Index: projects/routing/sys/net/rt_nhops.c =================================================================== --- projects/routing/sys/net/rt_nhops.c (revision 287067) +++ projects/routing/sys/net/rt_nhops.c (revision 287068) @@ -1,1191 +1,1385 @@ /*- * Copyright (c) 2014 * Alexander V. Chernikov * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ /* * Temporary file. In future it should be split between net/route.c * and per-AF files like netinet/in_rmx.c | netinet6/in6_rmx.c */ #include "opt_inet.h" #include "opt_inet6.h" #include "opt_route.h" #include "opt_mpath.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef RADIX_MPATH #include #endif #include #include #include #include #include #include #include #include #include #include #include #include #include struct fwd_info { fib_lookup_t *lookup; void *state; }; #define FWD_FSM_NONE 0 #define FWD_FSM_INIT 1 #define FWD_FSM_FWD 2 struct fwd_control { int fwd_state; /* FSM */ struct fwd_module *fm; }; #if 0 static struct fwd_info *fwd_db[FWD_SIZE]; static struct fwd_control *fwd_ctl[FWD_SIZE]; static TAILQ_HEAD(fwd_module_list, fwd_module) modulehead = TAILQ_HEAD_INITIALIZER(modulehead); static struct fwd_module_list fwd_modules[FWD_SIZE]; static uint8_t fwd_map_af[] = { AF_INET, AF_INET6, }; static struct rwlock fwd_lock; #define FWD_LOCK_INIT() rw_init(&fwd_lock, "fwd_lock") #define FWD_RLOCK() rw_rlock(&fwd_lock) #define FWD_RUNLOCK() rw_runlock(&fwd_lock) #define FWD_WLOCK() rw_wlock(&fwd_lock) #define FWD_WUNLOCK() rw_wunlock(&fwd_lock) int fwd_attach_fib(struct fwd_module *fm, u_int fib); int fwd_destroy_fib(struct fwd_module *fm, u_int fib); #endif static inline uint16_t fib_rte_to_nh_flags(int rt_flags); #ifdef INET +static void rib4_rte_to_nh_extended(struct rtentry *rte, struct in_addr dst, + struct rt4_extended *prt4); static void fib4_rte_to_nh_extended(struct rtentry *rte, struct in_addr dst, struct nhop4_extended *pnh4); static void fib4_rte_to_nh_basic(struct rtentry *rte, struct in_addr dst, struct nhop4_basic *pnh4); #endif #ifdef INET6 static void fib6_rte_to_nh_extended(struct rtentry *rte, struct in6_addr *dst, struct nhop6_extended *pnh6); static void fib6_rte_to_nh_basic(struct rtentry *rte, const struct in6_addr *dst, struct nhop6_basic *pnh6); static int fib6_storelladdr(struct ifnet *ifp, struct in6_addr *dst, int mm_flags, u_char *desten); static uint16_t fib6_get_ifa(struct rtentry *rte); static int fib6_lla_to_nh_basic(const struct in6_addr *dst, uint32_t scopeid, struct nhop6_basic *pnh6); static int fib6_lla_to_nh_extended(struct in6_addr *dst, uint32_t scopeid, struct nhop6_extended *pnh6); static int fib6_lla_to_nh(struct in6_addr *dst, uint32_t scopeid, struct nhop_prepend *nh, struct ifnet **lifp); #endif MALLOC_DEFINE(M_RTFIB, "rtfib", "routing fwd"); /* * Per-AF fast routines returning minimal needed info. * It is not safe to dereference any pointers since it * may end up with use-after-free case. * Typically it may be used to check if outgoing * interface matches or to calculate proper MTU. * * Note that returned interface pointer is logical one, * e.g. actual transmit ifp may be different. * Difference may be triggered by * 1) loopback routes installed for interface addresses. * e.g. for address 10.0.0.1 with prefix /24 bound to * interface ix0, "logical" interface will be "ix0", * while "trasmit" interface will be "lo0" since this is * loopback route. You should consider using other * functions if you need "transmit" interface or both. * * * Returns 0 on match, error code overwise. */ //#define NHOP_DIRECT #define RNTORT(p) ((struct rtentry *)(p)) /* * Copies proper nexthop data based on @nh_src nexthop. * * For non-ECMP nexthop function simply copies @nh_src. * For ECMP nexthops flowid is used to select proper * nexthop. * */ static inline void fib_choose_prepend(uint32_t fibnum, struct nhop_prepend *nh_src, uint32_t flowid, struct nhop_prepend *nh, int af) { struct nhop_multi *nh_multi; int idx; if ((nh_src->nh_flags & NHF_RECURSE) != 0) { /* * Recursive nexthop. Choose direct nexthop * based on flowid. */ nh_multi = (struct nhop_multi *)nh_src; idx = nh_multi->nh_nhops[flowid % nh_multi->nh_count]; #if 0 KASSERT((fibnum < rt_numfibs), ("fib4_lookup_prepend§: bad fibnum")); rnh = rt_tables_get_rnh(fibnum, AF_INET); //nh_src = &rnh->nhops[i]; #endif } *nh = *nh_src; /* TODO: Do some light-weight refcounting on egress ifp's */ } static inline void fib_free_nh_prepend(uint32_t fibnum, struct nhop_prepend *nh, int af) { /* TODO: Do some light-weight refcounting on egress ifp's */ } #ifdef INET void fib4_free_nh_prepend(uint32_t fibnum, struct nhop_prepend *nh) { fib_free_nh_prepend(fibnum, nh, AF_INET); } void fib4_choose_prepend(uint32_t fibnum, struct nhop_prepend *nh_src, uint32_t flowid, struct nhop_prepend *nh, struct nhop4_extended *nh_ext) { fib_choose_prepend(fibnum, nh_src, flowid, nh, AF_INET); if (nh_ext == NULL) return; nh_ext->nh_ifp = NH_LIFP(nh); nh_ext->nh_mtu = nh->nh_mtu; nh_ext->nh_flags = nh->nh_flags; #if 0 /* TODO: copy source/gw address from extended nexthop data */ nh_ext->nh_addr = ; nh_ext->nh_src= ; #endif } /* * Function performs lookup in IPv4 table fib @fibnum. * * In case of successful lookup @nh header is filled with * appropriate interface info and full L2 header to prepend. * * If no valid ARP record is present, NHF_L2_INCOMPLETE flag * is set and gateway address is stored into nh->d.gw4 * * If @nh_ext is not NULL, additional nexthop data is stored there. * * Returns 0 on success. * */ int fib4_lookup_prepend(uint32_t fibnum, struct in_addr dst, struct mbuf *m, struct nhop_prepend *nh, struct nhop4_extended *nh_ext) { struct radix_node_head *rnh; struct radix_node *rn; struct sockaddr_in *gw_sa, sin; struct ifnet *lifp; struct in_addr gw; struct ether_header *eh; int error, flags; //uint32_t flowid; struct rtentry *rte; KASSERT((fibnum < rt_numfibs), ("fib4_lookup_prepend: bad fibnum")); rnh = rt_tables_get_rnh(fibnum, AF_INET); if (rnh == NULL) return (EHOSTUNREACH); /* Prepare lookup key */ memset(&sin, 0, sizeof(sin)); sin.sin_len = sizeof(struct sockaddr_in); sin.sin_addr = dst; RADIX_NODE_HEAD_RLOCK(rnh); rn = rnh->rnh_matchaddr((void *)&sin, rnh); rte = RNTORT(rn); if (rn == NULL || ((rn->rn_flags & RNF_ROOT) != 0) || RT_LINK_IS_UP(rte->rt_ifp) == 0) { RADIX_NODE_HEAD_RUNLOCK(rnh); return (EHOSTUNREACH); } /* * Currently we fill in @nh ourselves. * In near future rte will have nhop index to copy from. */ /* Calculate L3 info */ flags = 0; nh->nh_mtu = min(rte->rt_mtu, rte->rt_ifp->if_mtu); if (rte->rt_flags & RTF_GATEWAY) { gw_sa = (struct sockaddr_in *)rte->rt_gateway; gw = gw_sa->sin_addr; } else gw = dst; /* Set flags */ flags = fib_rte_to_nh_flags(rte->rt_flags); gw_sa = (struct sockaddr_in *)rt_key(rte); if (gw_sa->sin_addr.s_addr == 0) flags |= NHF_DEFAULT; /* * TODO: nh L2/L3 resolve. * Currently all we have is rte ifp. * Simply use it. */ /* Save interface address ifp */ lifp = rte->rt_ifa->ifa_ifp; nh->aifp_idx = lifp->if_index; /* Save both logical and transmit interface indexes */ lifp = rte->rt_ifp; nh->lifp_idx = lifp->if_index; nh->i.ifp_idx = nh->lifp_idx; if (nh_ext != NULL) { /* Fill in extended info */ fib4_rte_to_nh_extended(rte, dst, nh_ext); } RADIX_NODE_HEAD_RUNLOCK(rnh); nh->nh_flags = flags; /* * Try to lookup L2 info. * Do this using separate LLE locks. * TODO: move this under radix lock. */ if (lifp->if_type == IFT_ETHER) { eh = (struct ether_header *)nh->d.data; /* * Fill in ethernet header. * It should be already presented if we're * sending