/*- * 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: src/sys/netinet/ip_divert.c,v 1.113.2.3 2006/12/29 19:25:48 jhb Exp $ */ #if !defined(KLD_MODULE) #include "opt_inet.h" #include "opt_ipfw.h" #include "opt_mac.h" #ifndef INET6 #error "IPD6IVERT requires INET.6" #endif #ifndef IP6FIREWALL #error "IP6DIVERT requires IP6FIREWALL" #endif #endif #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 /* * Divert sockets */ /* * Allocate enough space to hold a full IP packet */ #define DIVSNDQ (65536 + 100) #define DIVRCVQ (65536 + 100) /* * Divert sockets work in conjunction with ipfw, see the divert(4) * manpage for features. * Internally, packets selected by ipfw in ip6_input() or ip6_output(), * and never diverted before, are passed to the input queue of the * divert socket with a given 'divert_port' number (as specified in * the matching ipfw rule), and they are tagged with a 16 bit cookie * (representing the rule number of the matching ipfw rule), which * is passed to process reading from the socket. * * Packets written to the divert socket are again tagged with a cookie * (usually the same as above) and a destination address. * If the destination address is INADDR_ANY then the packet is * treated as outgoing and sent to ip6_output(), otherwise it is * treated as incoming and sent to ip6_input(). * In both cases, the packet is tagged with the cookie. * * On reinjection, processing in ip6_input() and ip6_output() * will be exactly the same as for the original packet, except that * ipfw processing will start at the rule number after the one * written in the cookie (so, tagging a packet with a cookie of 0 * will cause it to be effectively considered as a standard packet). */ /* Internal variables. */ static struct inpcbhead divcb; static struct inpcbinfo divcbinfo; static u_long div_sendspace = DIVSNDQ; /* XXX sysctl ? */ static u_long div_recvspace = DIVRCVQ; /* XXX sysctl ? */ /* * Initialize divert connection block queue. */ static void div6_zone_change(void *tag) { uma_zone_set_max(divcbinfo.ipi_zone, maxsockets); } static int div6_inpcb_init(void *mem, int size, int flags) { struct inpcb *inp = mem; INP_LOCK_INIT(inp, "inp", "divinp"); return (0); } static void div6_inpcb_fini(void *mem, int size) { struct inpcb *inp = mem; INP_LOCK_DESTROY(inp); } void div6_init(void) { INP_INFO_LOCK_INIT(&divcbinfo, "div"); LIST_INIT(&divcb); divcbinfo.listhead = &divcb; /* * XXX We don't use the hash list for divert IP, but it's easier * to allocate a one entry hash list than it is to check all * over the place for hashbase == NULL. */ divcbinfo.hashbase = hashinit(1, M_PCB, &divcbinfo.hashmask); divcbinfo.porthashbase = hashinit(1, M_PCB, &divcbinfo.porthashmask); divcbinfo.ipi_zone = uma_zcreate("div6cb", sizeof(struct in6pcb), NULL, NULL, div6_inpcb_init, div6_inpcb_fini, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); uma_zone_set_max(divcbinfo.ipi_zone, maxsockets); EVENTHANDLER_REGISTER(maxsockets_change, div6_zone_change, NULL, EVENTHANDLER_PRI_ANY); } /* * IPPROTO_DIVERT is not in the real IP protocol number space; this * function should never be called. Just in case, drop any