UNLOCK(sav); return crypto_dispatch(crp); } /* * IPComp input callback from the crypto driver. */ static int ipcomp_input_cb(struct cryptop *crp) { char buf[IPSEC_ADDRSTRLEN]; struct cryptodesc *crd; struct xform_data *xd; struct mbuf *m; struct secasvar *sav; struct secasindex *saidx; caddr_t addr; uint64_t cryptoid; int hlen = IPCOMP_HLENGTH, error, clen; int skip, protoff; uint8_t nproto; crd = crp->crp_desc; m = (struct mbuf *) crp->crp_buf; xd = (struct xform_data *) crp->crp_opaque; sav = xd->sav; skip = xd->skip; protoff = xd->protoff; cryptoid = xd->cryptoid; saidx = &sav->sah->saidx; IPSEC_ASSERT(saidx->dst.sa.sa_family == AF_INET || saidx->dst.sa.sa_family == AF_INET6, ("unexpected protocol family %u", saidx->dst.sa.sa_family)); /* Check for crypto errors */ if (crp->crp_etype) { if (crp->crp_etype == EAGAIN) { /* Reset the session ID */ if (ipsec_updateid(sav, &crp->crp_sid, &cryptoid) != 0) crypto_freesession(cryptoid); xd->cryptoid = crp->crp_sid; return (crypto_dispatch(crp)); } IPCOMPSTAT_INC(ipcomps_noxform); DPRINTF(("%s: crypto error %d\n", __func__, crp->crp_etype)); error = crp->crp_etype; goto bad; } /* Shouldn't happen... */ if (m == NULL) { IPCOMPSTAT_INC(ipcomps_crypto); DPRINTF(("%s: null mbuf returned from crypto\n", __func__)); error = EINVAL; goto bad; } IPCOMPSTAT_INC(ipcomps_hist[sav->alg_comp]); clen = crp->crp_olen; /* Length of data after processing */ /* Release the crypto descriptors */ free(xd, M_XDATA), xd = NULL; crypto_freereq(crp), crp = NULL; /* In case it's not done already, adjust the size of the mbuf chain */ m->m_pkthdr.len = clen + hlen + skip; if (m->m_len < skip + hlen && (m = m_pullup(m, skip + hlen)) == NULL) { IPCOMPSTAT_INC(ipcomps_hdrops); /*XXX*/ DPRINTF(("%s: m_pullup failed\n", __func__)); error = EINVAL; /*XXX*/ goto bad; } /* Keep the next protocol field */ addr = (caddr_t) mtod(m, struct ip *) + skip; nproto = ((struct ipcomp *) addr)->comp_nxt; /* Remove the IPCOMP header */ error = m_striphdr(m, skip, hlen); if (error) { IPCOMPSTAT_INC(ipcomps_hdrops); DPRINTF(("%s: bad mbuf chain, IPCA %s/%08lx\n", __func__, ipsec_address(&sav->sah->saidx.dst, buf, sizeof(buf)), (u_long) ntohl(sav->spi))); goto bad; } /* Restore the Next Protocol field */ m_copyback(m, protoff, sizeof (u_int8_t), (u_int8_t *) &nproto); switch (saidx->dst.sa.sa_family) { #ifdef INET6 case AF_INET6: error = ipsec6_common_input_cb(m, sav, skip, protoff); break; #endif #ifdef INET case AF_INET: error = ipsec4_common_input_cb(m, sav, skip, protoff); break; #endif default: panic("%s: Unexpected address family: %d saidx=%p", __func__, saidx->dst.sa.sa_family, saidx); } return error; bad: if (sav != NULL) key_freesav(&sav); if (m != NULL) m_freem(m); if (xd != NULL) free(xd, M_XDATA); if (crp != NULL) crypto_freereq(crp); return error; } /* * IPComp output routine, called by ipsec[46]_perform_request() */ static int ipcomp_output(struct mbuf *m, struct secpolicy *sp, struct secasvar *sav, u_int idx, int skip, int protoff) { char buf[IPSEC_ADDRSTRLEN]; struct comp_algo *ipcompx; struct cryptodesc *crdc; struct cryptop *crp; struct xform_data *xd; int error, ralen, maxpacketsize; IPSEC_ASSERT(sav != NULL, ("null SA")); ipcompx = sav->tdb_compalgxform; IPSEC_ASSERT(ipcompx != NULL, ("null compression xform")); /* * Do not touch the packet in case our payload to compress * is lower than the minimal threshold of the compression * alogrithm. We will just send out the data uncompressed. * See RFC 3173, 2.2. Non-Expansion Policy. */ if (m->m_pkthdr.len <= ipcompx->minlen) { IPCOMPSTAT_INC(ipcomps_threshold); return ipsec_process_done(m, sp, sav, idx); } ralen = m->m_pkthdr.len - skip; /* Raw payload length before comp. */ IPCOMPSTAT_INC(ipcomps_output); /* Check for maximum packet size violations. */ switch (sav->sah->saidx.dst.sa.sa_family) { #ifdef INET case AF_INET: maxpacketsize = IP_MAXPACKET; break; #endif /* INET */ #ifdef INET6 case AF_INET6: maxpacketsize = IPV6_MAXPACKET; break; #endif /* INET6 */ default: IPCOMPSTAT_INC(ipcomps_nopf); DPRINTF(("%s: unknown/unsupported protocol family %d, " "IPCA %s/%08lx\n", __func__, sav->sah->saidx.dst.sa.sa_family, ipsec_address(&sav->sah->saidx.dst, buf, sizeof(buf)), (u_long) ntohl(sav->spi))); error = EPFNOSUPPORT; goto bad; } if (ralen + skip + IPCOMP_HLENGTH > maxpacketsize) { IPCOMPSTAT_INC(ipcomps_toobig); DPRINTF(("%s: packet in IPCA %s/%08lx got too big " "(len %u, max len %u)\n", __func__, ipsec_address(&sav->sah->saidx.dst, buf, sizeof(buf)), (u_long) ntohl(sav->spi), ralen + skip + IPCOMP_HLENGTH, maxpacketsize)); error = EMSGSIZE; goto bad; } /* Update the counters */ IPCOMPSTAT_ADD(ipcomps_obytes, m->m_pkthdr.len - skip); m = m_unshare(m, M_NOWAIT); if (m == NULL) { IPCOMPSTAT_INC(ipcomps_hdrops); DPRINTF(("%s: cannot clone mbuf chain, IPCA %s/%08lx\n", __func__, ipsec_address(&sav->sah->saidx.dst, buf, sizeof(buf)), (u_long) ntohl(sav->spi))); error = ENOBUFS; goto bad; } /* Ok now, we can pass to the crypto processing. */ /* Get crypto descriptors */ crp = crypto_getreq(1); if (crp == NULL) { IPCOMPSTAT_INC(ipcomps_crypto); DPRINTF(("%s: failed to acquire crypto descriptor\n",__func__)); error = ENOBUFS; goto bad; } crdc = crp->crp_desc; /* Compression descriptor */ crdc->crd_skip = skip; crdc->crd_len = ralen; crdc->crd_flags = CRD_F_COMP; crdc->crd_inject = skip; /* Compression operation */ crdc->crd_alg = ipcompx->type; /* IPsec-specific opaque crypto info */ xd = malloc(sizeof(struct xform_data), M_XDATA, M_NOWAIT | M_ZERO); if (xd == NULL) { IPCOMPSTAT_INC(ipcomps_crypto); DPRINTF(("%s: failed to allocate xform_data\n", __func__)); crypto_freereq(crp); error = ENOBUFS; goto bad; } xd->sp = sp; xd->sav = sav; xd->idx = idx; xd->skip = skip; xd->protoff = protoff; /* Crypto operation descriptor */ crp->crp_ilen = m->m_pkthdr.len; /* Total input length */ crp->crp_flags = CRYPTO_F_IMBUF | CRYPTO_F_CBIFSYNC; crp->crp_buf = (caddr_t) m; crp->crp_callback = ipcomp_output_cb; crp->crp_opaque = (caddr_t) xd; SECASVAR_LOCK(sav); crp->crp_sid = xd->cryptoid = sav->tdb_cryptoid; SECASVAR_UNLOCK(sav); return crypto_dispatch(crp); bad: if (m) m_freem(m); return (error); } /* * IPComp output callback from the crypto driver. */ static int ipcomp_output_cb(struct cryptop *crp) { char buf[IPSEC_ADDRSTRLEN]; struct xform_data *xd; struct secpolicy *sp; struct secasvar *sav; struct mbuf *m; uint64_t cryptoid; u_int idx; int error, skip, protoff; m = (struct mbuf *) crp->crp_buf; xd = (struct xform_data *) crp->crp_opaque; idx = xd->idx; sp = xd->sp; sav = xd->sav; skip = xd->skip; protoff = xd->protoff; cryptoid = xd->cryptoid; /* Check for crypto errors */ if (crp->crp_etype) { if (crp->crp_etype == EAGAIN) { /* Reset the session ID */ if (ipsec_updateid(sav, &crp->crp_sid, &cryptoid) != 0) crypto_freesession(cryptoid); xd->cryptoid = crp->crp_sid; return (crypto_dispatch(crp)); } IPCOMPSTAT_INC(ipcomps_noxform); DPRINTF(("%s: crypto error %d\n", __func__, crp->crp_etype)); error = crp->crp_etype; goto bad; } /* Shouldn't happen... */ if (m == NULL) { IPCOMPSTAT_INC(ipcomps_crypto); DPRINTF(("%s: bogus return buffer from crypto\n", __func__)); error = EINVAL; goto bad; } IPCOMPSTAT_INC(ipcomps_hist[sav->alg_comp]); if (crp->crp_ilen - skip > crp->crp_olen) { struct mbuf *mo; struct ipcomp *ipcomp; int roff; uint8_t prot; /* Compression helped, inject IPCOMP header. */ mo = m_makespace(m, skip, IPCOMP_HLENGTH, &roff); if (mo == NULL) { IPCOMPSTAT_INC(ipcomps_wrap); DPRINTF(("%s: IPCOMP header inject failed " "for IPCA %s/%08lx\n", __func__, ipsec_address(&sav->sah->saidx.dst, buf, sizeof(buf)), (u_long) ntohl(sav->spi))); error = ENOBUFS; goto bad; } ipcomp = (struct ipcomp *)(mtod(mo, caddr_t) + roff); /* Initialize the IPCOMP header */ /* XXX alignment always correct? */ switch (sav->sah->saidx.dst.sa.sa_family) { #ifdef INET case AF_INET: ipcomp->comp_nxt = mtod(m, struct ip *)->ip_p; break; #endif /* INET */ #ifdef INET6 case AF_INET6: ipcomp->comp_nxt = mtod(m, struct ip6_hdr *)->ip6_nxt; break; #endif } ipcomp->comp_flags = 0; ipcomp->comp_cpi = htons((u_int16_t) ntohl(sav->spi)); /* Fix Next Protocol in IPv4/IPv6 header */ prot = IPPROTO_IPCOMP; m_copyback(m, protoff, sizeof(u_int8_t), (u_char *)&prot); /* Adjust the length in the IP header */ switch (sav->sah->saidx.dst.sa.sa_family) { #ifdef INET case AF_INET: mtod(m, struct ip *)->ip_len = htons(m->m_pkthdr.len); break; #endif /* INET */ #ifdef INET6 case AF_INET6: mtod(m, struct ip6_hdr *)->ip6_plen = htons(m->m_pkthdr.len) - sizeof(struct ip6_hdr); break; #endif /* INET6 */ default: IPCOMPSTAT_INC(ipcomps_nopf); DPRINTF(("%s: unknown/unsupported protocol " "family %d, IPCA %s/%08lx\n", __func__, sav->sah->saidx.dst.sa.sa_family, ipsec_address(&sav->sah->saidx.dst, buf, sizeof(buf)), (u_long) ntohl(sav->spi))); error = EPFNOSUPPORT; goto bad; } } else { /* Compression was useless, we have lost time. */ IPCOMPSTAT_INC(ipcomps_uncompr); DPRINTF(("%s: compressions was useless %d - %d <= %d\n", __func__, crp->crp_ilen, skip, crp->crp_olen)); /* XXX remember state to not compress the next couple * of packets, RFC 3173, 2.2. Non-Expansion Policy */ } /* Release the crypto descriptor */ free(xd, M_XDATA); crypto_freereq(crp); /* NB: m is reclaimed by ipsec_process_done. */ error = ipsec_process_done(m, sp, sav, idx); return (error); bad: if (m) m_freem(m); free(xd, M_XDATA); crypto_freereq(crp); key_freesav(&sav); key_freesp(&sp); return (error); } static struct xformsw ipcomp_xformsw = { XF_IPCOMP, XFT_COMP, "IPcomp", ipcomp_init, ipcomp_zeroize, ipcomp_input, ipcomp_output }; #ifdef INET static const struct encaptab *ipe4_cookie = NULL; extern struct domain inetdomain; static struct protosw ipcomp4_protosw = { .pr_type = SOCK_RAW, .pr_domain = &inetdomain, .pr_protocol = 0 /* IPPROTO_IPV[46] */, .pr_flags = PR_ATOMIC | PR_ADDR | PR_LASTHDR, .pr_input = ipcomp_nonexp_input, .pr_output = rip_output, .pr_ctloutput = rip_ctloutput, .pr_usrreqs = &rip_usrreqs }; static int ipcomp4_nonexp_encapcheck(const struct mbuf *m, int off, int proto, void *arg __unused) { union sockaddr_union src, dst; const struct ip *ip; if (V_ipcomp_enable == 0) return (0); if (proto != IPPROTO_IPV4 && proto != IPPROTO_IPV6) return (0); bzero(&src, sizeof(src)); bzero(&dst, sizeof(dst)); src.sa.sa_family = dst.sa.sa_family = AF_INET; src.sin.sin_len = dst.sin.sin_len = sizeof(struct sockaddr_in); ip = mtod(m, const struct ip *); src.sin.sin_addr = ip->ip_src; dst.sin.sin_addr = ip->ip_dst; return (ipcomp_encapcheck(&src, &dst)); } #endif #ifdef INET6 static const struct encaptab *ipe6_cookie = NULL; extern struct domain inet6domain; static struct protosw ipcomp6_protosw = { .pr_type = SOCK_RAW, .pr_domain = &inet6domain, .pr_protocol = 0 /* IPPROTO_IPV[46] */, .pr_flags = PR_ATOMIC | PR_ADDR | PR_LASTHDR, .pr_input = ipcomp_nonexp_input, .pr_output = rip6_output, .pr_ctloutput = rip6_ctloutput, .pr_usrreqs = &rip6_usrreqs }; static int ipcomp6_nonexp_encapcheck(const struct mbuf *m, int off, int proto, void *arg __unused) { union sockaddr_union src, dst; const struct ip6_hdr *ip6; if (V_ipcomp_enable == 0) return (0); if (proto != IPPROTO_IPV4 && proto != IPPROTO_IPV6) return (0); bzero(&src, sizeof(src)); bzero(&dst, sizeof(dst)); src.sa.sa_family = dst.sa.sa_family = AF_INET; src.sin6.sin6_len = dst.sin6.sin6_len = sizeof(struct sockaddr_in6); ip6 = mtod(m, const struct ip6_hdr *); src.sin6.sin6_addr = ip6->ip6_src; dst.sin6.sin6_addr = ip6->ip6_dst; if (IN6_IS_SCOPE_LINKLOCAL(&src.sin6.sin6_addr)) { /* XXX: sa6_recoverscope() */ src.sin6.sin6_scope_id = ntohs(src.sin6.sin6_addr.s6_addr16[1]); src.sin6.sin6_addr.s6_addr16[1] = 0; } if (IN6_IS_SCOPE_LINKLOCAL(&dst.sin6.sin6_addr)) { /* XXX: sa6_recoverscope() */ dst.sin6.sin6_scope_id = ntohs(dst.sin6.sin6_addr.s6_addr16[1]); dst.sin6.sin6_addr.s6_addr16[1] = 0; } return (ipcomp_encapcheck(&src, &dst)); } #endif static void ipcomp_attach(void) { #ifdef INET ipe4_cookie = encap_attach_func(AF_INET, -1, ipcomp4_nonexp_encapcheck, &ipcomp4_protosw, NULL); #endif #ifdef INET6 ipe6_cookie = encap_attach_func(AF_INET6, -1, ipcomp6_nonexp_encapcheck, &ipcomp6_protosw, NULL); #endif xform_register(&ipcomp_xformsw); } SYSINIT(ipcomp_xform_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ipcomp_attach, NULL);