data via known gateway. */ error = arpresolve_fast(lifp, gw, m ? m->m_flags : 0, eh->ether_dhost); if (error == 0) { memcpy(&eh->ether_shost, IF_LLADDR(lifp), ETHER_ADDR_LEN); eh->ether_type = htons(ETHERTYPE_IP); nh->nh_count = ETHER_HDR_LEN; return (0); } } /* Notify caller that no L2 info is linked */ nh->nh_count = 0; nh->nh_flags |= NHF_L2_INCOMPLETE; /* ..And save gateway address */ nh->d.gw4 = gw; return (0); } int fib4_sendmbuf(struct ifnet *ifp, struct mbuf *m, struct nhop_prepend *nh, struct in_addr dst) { int error; if (nh != NULL && (nh->nh_flags & NHF_L2_INCOMPLETE) == 0) { /* * Fast path case. Most packets should * be sent from here. * TODO: Make special ifnet * 'if_output_frame' handler for that. */ struct route_compat rc; struct ether_header *eh; rc.ro_flags = AF_INET << 8 | RT_NHOP; rc.ro_nh = nh; M_PREPEND(m, nh->nh_count, M_NOWAIT); if (m == NULL) return (ENOBUFS); eh = mtod(m, struct ether_header *); memcpy(eh, nh->d.data, nh->nh_count); error = (*ifp->if_output)(ifp, m, NULL, (struct route *)&rc); } else { struct sockaddr_in gw_out; memset(&gw_out, 0, sizeof(gw_out)); gw_out.sin_len = sizeof(gw_out); gw_out.sin_family = AF_INET; gw_out.sin_addr = nh ? nh->d.gw4 : dst; error = (*ifp->if_output)(ifp, m, (const struct sockaddr *)&gw_out, NULL); } return (error); } static inline uint16_t fib_rte_to_nh_flags(int rt_flags) { uint16_t res; res = (rt_flags & RTF_REJECT) ? NHF_REJECT : 0; res |= (rt_flags & RTF_BLACKHOLE) ? NHF_BLACKHOLE : 0; res |= (rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) ? NHF_REDIRECT : 0; res |= (rt_flags & RTF_BROADCAST) ? NHF_BROADCAST : 0; res |= (rt_flags & RTF_GATEWAY) ? NHF_GATEWAY : 0; return (res); } static void fib4_rte_to_nh_basic(struct rtentry *rte, struct in_addr dst, struct nhop4_basic *pnh4) { struct sockaddr_in *gw; pnh4->nh_ifp = rte->rt_ifa->ifa_ifp; pnh4->nh_mtu = min(rte->rt_mtu, rte->rt_ifp->if_mtu); if (rte->rt_flags & RTF_GATEWAY) { gw = (struct sockaddr_in *)rte->rt_gateway; pnh4->nh_addr = gw->sin_addr; } else pnh4->nh_addr = dst; /* Set flags */ pnh4->nh_flags = fib_rte_to_nh_flags(rte->rt_flags); gw = (struct sockaddr_in *)rt_key(rte); if (gw->sin_addr.s_addr == 0) pnh4->nh_flags |= NHF_DEFAULT; /* XXX: Set RTF_BROADCAST if GW address is broadcast */ } static void fib4_rte_to_nh_extended(struct rtentry *rte, struct in_addr dst, struct nhop4_extended *pnh4) { struct sockaddr_in *gw; struct in_ifaddr *ia; pnh4->nh_ifp = rte->rt_ifa->ifa_ifp; pnh4->nh_mtu = min(rte->rt_mtu, rte->rt_ifp->if_mtu); if (rte->rt_flags & RTF_GATEWAY) { gw = (struct sockaddr_in *)rte->rt_gateway; pnh4->nh_addr = gw->sin_addr; } else pnh4->nh_addr = dst; /* Set flags */ pnh4->nh_flags = fib_rte_to_nh_flags(rte->rt_flags); gw = (struct sockaddr_in *)rt_key(rte); if (gw->sin_addr.s_addr == 0) pnh4->nh_flags |= NHF_DEFAULT; /* XXX: Set RTF_BROADCAST if GW address is broadcast */ ia = ifatoia(rte->rt_ifa); pnh4->nh_src = IA_SIN(ia)->sin_addr; } +static void +rib4_rte_to_nh_extended(struct rtentry *rte, struct in_addr dst, + struct rt4_extended *prt4) +{ + struct sockaddr_in *gw; + struct in_ifaddr *ia; + + /* Do explicit nexthop zero unless we're copying it */ + memset(prt4, 0, sizeof(*prt4)); + + gw = ((struct sockaddr_in *)rt_key(rte)); + prt4->rt_addr = gw->sin_addr; + gw = ((struct sockaddr_in *)rt_mask(rte)); + prt4->rt_mask.s_addr = (gw != NULL) ? + gw->sin_addr.s_addr : INADDR_BROADCAST; + + if (rte->rt_flags & RTF_GATEWAY) { + gw = (struct sockaddr_in *)rte->rt_gateway; + prt4->rt_gateway = gw->sin_addr; + } else + prt4->rt_gateway = dst; + + prt4->rt_lifp = rte->rt_ifp; + prt4->rt_aifp = rte->rt_ifa->ifa_ifp; + prt4->rt_flags = rte->rt_flags; + prt4->rt_mtu = min(rte->rt_mtu, rte->rt_ifp->if_mtu); + + prt4->rt_nhop = 0; /* XXX: fill real nexthop */ + + ia = ifatoia(rte->rt_ifa); + prt4->rt_src = IA_SIN(ia)->sin_addr; +} + /* * Performs IPv4 route table lookup on @dst. Returns 0 on success. * Stores nexthop info provided @pnh4 structure. * Note that * - nh_ifp cannot be safely dereferenced * - nh_ifp represents ifaddr ifp (e.g. if looking up address on * interface "ix0" pointer to "ix0" interface will be returned instead * of "lo0") * - howewer mtu from "transmit" interface will be returned. */ int fib4_lookup_nh_basic(uint32_t fibnum, struct in_addr dst, uint32_t flowid, struct nhop4_basic *pnh4) { struct radix_node_head *rnh; struct radix_node *rn; struct sockaddr_in sin; struct rtentry *rte; KASSERT((fibnum < rt_numfibs), ("fib4_lookup_nh_basic: bad fibnum")); rnh = rt_tables_get_rnh(fibnum, AF_INET); if (rnh == NULL) return (ENOENT); /* Prepare lookup key */ memset(&sin, 0, sizeof(sin)); sin.sin_len = sizeof(struct sockaddr_in); sin.sin_addr = dst; RADIX_NODE_HEAD_RLOCK(rnh); rn = rnh->rnh_matchaddr((void *)&sin, rnh); if (rn != NULL && ((rn->rn_flags & RNF_ROOT) == 0)) { rte = RNTORT(rn); /* Ensure route & ifp is UP */ if (RT_LINK_IS_UP(rte->rt_ifp)) { fib4_rte_to_nh_basic(rte, dst, pnh4); RADIX_NODE_HEAD_RUNLOCK(rnh); return (0); } } RADIX_NODE_HEAD_RUNLOCK(rnh); return (ENOENT); } int fib4_lookup_nh_ifp(uint32_t fibnum, struct in_addr dst, uint32_t flowid, struct nhop4_basic *pnh4) { struct radix_node_head *rnh; struct radix_node *rn; struct sockaddr_in sin; struct rtentry *rte; KASSERT((fibnum < rt_numfibs), ("fib4_lookup_nh_ifp: bad fibnum")); rnh = rt_tables_get_rnh(fibnum, AF_INET); if (rnh == NULL) return (ENOENT); /* Prepare lookup key */ memset(&sin, 0, sizeof(sin)); sin.sin_len = sizeof(struct sockaddr_in); sin.sin_addr = dst; RADIX_NODE_HEAD_RLOCK(rnh); rn = rnh->rnh_matchaddr((void *)&sin, rnh); if (rn != NULL && ((rn->rn_flags & RNF_ROOT) == 0)) { rte = RNTORT(rn); /* Ensure route & ifp is UP */ if (RT_LINK_IS_UP(rte->rt_ifp)) { fib4_rte_to_nh_basic(rte, dst, pnh4); RADIX_NODE_HEAD_RUNLOCK(rnh); pnh4->nh_ifp = rte->rt_ifp; return (0); } } RADIX_NODE_HEAD_RUNLOCK(rnh); return (ENOENT); } /* * Performs IPv4 route table lookup on @dst. Returns 0 on success. * Stores extende nexthop info provided @pnh4 structure. * Note that * - nh_ifp cannot be safely dereferenced unless NHOP_LOOKUP_REF is specified. * - in that case you need to call fib4_free_nh_ext() * - nh_ifp represents logical transmit interface (rt_ifp) * - mtu from logical transmit interface will be returned. */ int fib4_lookup_nh_ext(uint32_t fibnum, struct in_addr dst, uint32_t flowid, uint32_t flags, struct nhop4_extended *pnh4) { struct radix_node_head *rnh; struct radix_node *rn; struct sockaddr_in sin; struct rtentry *rte; KASSERT((fibnum < rt_numfibs), ("fib4_lookup_nh_ext: bad fibnum")); rnh = rt_tables_get_rnh(fibnum, AF_INET); if (rnh == NULL) return (ENOENT); /* Prepare lookup key */ memset(&sin, 0, sizeof(sin)); sin.sin_len = sizeof(struct sockaddr_in); sin.sin_addr = dst; RADIX_NODE_HEAD_RLOCK(rnh); rn = rnh->rnh_matchaddr((void *)&sin, rnh); if (rn != NULL && ((rn->rn_flags & RNF_ROOT) == 0)) { rte = RNTORT(rn); /* Ensure route & ifp is UP */ if (RT_LINK_IS_UP(rte->rt_ifp)) { fib4_rte_to_nh_extended(rte, dst, pnh4); if ((flags & NHOP_LOOKUP_REF) != 0) { /* TODO: Do lwref on egress ifp's */ } RADIX_NODE_HEAD_RUNLOCK(rnh); return (0); } } RADIX_NODE_HEAD_RUNLOCK(rnh); return (ENOENT); } void fib4_free_nh_ext(uint32_t fibnum, struct nhop4_extended *pnh4) { } void fib4_source_to_sa_ext(const struct nhopu_extended *pnhu, struct sockaddr_in *sin) { sin->sin_family = AF_INET; sin->sin_len = sizeof(*sin); sin->sin_addr = pnhu->u.nh4.nh_src; } +int +rib4_lookup_nh_ext(uint32_t fibnum, struct in_addr dst, uint32_t flowid, + uint32_t flags, struct rt4_extended *prt4) +{ + struct radix_node_head *rnh; + struct radix_node *rn; + struct sockaddr_in sin; + struct rtentry *rte; + + KASSERT((fibnum < rt_numfibs), ("rib4_lookup_nh_ext: bad fibnum")); + rnh = rt_tables_get_rnh(fibnum, AF_INET); + if (rnh == NULL) + return (ENOENT); + + /* Prepare lookup key */ + memset(&sin, 0, sizeof(sin)); + sin.sin_len = sizeof(struct sockaddr_in); + sin.sin_addr = dst; + + RADIX_NODE_HEAD_RLOCK(rnh); + rn = rnh->rnh_matchaddr((void *)&sin, rnh); + if (rn != NULL && ((rn->rn_flags & RNF_ROOT) == 0)) { + rte = RNTORT(rn); + /* Ensure route & ifp is UP */ + if (RT_LINK_IS_UP(rte->rt_ifp)) { + rib4_rte_to_nh_extended(rte, dst, prt4); + if ((flags & NHOP_LOOKUP_REF) != 0) { + /* TODO: Do lwref on egress ifp's */ + } + RADIX_NODE_HEAD_RUNLOCK(rnh); + return (0); + } + } + RADIX_NODE_HEAD_RUNLOCK(rnh); + + return (ENOENT); +} + +void +rib4_free_nh_ext(uint32_t fibnum, struct rt4_extended *prt4) +{ + +} + #endif #ifdef INET6 void fib6_free_nh_prepend(uint32_t fibnum, struct nhop_prepend *nh) { fib_free_nh_prepend(fibnum, nh, AF_INET6); } void fib6_choose_prepend(uint32_t fibnum, struct nhop_prepend *nh_src, uint32_t flowid, struct nhop_prepend *nh, struct nhop6_extended *nh_ext) { fib_choose_prepend(fibnum, nh_src, flowid, nh, AF_INET6); if (nh_ext == NULL) return; nh_ext->nh_ifp = NH_LIFP(nh); nh_ext->nh_mtu = nh->nh_mtu; nh_ext->nh_flags = nh->nh_flags; /* nh_ext->nh_addr = ; nh_ext->nh_src= ; */ } /* * Temporary function to copy ethernet address from valid lle */ static int fib6_storelladdr(struct ifnet *ifp, struct in6_addr *dst, int mm_flags, u_char *desten) { struct llentry *ln; struct sockaddr_in6 dst_sa; if (mm_flags & M_MCAST) { ETHER_MAP_IPV6_MULTICAST(&dst, desten); return (0); } memset(&dst_sa, 0, sizeof(dst_sa)); dst_sa.sin6_family = AF_INET6; dst_sa.sin6_len = sizeof(dst_sa); dst_sa.sin6_addr = *dst; dst_sa.sin6_scope_id = ifp->if_index; /* * the entry should have been created in nd6_store_lladdr */ IF_AFDATA_RLOCK(ifp); ln = lla_lookup(LLTABLE6(ifp), 0, (struct sockaddr *)&dst_sa); /* * Perform fast path for the following cases: * 1) lle state is REACHABLE * 2) lle state is DELAY (NS message sentNS message sent) * * Every other case involves lle modification, so we handle * them separately. */ if (ln == NULL || (ln->ln_state != ND6_LLINFO_REACHABLE && ln->ln_state != ND6_LLINFO_DELAY)) { if (ln != NULL) LLE_RUNLOCK(ln); IF_AFDATA_RUNLOCK(ifp); return (1); } bcopy(&ln->ll_addr, desten, ifp->if_addrlen); LLE_RUNLOCK(ln); IF_AFDATA_RUNLOCK(ifp); return (0); } int fib6_lookup_prepend(uint32_t fibnum, struct in6_addr *dst, uint32_t scopeid, struct mbuf *m, struct nhop_prepend *nh, struct nhop6_extended *nh_ext) { struct radix_node_head *rnh; struct radix_node *rn; struct sockaddr_in6 sin6, *gw_sa; struct in6_addr gw6; struct rtentry *rte; struct ifnet *lifp; struct ether_header *eh; uint32_t flags; int error; if (IN6_IS_SCOPE_LINKLOCAL(dst)) { /* Do not lookup link-local addresses in rtable */ error = fib6_lla_to_nh(dst, scopeid, nh, &lifp); if (error != 0) return (error); /* */ gw6 = *dst; goto do_l2; } KASSERT((fibnum < rt_numfibs), ("fib6_lookup_prepend: bad fibnum")); rnh = rt_tables_get_rnh(fibnum, AF_INET6); if (rnh == NULL) return (ENOENT); /* Prepare lookup key */ memset(&sin6, 0, sizeof(sin6)); sin6.sin6_len = sizeof(struct sockaddr_in6); sin6.sin6_addr = *dst; sin6.sin6_scope_id = scopeid; sa6_embedscope(&sin6, 0); RADIX_NODE_HEAD_RLOCK(rnh); rn = rnh->rnh_matchaddr((void *)&sin6, rnh); rte = RNTORT(rn); if (rn == NULL || ((rn->rn_flags & RNF_ROOT) != 0) || RT_LINK_IS_UP(rte->rt_ifp) == 0) { RADIX_NODE_HEAD_RUNLOCK(rnh); return (EHOSTUNREACH); } /* Explicitly zero nexthop */ memset(nh, 0, sizeof(*nh)); flags = 0; nh->nh_mtu = min(rte->rt_mtu, IN6_LINKMTU(rte->rt_ifp)); if (rte->rt_flags & RTF_GATEWAY) { gw_sa = (struct sockaddr_in6 *)rte->rt_gateway; gw6 = gw_sa->sin6_addr; in6_clearscope(&gw6); } else gw6 = *dst; /* Set flags */ flags = fib_rte_to_nh_flags(rte->rt_flags); gw_sa = (struct sockaddr_in6 *)rt_key(rte); if (IN6_IS_ADDR_UNSPECIFIED(&gw_sa->sin6_addr)) flags |= NHF_DEFAULT; /* * TODO: nh L2/L3 resolve. * Currently all we have is rte ifp. * Simply use it. */ /* Save interface address ifp */ nh->aifp_idx = fib6_get_ifa(rte); /* Save both logical and transmit interface indexes */ lifp = rte->rt_ifp; nh->lifp_idx = lifp->if_index; nh->i.ifp_idx = nh->lifp_idx; RADIX_NODE_HEAD_RUNLOCK(rnh); nh->nh_flags = flags; do_l2: /* * Try to lookup L2 info. * Do this using separate LLE locks. * TODO: move this under radix lock. */ if (lifp->if_type == IFT_ETHER) { eh = (struct ether_header *)nh->d.data; /* * Fill in ethernet header. * It should be already presented if we're * sending data via known gateway. */ error = fib6_storelladdr(lifp, &gw6, m ? m->m_flags : 0, eh->ether_dhost); if (error == 0) { memcpy(&eh->ether_shost, IF_LLADDR(lifp), ETHER_ADDR_LEN); eh->ether_type = htons(ETHERTYPE_IPV6); nh->nh_count = ETHER_HDR_LEN; return (0); } } /* Notify caller that no L2 info is linked */ nh->nh_count = 0; nh->nh_flags |= NHF_L2_INCOMPLETE; /* ..And save gateway address */ nh->d.gw6 = gw6; return (0); } int fib6_sendmbuf(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m, struct nhop_prepend *nh) { int error; if (nh != NULL && (nh->nh_flags & NHF_L2_INCOMPLETE) == 0) { /* * Fast path case. Most packets should * be sent from here. * TODO: Make special ifnet * 'if_output_frame' handler for that. */ struct route_compat rc; struct ether_header *eh; rc.ro_flags = AF_INET6 << 8 | RT_NHOP; rc.ro_nh = nh; M_PREPEND(m, nh->nh_count, M_NOWAIT); if (m == NULL) return (ENOBUFS); eh = mtod(m, struct ether_header *); memcpy(eh, nh->d.data, nh->nh_count); error = (*ifp->if_output)(ifp, m, NULL, (struct route *)&rc); } else { /* We need to perform ND lookup */ struct sockaddr_in6 gw_out; memset(&gw_out, 0, sizeof(gw_out)); gw_out.sin6_family = AF_INET6; gw_out.sin6_len = sizeof(gw_out); gw_out.sin6_addr = nh->d.gw6; gw_out.sin6_scope_id = ifp->if_index; sa6_embedscope(&gw_out, 0); error = nd6_output(ifp, origifp, m, &gw_out, NULL); } return (error); } static uint16_t fib6_get_ifa(struct rtentry *rte) { struct ifnet *ifp; struct sockaddr_dl *sdl; ifp = rte->rt_ifp; if ((ifp->if_flags & IFF_LOOPBACK) && rte->rt_gateway->sa_family == AF_LINK) { sdl = (struct sockaddr_dl *)rte->rt_gateway; return (sdl->sdl_index); } return (ifp->if_index); #if 0 /* IPv6 case */ /* Alternative way to get interface address ifp */ /* * Adjust the "outgoing" interface. If we're going to loop * the packet back to ourselves, the ifp would be the loopback * interface. However, we'd rather know the interface associated * to the destination address (which should probably be one of * our own addresses.) */ if (rt) { if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) && (rt->rt_gateway->sa_family == AF_LINK)) *retifp = ifnet_byindex(((struct sockaddr_dl *) rt->rt_gateway)->sdl_index); } /* IPv4 case */ //pnh6->nh_ifp = rte->rt_ifa->ifa_ifp; #endif } static int fib6_lla_to_nh_basic(const struct in6_addr *dst, uint32_t scopeid, struct nhop6_basic *pnh6) { struct ifnet *ifp; ifp = ifnet_byindex_locked(scopeid); if (ifp == NULL) return (ENOENT); /* Do explicit nexthop zero unless we're copying it */ memset(pnh6, 0, sizeof(*pnh6)); pnh6->nh_ifp = ifp; pnh6->nh_mtu = IN6_LINKMTU(ifp); /* No flags set */ pnh6->nh_addr = *dst; return (0); } static int fib6_lla_to_nh_extended(struct in6_addr *dst, uint32_t scopeid, struct nhop6_extended *pnh6) { struct ifnet *ifp; ifp = ifnet_byindex_locked(scopeid); if (ifp == NULL) return (ENOENT); /* Do explicit nexthop zero unless we're copying it */ memset(pnh6, 0, sizeof(*pnh6)); pnh6->nh_ifp = ifp; pnh6->nh_mtu = IN6_LINKMTU(ifp); /* No flags set */ pnh6->nh_addr = *dst; return (0); } static int +rib6_lla_to_nh_extended(struct in6_addr *dst, uint32_t scopeid, + struct rt6_extended *prt6) +{ + struct ifnet *ifp; + + ifp = ifnet_byindex_locked(scopeid); + if (ifp == NULL) + return (ENOENT); + + /* Do explicit nexthop zero unless we're copying it */ + memset(prt6, 0, sizeof(*prt6)); + + prt6->rt_addr.s6_addr16[0] = htons(0xFE80); + prt6->rt_mask = 64; /* XXX check RFC */ + + prt6->rt_aifp = ifp; + prt6->rt_lifp = ifp; + /* Check id this is for-us address */ + if (in6_ifawithifp_lla(ifp, dst)) { + if ((ifp = V_loif) != NULL) + prt6->rt_lifp = ifp; + } + + prt6->rt_mtu = IN6_LINKMTU(ifp); + /* No flags set */ + + return (0); +} + +static int fib6_lla_to_nh(struct in6_addr *dst, uint32_t scopeid, struct nhop_prepend *nh, struct ifnet **lifp) { struct ifnet *ifp; ifp = ifnet_byindex_locked(scopeid); if (ifp == NULL) return (ENOENT); /* Do explicit nexthop zero unless we're copying it */ memset(nh, 0, sizeof(*nh)); /* No flags set */ nh->nh_mtu = IN6_LINKMTU(ifp); /* Save lifp */ *lifp = ifp; nh->aifp_idx = scopeid; nh->lifp_idx = scopeid; /* Check id this is for-us address */ if (in6_ifawithifp_lla(ifp, dst)) { if ((ifp = V_loif) != NULL) nh->lifp_idx = ifp->if_index; } return (0); } static void fib6_rte_to_nh_basic(struct rtentry *rte, const struct in6_addr *dst, struct nhop6_basic *pnh6) { struct sockaddr_in6 *gw; /* Do explicit nexthop zero unless we're copying it */ memset(pnh6, 0, sizeof(*pnh6)); pnh6->nh_ifp = ifnet_byindex(fib6_get_ifa(rte)); pnh6->nh_mtu = min(rte->rt_mtu, IN6_LINKMTU(rte->rt_ifp)); if (rte->rt_flags & RTF_GATEWAY) { gw = (struct sockaddr_in6 *)rte->rt_gateway; pnh6->nh_addr = gw->sin6_addr; in6_clearscope(&pnh6->nh_addr); } else pnh6->nh_addr = *dst; /* Set flags */ pnh6->nh_flags = fib_rte_to_nh_flags(rte->rt_flags); gw = (struct sockaddr_in6 *)rt_key(rte); if (IN6_IS_ADDR_UNSPECIFIED(&gw->sin6_addr)) pnh6->nh_flags |= NHF_DEFAULT; } static void fib6_rte_to_nh_extended(struct rtentry *rte, struct in6_addr *dst, struct nhop6_extended *pnh6) { struct sockaddr_in6 *gw; struct in6_ifaddr *ia; /* Do explicit nexthop zero unless we're copying it */ memset(pnh6, 0, sizeof(*pnh6)); pnh6->nh_ifp = ifnet_byindex(fib6_get_ifa(rte)); pnh6->nh_mtu = min(rte->rt_mtu, IN6_LINKMTU(rte->rt_ifp)); if (rte->rt_flags & RTF_GATEWAY) { gw = (struct sockaddr_in6 *)rte->rt_gateway; pnh6->nh_addr = gw->sin6_addr; in6_clearscope(&pnh6->nh_addr); } else pnh6->nh_addr = *dst; /* Set flags */ pnh6->nh_flags = fib_rte_to_nh_flags(rte->rt_flags); gw = (struct sockaddr_in6 *)rt_key(rte); if (IN6_IS_ADDR_UNSPECIFIED(&gw->sin6_addr)) pnh6->nh_flags |= NHF_DEFAULT; ia = ifatoia6(rte->rt_ifa); } +#define ipv6_masklen(x) bitcount32((x).__u6_addr.__u6_addr32[0]) + \ + bitcount32((x).__u6_addr.__u6_addr32[1]) + \ + bitcount32((x).__u6_addr.__u6_addr32[2]) + \ + bitcount32((x).__u6_addr.__u6_addr32[3]) +static void +rib6_rte_to_nh_extended(struct rtentry *rte, struct in6_addr *dst, + struct rt6_extended *prt6) +{ + struct sockaddr_in6 *gw; + + /* Do explicit nexthop zero unless we're copying it */ + memset(prt6, 0, sizeof(*prt6)); + + gw = ((struct sockaddr_in6 *)rt_key(rte)); + prt6->rt_addr = gw->sin6_addr; + gw = ((struct sockaddr_in6 *)rt_mask(rte)); + prt6->rt_mask = (gw != NULL) ? ipv6_masklen(gw->sin6_addr) : 128; + + if (rte->rt_flags & RTF_GATEWAY) { + gw = (struct sockaddr_in6 *)rte->rt_gateway; + prt6->rt_gateway = gw->sin6_addr; + in6_clearscope(&prt6->rt_gateway); + } else + prt6->rt_gateway = *dst; + + prt6->rt_lifp = rte->rt_ifp; + prt6->rt_aifp = ifnet_byindex(fib6_get_ifa(rte)); + prt6->rt_flags = fib_rte_to_nh_flags(rte->rt_flags); + prt6->rt_mtu = min(rte->rt_mtu, IN6_LINKMTU(rte->rt_ifp)); +} + int fib6_lookup_nh_ifp(uint32_t fibnum, struct in6_addr *dst, uint32_t scopeid, uint32_t flowid, struct nhop6_basic *pnh6) { struct radix_node_head *rnh; struct radix_node *rn; struct sockaddr_in6 sin6; struct rtentry *rte; if (IN6_IS_SCOPE_LINKLOCAL(dst)) { /* Do not lookup link-local addresses in rtable */ + /* XXX: Check if dst is local */ return (fib6_lla_to_nh_basic(dst, scopeid, pnh6)); } KASSERT((fibnum < rt_numfibs), ("fib6_lookup_nh_basic: bad fibnum")); rnh = rt_tables_get_rnh(fibnum, AF_INET6); if (rnh == NULL) return (ENOENT); /* Prepare lookup key */ memset(&sin6, 0, sizeof(sin6)); sin6.sin6_addr = *dst; sin6.sin6_scope_id = scopeid; sa6_embedscope(&sin6, 0); RADIX_NODE_HEAD_RLOCK(rnh); rn = rnh->rnh_matchaddr((void *)&sin6, rnh); if (rn != NULL && ((rn->rn_flags & RNF_ROOT) == 0)) { rte = RNTORT(rn); /* Ensure route & ifp is UP */ if (RT_LINK_IS_UP(rte->rt_ifp)) { fib6_rte_to_nh_basic(rte, dst, pnh6); pnh6->nh_ifp = rte->rt_ifp; RADIX_NODE_HEAD_RUNLOCK(rnh); return (0); } } RADIX_NODE_HEAD_RUNLOCK(rnh); return (ENOENT); } int fib6_lookup_nh_basic(uint32_t fibnum, const struct in6_addr *dst, uint32_t scopeid, uint32_t flowid, struct nhop6_basic *pnh6) { struct radix_node_head *rnh; struct radix_node *rn; struct sockaddr_in6 sin6; struct rtentry *rte; if (IN6_IS_SCOPE_LINKLOCAL(dst)) { /* Do not lookup link-local addresses in rtable */ return (fib6_lla_to_nh_basic(dst, scopeid, pnh6)); } KASSERT((fibnum < rt_numfibs), ("fib6_lookup_nh_basic: bad fibnum")); rnh = rt_tables_get_rnh(fibnum, AF_INET6); if (rnh == NULL) return (ENOENT); /* Prepare lookup key */ memset(&sin6, 0, sizeof(sin6)); sin6.sin6_addr = *dst; sin6.sin6_scope_id = scopeid; sa6_embedscope(&sin6, 0); RADIX_NODE_HEAD_RLOCK(rnh); rn = rnh->rnh_matchaddr((void *)&sin6, rnh); if (rn != NULL && ((rn->rn_flags & RNF_ROOT) == 0)) { rte = RNTORT(rn); /* Ensure route & ifp is UP */ if (RT_LINK_IS_UP(rte->rt_ifp)) { fib6_rte_to_nh_basic(rte, dst, pnh6); RADIX_NODE_HEAD_RUNLOCK(rnh); return (0); } } RADIX_NODE_HEAD_RUNLOCK(rnh); return (ENOENT); } /* * Performs IPv6 route table lookup on @dst. Returns 0 on success. * Stores extende nexthop info provided @pnh4 structure. * Note that * - nh_ifp cannot be safely dereferenced unless NHOP_LOOKUP_REF is specified. * - in that case you need to call fib6_free_nh_ext() * - nh_ifp represents logical transmit interface (rt_ifp) * - mtu from logical transmit interface will be returned. */ int fib6_lookup_nh_ext(uint32_t fibnum, struct in6_addr *dst, uint32_t scopeid, uint32_t flowid, uint32_t flags, struct nhop6_extended *pnh6) { struct radix_node_head *rnh; struct radix_node *rn; struct sockaddr_in6 sin6; struct rtentry *rte; if (IN6_IS_SCOPE_LINKLOCAL(dst)) { /* Do not lookup link-local addresses in rtable */ /* XXX: Do lwref on egress ifp */ return (fib6_lla_to_nh_extended(dst, scopeid, pnh6)); } KASSERT((fibnum < rt_numfibs), ("fib4_lookup_nh_ext: bad fibnum")); rnh = rt_tables_get_rnh(fibnum, AF_INET6); if (rnh == NULL) return (ENOENT); /* Prepare lookup key */ memset(&sin6, 0, sizeof(sin6)); sin6.sin6_len = sizeof(struct sockaddr_in6); sin6.sin6_addr = *dst; sin6.sin6_scope_id = scopeid; sa6_embedscope(&sin6, 0); RADIX_NODE_HEAD_RLOCK(rnh); rn = rnh->rnh_matchaddr((void *)&sin6, rnh); if (rn != NULL && ((rn->rn_flags & RNF_ROOT) == 0)) { rte = RNTORT(rn); /* Ensure route & ifp is UP */ if (RT_LINK_IS_UP(rte->rt_ifp)) { fib6_rte_to_nh_extended(rte, dst, pnh6); if ((flags & NHOP_LOOKUP_REF) != 0) { /* TODO: Do lwref on egress ifp's */ } RADIX_NODE_HEAD_RUNLOCK(rnh); return (0); } } RADIX_NODE_HEAD_RUNLOCK(rnh); return (ENOENT); } void fib6_free_nh_ext(uint32_t fibnum, struct