packets. */ void div6_input(struct mbuf *m, int off) { ip6stat.ip6s_noproto++; m_freem(m); } /* * Divert a packet by passing it up to the divert socket at port 'port'. * * Setup generic address and protocol structures for div_input routine, * then pass them along with mbuf chain. */ static void divert6_packet(struct mbuf *m, int incoming) { struct in6pcb *in6p; struct socket *sa; u_int16_t nport; struct sockaddr_div div6src; struct m_tag *mtag; mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL); if (mtag == NULL) { printf("%s: no divert tag\n", __func__); m_freem(m); return; } /* * Record receive interface address, if any. * But only for incoming packets. */ memset(&div6src, 0, sizeof(div6src)); div6src.div_len = sizeof(div6src); div6src.div_family = AF_INET6; div6src.div_cookie = divert6_cookie(mtag); /* record matching rule */ if (incoming) { struct ifaddr *ifa; /* Sanity check */ M_ASSERTPKTHDR(m); /* Find IP address for receive interface */ TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) { if (ifa->ifa_addr == NULL) continue; if (ifa->ifa_addr->sa_family != AF_INET6) continue; memcpy(&div6src.div6_addr, &((struct sockaddr_div *) ifa->ifa_addr)->div6_addr, sizeof(struct in6_addr)); break; } } /* * Record the incoming interface name whenever we have one. */ if (m->m_pkthdr.rcvif) { memset(div6src.div_iface, 0, sizeof(div6src.div_iface)); strlcpy(div6src.div_iface, m->m_pkthdr.rcvif->if_xname, sizeof(div6src.div_iface)); } /* Put packet on socket queue, if any */ sa = NULL; nport = htons((u_int16_t)divert6_info(mtag)); INP_INFO_RLOCK(&divcbinfo); LIST_FOREACH(in6p, &divcb, inp_list) { INP_LOCK(in6p); /* XXX why does only one socket match? */ if (in6p->inp_lport == nport) { sa = in6p->inp_socket; SOCKBUF_LOCK(&sa->so_rcv); if (sbappendaddr_locked(&sa->so_rcv, (struct sockaddr *)&div6src, m, (struct mbuf *)0) == 0) { SOCKBUF_UNLOCK(&sa->so_rcv); sa = NULL; /* force mbuf reclaim below */ } else sorwakeup_locked(sa); INP_UNLOCK(in6p); break; } INP_UNLOCK(in6p); } INP_INFO_RUNLOCK(&divcbinfo); if (sa == NULL) { m_freem(m); ip6stat.ip6s_noproto++; ip6stat.ip6s_delivered--; } } /* * Deliver packet back into the IP processing machinery. * * If no address specified, or address is 0.0.0.0, send to ip_output(); * otherwise, send to ip_input() and mark as having been received on * the interface with that address. */ static int div6_output(struct socket *so, struct mbuf *m, struct sockaddr_div *sdiv, struct mbuf *control) { struct m_tag *mtag; struct divert6_tag *dt; int error = 0; printf("div6_output\n"); /* * An mbuf may hasn't come from userland, but we pretend * that it has. */ m->m_pkthdr.rcvif = NULL; m->m_nextpkt = NULL; if (control) m_freem(control); /* XXX */ if ((mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL)) == NULL) { mtag = m_tag_get(PACKET_TAG_DIVERT, sizeof(struct divert6_tag), M_NOWAIT | M_ZERO); if (mtag == NULL) { error = ENOBUFS; goto cantsend; } dt = (struct divert6_tag *)(mtag+1); m_tag_prepend(m, mtag); } else dt = (struct divert6_tag *)(mtag+1); /* Loopback avoidance and state recovery */ if (sdiv) { int i; dt->cookie = sdiv->div_cookie; for (i = 0; i < sizeof(sdiv->div_iface) && sdiv->div_iface[i]; i++) ; if ( i > 0 && i < sizeof(sdiv->div_iface)) m->m_pkthdr.rcvif = ifunit(sdiv->div_iface); } /* Reinject packet into the system as incoming or outgoing */ if (!sdiv || IN6_IS_ADDR_UNSPECIFIED(&(sdiv->div6_addr))) { //struct ip6_hdr *const ip6 = mtod(m, struct ip6_hdr *); struct in6pcb *in6p; dt->info |= IPV6_FW_DIVERT_OUTPUT_FLAG; INP_INFO_WLOCK(&divcbinfo); in6p = sotoinpcb(so); INP_LOCK(in6p); /* * Don't allow both user specified and setsockopt options, * and don't allow packet length sizes that will crash */ // EG - NFI #ifdef NFI if (((ip6->ip_hl != (sizeof (*ip6) >> 2)) && in6p->inp_options) || ((u_short)ntohs(ip6->ip_len) > m->m_pkthdr.len)) { error = EINVAL; m_freem(m); } else { /* Convert fields to host order for ip_output() */ ip6->ip_len = ntohs(ip6->ip_len); ip6->ip_off = ntohs(ip6->ip_off); /* Send packet to output processing */ ip6stat.ip6s_rawout++; /* XXX */ #ifdef MAC mac_create_mbuf_from_inpcb(inp, m); #endif error = ip_output(m, in6p->inp_options, NULL, ((so->so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0) | IP_ALLOWBROADCAST | IP_RAWOUTPUT, inp->inp_moptions, NULL); } #endif /* NFI */ INP_UNLOCK(in6p); INP_INFO_WUNLOCK(&divcbinfo); } else { dt->info |= IPV6_FW_DIVERT_LOOPBACK_FLAG; if (m->m_pkthdr.rcvif == NULL) { /* * No luck with the name, check by IP address. * Clear the port and the ifname to make sure * there are no distractions for ifa_ifwithaddr. */ struct ifaddr *ifa; memset(sdiv->div_iface, 0, sizeof(sdiv->div_iface)); sdiv->div_cookie = 0; ifa = ifa_ifwithaddr((struct sockaddr *) sdiv); if (ifa == NULL) { error = EADDRNOTAVAIL; goto cantsend; } m->m_pkthdr.rcvif = ifa->ifa_ifp; } #ifdef MAC SOCK_LOCK(so); mac_create_mbuf_from_socket(so, m); SOCK_UNLOCK(so); #endif /* Send packet to input processing */ ip6_input(m); } return error; cantsend: m_freem(m); return error; } static int div6_attach(struct socket *so, int proto, struct thread *td) { struct in6pcb *in6p; int error; INP_INFO_WLOCK(&divcbinfo); in6p = sotoin6pcb(so); if (in6p != 0) { INP_INFO_WUNLOCK(&divcbinfo); return EINVAL; } if (td && (error = suser(td)) != 0) { INP_INFO_WUNLOCK(&divcbinfo); return error; } error = soreserve(so, div_sendspace, div_recvspace); if (error) { INP_INFO_WUNLOCK(&divcbinfo); return error; } error = in_pcballoc(so, &divcbinfo); if (error) { INP_INFO_WUNLOCK(&divcbinfo); return error; } in6p = (struct in6pcb *)so->so_pcb; INP_INFO_WUNLOCK(&divcbinfo); in6p->inp_ip_p = proto; in6p->inp_vflag |= INP_IPV6; in6p->inp_flags |= INP_HDRINCL; INP_UNLOCK(in6p); return 0; } static int div6_detach(struct socket *so) { struct in6pcb *in6p; INP_INFO_WLOCK(&divcbinfo); in6p = sotoin6pcb(so); if (in6p == 0) { INP_INFO_WUNLOCK(&divcbinfo); return EINVAL; } INP_LOCK(in6p); in_pcbdetach(in6p); INP_INFO_WUNLOCK(&divcbinfo); return 0; } static int div6_bind(struct socket *so, struct sockaddr *nam, struct thread *td) { struct in6pcb *in6p; int error; printf("div6_bind\n"); INP_INFO_WLOCK(&divcbinfo); in6p = sotoin6pcb(so); if (in6p == 0) { INP_INFO_WUNLOCK(&divcbinfo); return EINVAL; } /* in_pcbbind assumes that nam is a sockaddr_in6 * and in_pcbbind