nhop6_extended *pnh6) +{ + +} + +int +rib6_lookup_nh_ext(uint32_t fibnum, struct in6_addr *dst, uint32_t scopeid, + uint32_t flowid, uint32_t flags, struct rt6_extended *prt6) +{ + struct radix_node_head *rnh; + struct radix_node *rn; + struct sockaddr_in6 sin6; + struct rtentry *rte; + + if (IN6_IS_SCOPE_LINKLOCAL(dst)) { + /* Do not lookup link-local addresses in rtable */ + /* XXX: Do lwref on egress ifp */ + return (rib6_lla_to_nh_extended(dst, scopeid, prt6)); + } + + KASSERT((fibnum < rt_numfibs), ("rib6_lookup_nh_ext: bad fibnum")); + rnh = rt_tables_get_rnh(fibnum, AF_INET6); + if (rnh == NULL) + return (ENOENT); + + /* Prepare lookup key */ + memset(&sin6, 0, sizeof(sin6)); + sin6.sin6_len = sizeof(struct sockaddr_in6); + sin6.sin6_addr = *dst; + sin6.sin6_scope_id = scopeid; + sa6_embedscope(&sin6, 0); + + RADIX_NODE_HEAD_RLOCK(rnh); + rn = rnh->rnh_matchaddr((void *)&sin6, rnh); + if (rn != NULL && ((rn->rn_flags & RNF_ROOT) == 0)) { + rte = RNTORT(rn); + /* Ensure route & ifp is UP */ + if (RT_LINK_IS_UP(rte->rt_ifp)) { + rib6_rte_to_nh_extended(rte, dst, prt6); + if ((flags & NHOP_LOOKUP_REF) != 0) { + /* TODO: Do lwref on egress ifp's */ + } + RADIX_NODE_HEAD_RUNLOCK(rnh); + + return (0); + } + } + RADIX_NODE_HEAD_RUNLOCK(rnh); + + return (ENOENT); +} + +void +rib6_free_nh_ext(uint32_t fibnum, struct nhop6_extended *prt6) { } #endif void fib_free_nh_ext(uint32_t fibnum, struct nhopu_extended *pnhu) { } #if 0 typedef void nhop_change_cb_t(void *state); struct nhop_tracker { TAILQ_ENTRY(nhop_tracker) next; nhop_change_cb_t *f; void *state; uint32_t fibnum; struct sockaddr_storage ss; }; struct nhop_tracker * nhop_alloc_tracked(uint32_t fibnum, struct sockaddr *sa, nhop_change_cb_t *f, void *state) { struct nhop_tracker *nt; nt = malloc(sizeof(struct nhop_tracker), M_RTFIB, M_WAITOK | M_ZERO); nt->f = f; nt-state = state; nt->fibnum = fibnum; memcpy(&nt->ss, sa, sa->sa_len); return (nt); } int nhop_bind(struct nhop_tracker *nt) { NHOP_LOCK(nnh); NHOP_UNLOCK(nnh); return (0); } #endif Index: projects/routing/sys/net/rt_nhops.h =================================================================== --- projects/routing/sys/net/rt_nhops.h (revision 287067) +++ projects/routing/sys/net/rt_nhops.h (revision 287068) @@ -1,318 +1,329 @@ /*- * Copyright (c) 2014 * Alexander V. Chernikov * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #ifndef _NET_RT_NHOPS_H_ #define _NET_RT_NHOPS_H_ #define NH_TYPE_DIRECT 1 /* Directly reachable, no data */ #define NH_TYPE_BLACKHOLE 2 /* Blackhole route */ #define NH_TYPE_REJECT 3 /* Send reject */ #define NH_TYPE_L2 4 /* Provides full prepend header */ #define NH_TYPE_MUTATOR 5 /* NH+callback function */ #define NH_TYPE_MULTIPATH 6 /* Multipath route */ struct nhop_ctl_info { uint64_t refcnt; /* Use references */ uint64_t flags; /* Options */ }; /* Multipath nhop info */ struct nhop_mpath_info { uint16_t nhop; /* Netxthop id */ }; /* mutator info */ struct nhop_mutator_info; struct nhop_prepend; typedef int nhop_mutate_t(struct mbuf **, struct nhop_prepend *nd, void *storage); struct nhop_mutator_info { nhop_mutate_t *func; char data[]; }; /* Structures used for forwarding purposes */ #define MAX_PREPEND_LEN 48 /* Max data that can be prepended */ /* Non-recursive nexthop */ struct nhop_prepend { uint16_t nh_flags; /* NH flags */ uint8_t nh_count; /* Number of nexthops or data length */ uint8_t spare0; uint16_t nh_mtu; /* given nhop MTU */ uint16_t lifp_idx; /* Logical interface index */ union { uint16_t ifp_idx; /* Transmit interface index */ uint16_t nhop_idx; /* L2 multipath nhop index */ } i; uint16_t aifp_idx; /* Interface address index */ uint16_t spare1[2]; union { char data[MAX_PREPEND_LEN]; /* data to prepend */ #ifdef INET struct in_addr gw4; /* IPv4 gw address */ #endif #ifdef INET6 struct in6_addr gw6; /* IPv4 gw address */ #endif } d; }; /* Internal flags */ #define NHF_RECURSE 0x0001 /* Nexthop structure is recursive */ #define NHF_L2_NHOP 0x0002 /* L2 interface has to be selected */ #define NHF_L2_ME 0x0004 /* dst L2 address is our address */ #define NHF_L2_INCOMPLETE 0x0008 /* L2 header not prepended */ /* External flags */ #define NHF_REJECT 0x0010 /* RTF_REJECT */ #define NHF_BLACKHOLE 0x0020 /* RTF_BLACKHOLE */ #define NHF_REDIRECT 0x0040 /* RTF_DYNAMIC|RTF_MODIFIED */ #define NHF_DEFAULT 0x0080 /* Default route */ #define NHF_BROADCAST 0x0100 /* RTF_BROADCAST */ #define NHF_GATEWAY 0x0200 /* RTF_GATEWAY */ #define NH_LIFP(nh) ifnet_byindex_locked((nh)->lifp_idx) #define NH_TIFP(nh) ifnet_byindex_locked((nh)->i.ifp_idx) #define NH_AIFP(nh) ifnet_byindex_locked((nh)->aifp_idx) /* L2/L3 recursive nexthop */ struct nhop_multi { uint8_t nh_flags; /* NH flags */ uint8_t nh_count; /* Number of nexthops or data length */ uint8_t spare[2]; uint16_t nh_nhops[30]; /* Nexthop indexes */ }; /* Control plane nexthop data */ struct nhop_info { }; /* Per-AF per-fib nhop table */ struct nhops_descr { uint32_t nhop_size; /* Nehthop data size */ uint32_t nhops_max; /* Max number of nhops */ void *nhops_data; /* Pointer to nhop data table */ void *nhops_info; /* Pointer to nhop info table */ }; #if 0 typedef int nhop_resolve_t(struct sockaddr *dst, u_int fib, struct nhop_prepend *nd, struct nhop_info *nf); int lla_create_notify(struct sockaddr *dst, u_int fib, lla_notify_t *func, void *state, int flags); #endif /* Basic nexthop info used for uRPF/mtu checks */ struct nhop4_basic { struct ifnet *nh_ifp; /* Logical egress interface */ uint16_t nh_mtu; /* nexthop mtu */ uint16_t nh_flags; /* nhop flags */ struct in_addr nh_addr; /* GW/DST IPv4 address */ }; struct nhop6_basic { struct ifnet *nh_ifp; /* Logical egress interface */ uint16_t nh_mtu; /* nexthop mtu */ uint16_t nh_flags; /* nhop flags */ uint8_t spare[4]; struct in6_addr nh_addr; /* GW/DST IPv4 address */ }; struct nhopu_basic { union { struct nhop4_basic nh4; struct nhop6_basic nh6; } u; }; /* Extended nexthop info used for control protocols */ struct nhop4_extended { struct ifnet *nh_ifp; /* Logical egress interface */ uint16_t nh_mtu; /* nexthop mtu */ uint16_t nh_flags; /* nhop flags */ uint8_t spare[4]; struct in_addr nh_addr; /* GW/DST IPv4 address */ struct in_addr nh_src; /* default source IPv4 address */ uint64_t spare2[2]; }; /* Does not differ from nhop6_basic */ struct nhop6_extended { struct ifnet *nh_ifp; /* Logical egress interface */ uint16_t nh_mtu; /* nexthop mtu */ uint16_t nh_flags; /* nhop flags */ uint8_t spare[4]; struct in6_addr nh_addr; /* GW/DST IPv6 address */ uint64_t spare2[2]; }; /* route info used for control plane purposes */ -struct rt4_basic { +struct rt4_extended { struct in_addr rt_addr; /* route prefix */ struct in_addr rt_gateway; /* GW used */ + struct ifnet *rt_lifp; /* logical interface */ + struct ifnet *rt_aifp; /* address interface */ int rt_flags; /* Copy of rte flags */ uint16_t rt_mtu; uint16_t rt_nhop; /* nexthop id (might bi mpath) */ struct in_addr rt_mask; /* route mask */ + struct in_addr rt_src; uint16_t spare[2]; }; -struct rt6_basic { +struct rt6_extended { struct in6_addr rt_addr; struct in6_addr rt_gateway; + struct ifnet *rt_lifp; /* logical interface */ + struct ifnet *rt_aifp; /* address interface */ int rt_flags; uint16_t rt_mtu; uint16_t rt_nhop; - uint8_t rt_mask; + uint8_t rt_mask; /*Hopefully, no more non-config masks */ uint8_t spare[7]; }; struct nhopu_extended { union { struct nhop4_extended nh4; struct nhop6_extended nh6; } u; }; struct route_info { struct nhop_prepend *ri_nh; /* Desired nexthop to use */ struct nhopu_basic *ri_nh_info; /* Get selected route info */ uint16_t ri_mtu; /* Get selected route MTU */ uint16_t spare; uint32_t scopeid; /* Desired scope id to use */ }; struct route_compat { struct nhop_prepend *ro_nh; void *spare0; void *spare1; int ro_flags; }; int fib4_lookup_nh_ifp(uint32_t fibnum, struct in_addr dst, uint32_t flowid, struct nhop4_basic *pnh4); int fib4_lookup_nh_basic(uint32_t fibnum, struct in_addr dst, uint32_t flowid, struct nhop4_basic *pnh4); int fib4_lookup_nh_ext(uint32_t fibnum, struct in_addr dst, uint32_t flowid, uint32_t flags, struct nhop4_extended *pnh4); void fib4_free_nh_ext(uint32_t fibnum, struct nhop4_extended *pnh4); #define NHOP_LOOKUP_REF 0x01 void fib4_source_to_sa_ext(const struct nhopu_extended *pnhu, struct sockaddr_in *sin); +int rib4_lookup_nh_ext(uint32_t fibnum, struct in_addr dst, uint32_t flowid, + uint32_t flags, struct rt4_extended *prt4); +void rib4_free_nh_ext(uint32_t fibnum, struct rt4_extended *prt4); int fib6_lookup_nh_ifp(uint32_t fibnum, struct in6_addr *dst, uint32_t scopeid, uint32_t flowid, struct nhop6_basic *pnh6); int fib6_lookup_nh_basic(uint32_t fibnum, const struct in6_addr *dst, uint32_t scopeid, uint32_t flowid, struct nhop6_basic *pnh6); int fib6_lookup_nh_ext(uint32_t fibnum, struct in6_addr *dst, uint32_t scopeid, uint32_t flowid, uint32_t flags, struct nhop6_extended *pnh6); void fib6_free_nh_ext(uint32_t fibnum, struct nhop6_extended *pnh6); +int rib6_lookup_nh_ext(uint32_t fibnum, struct in6_addr *dst, uint32_t scopeid, + uint32_t flowid, uint32_t flags, struct rt6_extended *prt6); +void rib6_free_nh_ext(uint32_t fibnum, struct nhop6_extended *prt6); void fib_free_nh_ext(uint32_t fibnum, struct nhopu_extended *pnhu); void fib4_free_nh_prepend(uint32_t fibnum, struct nhop_prepend *nh); void fib4_choose_prepend(uint32_t fibnum, struct nhop_prepend *nh_src, uint32_t flowid, struct nhop_prepend *nh, struct nhop4_extended *nh_ext); int fib4_lookup_prepend(uint32_t fibnum, struct in_addr dst, struct mbuf *m, struct nhop_prepend *nh, struct nhop4_extended *nh_ext); int fib4_sendmbuf(struct ifnet *ifp, struct mbuf *m, struct nhop_prepend *nh, struct in_addr dst); void fib6_free_nh_prepend(uint32_t fibnum, struct nhop_prepend *nh); void fib6_choose_prepend(uint32_t fibnum, struct nhop_prepend *nh_src, uint32_t flowid, struct nhop_prepend *nh, struct nhop6_extended *nh_ext); int fib6_lookup_prepend(uint32_t fibnum, struct in6_addr *dst, uint32_t scopeid, struct mbuf *m, struct nhop_prepend *nh, struct nhop6_extended *nh_ext); int fib6_sendmbuf(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m, struct nhop_prepend *nh); #define FWD_INET 0 #define FWD_INET6 1 #define FWD_SIZE 2 #define FWD_NAME_MAX 15 #define FWD_MULTIPATH 0x0001 /* has multipath support */ #define FWD_OLDMASKS 0x0002 /* has support for non-contig masks */ #define FWD_DEFAULT 0x0004 /* installs as default fib mechanism */ #define FWD_MANAGELOCK 0x0004 /* manage its own locking */ typedef void *fib_init_t(u_int fibnum); typedef void fib_destroy_t(void *state); typedef int fib_dump_t(void *state, struct radix_node_head *rnh); typedef int fib_change_t(void *state, int req, struct rtentry *rte, struct rt_addrinfo *info); typedef int fib_lookup_t(void *state, void *key, uint64_t *attr, u_int flowid, void *nhop); /* Structure used by external module */ struct fwd_module_info { uint8_t fwd_family; /* family we're registering to */ char name[FWD_NAME_MAX]; /* fwd module name */ uint32_t capabilities; fib_init_t *fib_init; fib_destroy_t *fib_destroy; fib_dump_t *fib_dump; fib_change_t *fib_change; fib_lookup_t *fib_lookup; }; /* Internal version of previous structure */ struct fwd_module { TAILQ_ENTRY(fwd_module) list; uint8_t fwd_family; char name[FWD_NAME_MAX]; uint32_t capabilities; fib_init_t *fib_init; fib_destroy_t *fib_destroy; fib_dump_t *fib_dump; fib_change_t *fib_change; fib_lookup_t *fib_lookup; }; int fwd_attach_module(struct fwd_module_info *m, void **); int fwd_destroy_module(void *state); int fwd_change_fib(struct radix_node_head *rnh, int req, struct rtentry *rte, struct rt_addrinfo *info); #endif Index: projects/routing/sys/netgraph/netflow/netflow.c =================================================================== --- projects/routing/sys/netgraph/netflow/netflow.c (revision 287067) +++ projects/routing/sys/netgraph/netflow/netflow.c (revision 287068) @@ -1,1196 +1,1134 @@ /*- * Copyright (c) 2010-2011 Alexander V. Chernikov * Copyright (c) 2004-2005 Gleb Smirnoff * Copyright (c) 2001-2003 Roman V. Palagin * 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. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * $SourceForge: netflow.c,v 1.41 2004/09/05 11:41:10 glebius Exp $ */ #include __FBSDID("$FreeBSD$"); #include "opt_inet6.h" #include "opt_route.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 #define NBUCKETS (65536) /* must be power of 2 */ /* This hash is for TCP or UDP packets. */ #define FULL_HASH(addr1, addr2, port1, port2) \ (((addr1 ^ (addr1 >> 16) ^ \ htons(addr2 ^ (addr2 >> 16))) ^ \ port1 ^ htons(port2)) & \ (NBUCKETS - 1)) /* This hash is for all other IP packets. */ #define ADDR_HASH(addr1, addr2) \ ((addr1 ^ (addr1 >> 16) ^ \ htons(addr2 ^ (addr2 >> 16))) & \ (NBUCKETS - 1)) /* Macros to shorten logical constructions */ /* XXX: priv must exist in namespace */ #define INACTIVE(fle) (time_uptime - fle->f.last > priv->nfinfo_inact_t) #define AGED(fle) (time_uptime - fle->f.first > priv->nfinfo_act_t) #define ISFREE(fle) (fle->f.packets == 0) /* * 4 is a magical number: statistically number of 4-packet flows is * bigger than 5,6,7...-packet flows by an order of magnitude. Most UDP/ICMP * scans are 1 packet (~ 90% of flow cache). TCP scans are 2-packet in case * of reachable host and 4-packet otherwise. */ #define SMALL(fle) (fle->f.packets <= 4) MALLOC_DEFINE(M_NETFLOW_HASH, "netflow_hash", "NetFlow hash"); static int export_add(item_p, struct flow_entry *); static int export_send(priv_p, fib_export_p, item_p, int); static int hash_insert(priv_p, struct flow_hash_entry *, struct flow_rec *, int, uint8_t, uint8_t); #ifdef INET6 static int hash6_insert(priv_p, struct flow_hash_entry *, struct flow6_rec *, int, uint8_t, uint8_t); #endif static void expire_flow(priv_p, fib_export_p, struct flow_entry *, int); /* * Generate hash for a given flow record. * * FIB is not used here, because: * most VRFS will carry public IPv4 addresses which are unique even * without FIB private addresses can overlap, but this is worked out * via flow_rec bcmp() containing fib id. In IPv6 world addresses are * all globally unique (it's not fully true, there is FC00::/7 for example, * but chances of address overlap are MUCH smaller) */ static inline uint32_t ip_hash(struct flow_rec *r) { switch (r->r_ip_p) { case IPPROTO_TCP: case IPPROTO_UDP: return FULL_HASH(r->r_src.s_addr, r->r_dst.s_addr, r->r_sport, r->r_dport); default: return ADDR_HASH(r->r_src.s_addr, r->r_dst.s_addr); } } #ifdef INET6 /* Generate hash for a given flow6 record. Use lower 4 octets from v6 addresses */ static inline uint32_t ip6_hash(struct flow6_rec *r) { switch (r->r_ip_p) { case IPPROTO_TCP: case IPPROTO_UDP: return FULL_HASH(r->src.r_src6.__u6_addr.__u6_addr32[3], r->dst.r_dst6.__u6_addr.__u6_addr32[3], r->r_sport, r->r_dport); default: return ADDR_HASH(r->src.r_src6.__u6_addr.__u6_addr32[3], r->dst.r_dst6.__u6_addr.