requires a valid address. Since divert * sockets don't we need to make sure the address is * filled in properly. * XXX -- divert should not be abusing in_pcbind * and should probably have its own family. */ if (nam->sa_family != AF_INET6) error = EAFNOSUPPORT; else { ((struct sockaddr_in6 *)nam)->sin6_addr = in6addr_any; INP_LOCK(in6p); error = in6_pcbbind(in6p, nam, td->td_ucred); INP_UNLOCK(in6p); } INP_INFO_WUNLOCK(&divcbinfo); return error; } static int div6_shutdown(struct socket *so) { struct in6pcb *in6p; INP_INFO_RLOCK(&divcbinfo); in6p = sotoin6pcb(so); if (in6p == 0) { INP_INFO_RUNLOCK(&divcbinfo); return EINVAL; } INP_LOCK(in6p); INP_INFO_RUNLOCK(&divcbinfo); socantsendmore(so); INP_UNLOCK(in6p); return 0; } static int div6_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, struct mbuf *control, struct thread *td) { printf("div6_send\n"); /* Packet must have a header (but that's about it) */ if (m->m_len < sizeof (struct ip6_hdr) && (m = m_pullup(m, sizeof (struct ip6_hdr))) == 0) { ip6stat.ip6s_toosmall++; m_freem(m); return EINVAL; } /* Send packet */ return div6_output(so, m, (struct sockaddr_div *)nam, control); } void div6_ctlinput(int cmd, struct sockaddr *sa, void *vip) { struct in6_addr faddr; faddr = ((struct sockaddr_in6 *)sa)->sin6_addr; if (sa->sa_family != AF_INET6 || IN6_IS_ADDR_UNSPECIFIED(&faddr)) return; if (PRC_IS_REDIRECT(cmd)) return; } #ifdef NFI static int div6_pcblist(SYSCTL_HANDLER_ARGS) { int error, i, n; struct in6pcb *in6p, **in6p_list; // EG - NFI 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 == 0) { n = divcbinfo.ipi_count; req->oldidx = 2 * (sizeof xig) + (n + n/8) * sizeof(struct xinpcb); return 0; } if (req->newptr != 0) return EPERM; /* * OK, now we're committed to doing something. */ INP_INFO_RLOCK(&divcbinfo); gencnt = divcbinfo.ipi_gencnt; n = divcbinfo.ipi_count; INP_INFO_RUNLOCK(&divcbinfo); error = sysctl_wire_old_buffer(req, 2 * sizeof(xig) + n*sizeof(struct xinpcb)); if (error != 0) return (error); xig.xig_len = sizeof xig; xig.xig_count = n; xig.xig_gen = gencnt; xig.xig_sogen = so_gencnt; error = SYSCTL_OUT(req, &xig, sizeof xig); if (error) return error; in6p_list = malloc(n * sizeof *in6p_list, M_TEMP, M_WAITOK); if (in6p_list == NULL) return ENOMEM; INP_INFO_RLOCK(&divcbinfo); for (in6p = LIST_FIRST(divcbinfo.listhead), i = 0; in6p && i < n; in6p = LIST_NEXT(in6p, in6p_list)) { INP_LOCK(in6p); if (in6p->in6p_gencnt <= gencnt && cr_canseesocket(req->td->td_ucred, in6p->in6p_socket) == 0) in6p_list[i++] = in6p; INP_UNLOCK(in6p); } INP_INFO_RUNLOCK(&divcbinfo); n = i; error = 0; for (i = 0; i < n; i++) { in6p = in6p_list[i]; INP_LOCK(in6p); if (in6p->in6p_gencnt <= gencnt) { struct xinpcb xi; memset(&xi, 0, sizeof(xi)); xi.xi_len = sizeof xi; /* XXX should avoid extra copy */ bcopy(in6p, &xi.xi_in6p, sizeof *in6p); if (in6p->in6p_socket) sotoxsocket(in6p->in6p_socket, &xi.xi_socket); INP_UNLOCK(in6p); error = SYSCTL_OUT(req, &xi, sizeof xi); } else INP_UNLOCK(in6p); } if (!error) { /* * Give the user an updated idea of our state. * If the generation differs from what we told * her before, she knows that something happened * while we were processing this request, and it * might be necessary to retry. */ INP_INFO_RLOCK(&divcbinfo); xig.xig_gen = divcbinfo.ipi_gencnt; xig.xig_sogen = so_gencnt; xig.xig_count = divcbinfo.ipi_count; INP_INFO_RUNLOCK(&divcbinfo); error = SYSCTL_OUT(req, &xig, sizeof xig); } free(in6p_list, M_TEMP); return error; } #endif /* * This is the wrapper function for in_setsockaddr. We just pass down * the pcbinfo for in_setpeeraddr to lock. */ static int div6_sockaddr(struct socket *so, struct sockaddr **nam) { return (in6_setsockaddr(so, nam)); } /* * This is the wrapper function for in_setpeeraddr. We just pass down * the pcbinfo for in_setpeeraddr to lock. */ static int div6_peeraddr(struct socket *so, struct sockaddr **nam) { return (in6_setpeeraddr(so, nam)); } #ifdef NFI #ifdef SYSCTL_NODE SYSCTL_NODE(_net_inet6, IPPROTO_DIVERT, divert6, CTLFLAG_RW, 0, "IP6DIVERT"); SYSCTL_PROC(_net_inet6_divert, OID_AUTO, pcblist, CTLFLAG_RD, 0, 0, div6_pcblist, "S,xinpcb", "List of active divert sockets"); #endif #endif struct pr_usrreqs div6_usrreqs = { .pru_attach = div6_attach, .pru_bind = div6_bind, .pru_control = in6_control, .pru_detach = div6_detach, .pru_peeraddr = div6_peeraddr, .pru_send = div6_send, .pru_shutdown = div6_shutdown, .pru_sockaddr = div6_sockaddr, .pru_sosetlabel = in_pcbsosetlabel }; struct protosw div6_protosw = { .pr_type = SOCK_RAW, .pr_protocol = IPPROTO_DIVERT, .pr_flags = PR_ATOMIC|PR_ADDR, .pr_input = div6_input, .pr_ctlinput = div6_ctlinput, .pr_ctloutput = ip6_ctloutput, .pr_init = div6_init, .pr_usrreqs = &div6_usrreqs }; static int div6_modevent(module_t mod, int type, void *unused) { int err = 0; int n; switch (type) { case MOD_LOAD: /* * Protocol will be initialized by pf_proto_register(). * We don't have to register ip_protox because we are not * a true IP protocol that goes over the wire. */ err = pf_proto_register(PF_INET6, &div6_protosw); if (err != 0) { printf("div6_modevent: pf_proto_register() returned %d\n", err); } ip6_divert_ptr = divert6_packet; break; case MOD_QUIESCE: /* * IPDIVERT may normally not be unloaded because of the * potential race conditions. Tell kldunload we can't be * unloaded unless the unload is forced. */ err = EPERM; break; case MOD_UNLOAD: /* * Forced unload. * * Module ipdivert can only be unloaded if no sockets are * connected. Maybe this can be changed later to forcefully * disconnect any open sockets. * * XXXRW: Note that there is a slight race here, as a new * socket open request could be spinning on the lock and then * we destroy the lock. */ INP_INFO_WLOCK(&divcbinfo); n = divcbinfo.ipi_count; if (n != 0) { err = EBUSY; INP_INFO_WUNLOCK(&divcbinfo); break; } ip6_divert_ptr = NULL; err = pf_proto_unregister(PF_INET6, IPPROTO_DIVERT, SOCK_RAW); INP_INFO_WUNLOCK(&divcbinfo); INP_INFO_LOCK_DESTROY(&divcbinfo); uma_zdestroy(divcbinfo.ipi_zone); break; default: err = EOPNOTSUPP; break; } return err; } static moduledata_t ip6divertmod = { "ip6divert", div6_modevent, 0 }; DECLARE_MODULE(ip6divert, ip6divertmod, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY); MODULE_DEPEND(dummynet, ipfw, 2, 2, 2); MODULE_VERSION(ip6divert, 1);