__u6_addr32[3]); } } #endif /* * Detach export datagram from priv, if there is any. * If there is no, allocate a new one. */ static item_p get_export_dgram(priv_p priv, fib_export_p fe) { item_p item = NULL; mtx_lock(&fe->export_mtx); if (fe->exp.item != NULL) { item = fe->exp.item; fe->exp.item = NULL; } mtx_unlock(&fe->export_mtx); if (item == NULL) { struct netflow_v5_export_dgram *dgram; struct mbuf *m; m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); if (m == NULL) return (NULL); item = ng_package_data(m, NG_NOFLAGS); if (item == NULL) return (NULL); dgram = mtod(m, struct netflow_v5_export_dgram *); dgram->header.count = 0; dgram->header.version = htons(NETFLOW_V5); dgram->header.pad = 0; } return (item); } /* * Re-attach incomplete datagram back to priv. * If there is already another one, then send incomplete. */ static void return_export_dgram(priv_p priv, fib_export_p fe, item_p item, int flags) { /* * It may happen on SMP, that some thread has already * put its item there, in this case we bail out and * send what we have to collector. */ mtx_lock(&fe->export_mtx); if (fe->exp.item == NULL) { fe->exp.item = item; mtx_unlock(&fe->export_mtx); } else { mtx_unlock(&fe->export_mtx); export_send(priv, fe, item, flags); } } /* * The flow is over. Call export_add() and free it. If datagram is * full, then call export_send(). */ static void expire_flow(priv_p priv, fib_export_p fe, struct flow_entry *fle, int flags) { struct netflow_export_item exp; uint16_t version = fle->f.version; if ((priv->export != NULL) && (version == IPVERSION)) { exp.item = get_export_dgram(priv, fe); if (exp.item == NULL) { priv->nfinfo_export_failed++; if (priv->export9 != NULL) priv->nfinfo_export9_failed++; /* fle definitely contains IPv4 flow. */ uma_zfree_arg(priv->zone, fle, priv); return; } if (export_add(exp.item, fle) > 0) export_send(priv, fe, exp.item, flags); else return_export_dgram(priv, fe, exp.item, NG_QUEUE); } if (priv->export9 != NULL) { exp.item9 = get_export9_dgram(priv, fe, &exp.item9_opt); if (exp.item9 == NULL) { priv->nfinfo_export9_failed++; if (version == IPVERSION) uma_zfree_arg(priv->zone, fle, priv); #ifdef INET6 else if (version == IP6VERSION) uma_zfree_arg(priv->zone6, fle, priv); #endif else panic("ng_netflow: Unknown IP proto: %d", version); return; } if (export9_add(exp.item9, exp.item9_opt, fle) > 0) export9_send(priv, fe, exp.item9, exp.item9_opt, flags); else return_export9_dgram(priv, fe, exp.item9, exp.item9_opt, NG_QUEUE); } if (version == IPVERSION) uma_zfree_arg(priv->zone, fle, priv); #ifdef INET6 else if (version == IP6VERSION) uma_zfree_arg(priv->zone6, fle, priv); #endif } /* Get a snapshot of node statistics */ void ng_netflow_copyinfo(priv_p priv, struct ng_netflow_info *i) { i->nfinfo_bytes = counter_u64_fetch(priv->nfinfo_bytes); i->nfinfo_packets = counter_u64_fetch(priv->nfinfo_packets); i->nfinfo_bytes6 = counter_u64_fetch(priv->nfinfo_bytes6); i->nfinfo_packets6 = counter_u64_fetch(priv->nfinfo_packets6); i->nfinfo_sbytes = counter_u64_fetch(priv->nfinfo_sbytes); i->nfinfo_spackets = counter_u64_fetch(priv->nfinfo_spackets); i->nfinfo_sbytes6 = counter_u64_fetch(priv->nfinfo_sbytes6); i->nfinfo_spackets6 = counter_u64_fetch(priv->nfinfo_spackets6); i->nfinfo_act_exp = counter_u64_fetch(priv->nfinfo_act_exp); i->nfinfo_inact_exp = counter_u64_fetch(priv->nfinfo_inact_exp); i->nfinfo_used = uma_zone_get_cur(priv->zone); #ifdef INET6 i->nfinfo_used6 = uma_zone_get_cur(priv->zone6); #endif i->nfinfo_alloc_failed = priv->nfinfo_alloc_failed; i->nfinfo_export_failed = priv->nfinfo_export_failed; i->nfinfo_export9_failed = priv->nfinfo_export9_failed; i->nfinfo_realloc_mbuf = priv->nfinfo_realloc_mbuf; i->nfinfo_alloc_fibs = priv->nfinfo_alloc_fibs; i->nfinfo_inact_t = priv->nfinfo_inact_t; i->nfinfo_act_t = priv->nfinfo_act_t; } /* * Insert a record into defined slot. * * First we get for us a free flow entry, then fill in all * possible fields in it. * * TODO: consider dropping hash mutex while filling in datagram, * as this was done in previous version. Need to test & profile * to be sure. */ static int hash_insert(priv_p priv, struct flow_hash_entry *hsh, struct flow_rec *r, int plen, uint8_t flags, uint8_t tcp_flags) { struct flow_entry *fle; - struct sockaddr_in sin; - struct rtentry *rt; + struct rt4_extended rt4; mtx_assert(&hsh->mtx, MA_OWNED); fle = uma_zalloc_arg(priv->zone, priv, M_NOWAIT); if (fle == NULL) { priv->nfinfo_alloc_failed++; return (ENOMEM); } /* * Now fle is totally ours. It is detached from all lists, * we can safely edit it. */ fle->f.version = IPVERSION; bcopy(r, &fle->f.r, sizeof(struct flow_rec)); fle->f.bytes = plen; fle->f.packets = 1; fle->f.tcp_flags = tcp_flags; fle->f.first = fle->f.last = time_uptime; /* * First we do route table lookup on destination address. So we can * fill in out_ifx, dst_mask, nexthop, and dst_as in future releases. */ if ((flags & NG_NETFLOW_CONF_NODSTLOOKUP) == 0) { - bzero(&sin, sizeof(sin)); - sin.sin_len = sizeof(struct sockaddr_in); - sin.sin_family = AF_INET; - sin.sin_addr = fle->f.r.r_dst; - rt = rtalloc1_fib((struct sockaddr *)&sin, 0, 0, r->fib); - if (rt != NULL) { - fle->f.fle_o_ifx = rt->rt_ifp->if_index; - - if (rt->rt_flags & RTF_GATEWAY && - rt->rt_gateway->sa_family == AF_INET) - fle->f.next_hop = - ((struct sockaddr_in *)(rt->rt_gateway))->sin_addr; - - if (rt_mask(rt)) - fle->f.dst_mask = - bitcount32(((struct sockaddr_in *)rt_mask(rt))->sin_addr.s_addr); - else if (rt->rt_flags & RTF_HOST) - /* Give up. We can't determine mask :( */ - fle->f.dst_mask = 32; - - RTFREE_LOCKED(rt); + if (rib4_lookup_nh_ext(r->fib, fle->f.r.r_dst, 0, 0, &rt4) != 0) { + fle->f.fle_o_ifx = rt4.rt_lifp->if_index; + if (rt4.rt_flags & RTF_GATEWAY) + fle->f.next_hop = rt4.rt_gateway; + fle->f.dst_mask = bitcount32(rt4.rt_mask.s_addr); } } /* Do route lookup on source address, to fill in src_mask. */ if ((flags & NG_NETFLOW_CONF_NOSRCLOOKUP) == 0) { - bzero(&sin, sizeof(sin)); - sin.sin_len = sizeof(struct sockaddr_in); - sin.sin_family = AF_INET; - sin.sin_addr = fle->f.r.r_src; - rt = rtalloc1_fib((struct sockaddr *)&sin, 0, 0, r->fib); - if (rt != NULL) { - if (rt_mask(rt)) - fle->f.src_mask = - bitcount32(((struct sockaddr_in *)rt_mask(rt))->sin_addr.s_addr); - else if (rt->rt_flags & RTF_HOST) - /* Give up. We can't determine mask :( */ - fle->f.src_mask = 32; - - RTFREE_LOCKED(rt); + if (rib4_lookup_nh_ext(r->fib, fle->f.r.r_src, 0, 0, &rt4) != 0) { + if (rt4.rt_flags & RTF_GATEWAY) + fle->f.next_hop = rt4.rt_gateway; + fle->f.src_mask = bitcount32(rt4.rt_mask.s_addr); } } /* Push new flow at the and of hash. */ TAILQ_INSERT_TAIL(&hsh->head, fle, fle_hash); return (0); } #ifdef INET6 /* XXX: make normal function, instead of.. */ #define ipv6_masklen(x) bitcount32((x).__u6_addr.__u6_addr32[0]) + \ bitcount32((x).__u6_addr.__u6_addr32[1]) + \ bitcount32((x).__u6_addr.__u6_addr32[2]) + \ bitcount32((x).__u6_addr.__u6_addr32[3]) #define RT_MASK6(x) (ipv6_masklen(((struct sockaddr_in6 *)rt_mask(x))->sin6_addr)) static int hash6_insert(priv_p priv, struct flow_hash_entry *hsh6, struct flow6_rec *r, int plen, uint8_t flags, uint8_t tcp_flags) { struct flow6_entry *fle6; - struct sockaddr_in6 *src, *dst; - struct rtentry *rt; - struct route_in6 rin6; + struct rt6_extended rt6; mtx_assert(&hsh6->mtx, MA_OWNED); fle6 = uma_zalloc_arg(priv->zone6, priv, M_NOWAIT); if (fle6 == NULL) { priv->nfinfo_alloc_failed++; return (ENOMEM); } /* * Now fle is totally ours. It is detached from all lists, * we can safely edit it. */ fle6->f.version = IP6VERSION; bcopy(r, &fle6->f.r, sizeof(struct flow6_rec)); fle6->f.bytes = plen; fle6->f.packets = 1; fle6->f.tcp_flags = tcp_flags; fle6->f.first = fle6->f.last = time_uptime; /* * First we do route table lookup on destination address. So we can * fill in out_ifx, dst_mask, nexthop, and dst_as in future releases. */ if ((flags & NG_NETFLOW_CONF_NODSTLOOKUP) == 0) { - bzero(&rin6, sizeof(struct route_in6)); - dst = (struct sockaddr_in6 *)&rin6.ro_dst; - dst->sin6_len = sizeof(struct sockaddr_in6); - dst->sin6_family = AF_INET6; - dst->sin6_addr = r->dst.r_dst6; - - rin6.ro_rt = rtalloc1_fib((struct sockaddr *)dst, 0, 0, r->fib); - - if (rin6.ro_rt != NULL) { - rt = rin6.ro_rt; - fle6->f.fle_o_ifx = rt->rt_ifp->if_index; - - if (rt->rt_flags & RTF_GATEWAY && - rt->rt_gateway->sa_family == AF_INET6) - fle6->f.n.next_hop6 = - ((struct sockaddr_in6 *)(rt->rt_gateway))->sin6_addr; - - if (rt_mask(rt)) - fle6->f.dst_mask = RT_MASK6(rt); - else - fle6->f.dst_mask = 128; - - RTFREE_LOCKED(rt); + if (rib6_lookup_nh_ext(r->fib, &r->dst.r_dst6, 0, 0, 0, &rt6) == 0) { + fle6->f.fle_o_ifx = rt6.rt_lifp->if_index; + if ((rt6.rt_flags & RTF_GATEWAY) != 0) + fle6->f.n.next_hop6 = rt6.rt_gateway; + fle6->f.dst_mask = rt6.rt_mask; } } if ((flags & NG_NETFLOW_CONF_NODSTLOOKUP) == 0) { /* Do route lookup on source address, to fill in src_mask. */ - bzero(&rin6, sizeof(struct route_in6)); - src = (struct sockaddr_in6 *)&rin6.ro_dst; - src->sin6_len = sizeof(struct sockaddr_in6); - src->sin6_family = AF_INET6; - src->sin6_addr = r->src.r_src6; - - rin6.ro_rt = rtalloc1_fib((struct sockaddr *)src, 0, 0, r->fib); - - if (rin6.ro_rt != NULL) { - rt = rin6.ro_rt; - - if (rt_mask(rt)) - fle6->f.src_mask = RT_MASK6(rt); - else - fle6->f.src_mask = 128; - - RTFREE_LOCKED(rt); + if (rib6_lookup_nh_ext(r->fib, &r->src.r_src6, 0, 0, 0, &rt6) == 0) { + fle6->f.dst_mask = rt6.rt_mask; } } /* Push new flow at the and of hash. */ TAILQ_INSERT_TAIL(&hsh6->head, (struct flow_entry *)fle6, fle_hash); return (0); } #undef ipv6_masklen #undef RT_MASK6 #endif /* * Non-static functions called from ng_netflow.c */ /* Allocate memory and set up flow cache */ void ng_netflow_cache_init(priv_p priv) { struct flow_hash_entry *hsh; int i; /* Initialize cache UMA zone. */ priv->zone = uma_zcreate("NetFlow IPv4 cache", sizeof(struct flow_entry), NULL, NULL, NULL, NULL, UMA_ALIGN_CACHE, 0); uma_zone_set_max(priv->zone, CACHESIZE); #ifdef INET6 priv->zone6 = uma_zcreate("NetFlow IPv6 cache", sizeof(struct flow6_entry), NULL, NULL, NULL, NULL, UMA_ALIGN_CACHE, 0); uma_zone_set_max(priv->zone6, CACHESIZE); #endif /* Allocate hash. */ priv->hash = malloc(NBUCKETS * sizeof(struct flow_hash_entry), M_NETFLOW_HASH, M_WAITOK | M_ZERO); /* Initialize hash. */ for (i = 0, hsh = priv->hash; i < NBUCKETS; i++, hsh++) { mtx_init(&hsh->mtx, "hash mutex", NULL, MTX_DEF); TAILQ_INIT(&hsh->head); } #ifdef INET6 /* Allocate hash. */ priv->hash6 = malloc(NBUCKETS * sizeof(struct flow_hash_entry), M_NETFLOW_HASH, M_WAITOK | M_ZERO); /* Initialize hash. */ for (i = 0, hsh = priv->hash6; i < NBUCKETS; i++, hsh++) { mtx_init(&hsh->mtx, "hash mutex", NULL, MTX_DEF); TAILQ_INIT(&hsh->head); } #endif priv->nfinfo_bytes = counter_u64_alloc(M_WAITOK); priv->nfinfo_packets = counter_u64_alloc(M_WAITOK); priv->nfinfo_bytes6 = counter_u64_alloc(M_WAITOK); priv->nfinfo_packets6 = counter_u64_alloc(M_WAITOK); priv->nfinfo_sbytes = counter_u64_alloc(M_WAITOK); priv->nfinfo_spackets = counter_u64_alloc(M_WAITOK); priv->nfinfo_sbytes6 = counter_u64_alloc(M_WAITOK); priv->nfinfo_spackets6 = counter_u64_alloc(M_WAITOK); priv->nfinfo_act_exp = counter_u64_alloc(M_WAITOK); priv->nfinfo_inact_exp = counter_u64_alloc(M_WAITOK); ng_netflow_v9_cache_init(priv); CTR0(KTR_NET, "ng_netflow startup()"); } /* Initialize new FIB table for v5 and v9 */ int ng_netflow_fib_init(priv_p priv, int fib) { fib_export_p fe = priv_to_fib(priv, fib); CTR1(KTR_NET, "ng_netflow(): fib init: %d", fib); if (fe != NULL) return (0); if ((fe = malloc(sizeof(struct fib_export), M_NETGRAPH, M_NOWAIT | M_ZERO)) == NULL) return (ENOMEM); mtx_init(&fe->export_mtx, "export dgram lock", NULL, MTX_DEF); mtx_init(&fe->export9_mtx, "export9 dgram lock", NULL, MTX_DEF); fe->fib = fib; fe->domain_id = fib; if (atomic_cmpset_ptr((volatile uintptr_t *)&priv->fib_data[fib], (uintptr_t)NULL, (uintptr_t)fe) == 0) { /* FIB already set up by other ISR */ CTR3(KTR_NET, "ng_netflow(): fib init: %d setup %p but got %p", fib, fe, priv_to_fib(priv, fib)); mtx_destroy(&fe->export_mtx); mtx_destroy(&fe->export9_mtx); free(fe, M_NETGRAPH); } else { /* Increase counter for statistics */ CTR3(KTR_NET, "ng_netflow(): fib %d setup to %p (%p)", fib, fe, priv_to_fib(priv, fib)); priv->nfinfo_alloc_fibs++; } return (0); } /* Free all flow cache memory. Called from node close method. */ void ng_netflow_cache_flush(priv_p priv) { struct flow_entry *fle, *fle1; struct flow_hash_entry *hsh; struct netflow_export_item exp; fib_export_p fe; int i; bzero(&exp, sizeof(exp)); /* * We are going to free probably billable data. * Expire everything before freeing it. * No locking is required since callout is already drained. */ for (hsh = priv->hash, i = 0; i < NBUCKETS; hsh++, i++) TAILQ_FOREACH_SAFE(fle, &hsh->head, fle_hash, fle1) { TAILQ_REMOVE(&hsh->head, fle, fle_hash); fe = priv_to_fib(priv, fle->f.r.fib); expire_flow(priv, fe, fle, NG_QUEUE); } #ifdef INET6 for (hsh = priv->hash6, i = 0; i < NBUCKETS; hsh++, i++) TAILQ_FOREACH_SAFE(fle, &hsh->head, fle_hash, fle1) { TAILQ_REMOVE(&hsh->head, fle, fle_hash); fe = priv_to_fib(priv, fle->f.r.fib); expire_flow(priv, fe, fle, NG_QUEUE); } #endif uma_zdestroy(priv->zone); /* Destroy hash mutexes. */ for (i = 0, hsh = priv->hash; i < NBUCKETS; i++, hsh++) mtx_destroy(&hsh->mtx); /* Free hash memory. */ if (priv->hash != NULL) free(priv->hash, M_NETFLOW_HASH); #ifdef INET6 uma_zdestroy(priv->zone6); /* Destroy hash mutexes. */ for (i = 0, hsh = priv->hash6; i < NBUCKETS; i++, hsh++) mtx_destroy(&hsh->mtx); /* Free hash memory. */ if (priv->hash6 != NULL) free(priv->hash6, M_NETFLOW_HASH); #endif for (i = 0; i < priv->maxfibs; i++) { if ((fe = priv_to_fib(priv, i)) == NULL) continue; if (fe->exp.item != NULL) export_send(priv, fe, fe->exp.item, NG_QUEUE); if (fe->exp.item9 != NULL) export9_send(priv, fe, fe->exp.item9, fe->exp.item9_opt, NG_QUEUE); mtx_destroy(&fe->export_mtx); mtx_destroy(&fe->export9_mtx); free(fe, M_NETGRAPH); } counter_u64_free(priv->nfinfo_bytes); counter_u64_free(priv->nfinfo_packets); counter_u64_free(priv->nfinfo_bytes6); counter_u64_free(priv->nfinfo_packets6); counter_u64_free(priv->nfinfo_sbytes); counter_u64_free(priv->nfinfo_spackets); counter_u64_free(priv->nfinfo_sbytes6); counter_u64_free(priv->nfinfo_spackets6); counter_u64_free(priv->nfinfo_act_exp); counter_u64_free(priv->nfinfo_inact_exp); ng_netflow_v9_cache_flush(priv); } /* Insert packet from into flow cache. */ int ng_netflow_flow_add(priv_p priv, fib_export_p fe, struct ip *ip, caddr_t upper_ptr, uint8_t upper_proto, uint8_t flags, unsigned int src_if_index) { struct flow_entry *fle, *fle1; struct flow_hash_entry *hsh; struct flow_rec r; int hlen, plen; int error = 0; uint16_t eproto; uint8_t tcp_flags = 0; bzero(&r, sizeof(r)); if (ip->ip_v != IPVERSION) return (EINVAL); hlen = ip->ip_hl << 2; if (hlen < sizeof(struct ip)) return (EINVAL); eproto = ETHERTYPE_IP; /* Assume L4 template by default */ r.flow_type = NETFLOW_V9_FLOW_V4_L4; r.r_src = ip->ip_src; r.r_dst = ip->ip_dst; r.fib = fe->fib; plen = ntohs(ip->ip_len); r.r_ip_p = ip->ip_p; r.r_tos = ip->ip_tos; r.r_i_ifx = src_if_index; /* * XXX NOTE: only first fragment of fragmented TCP, UDP and * ICMP packet will be recorded with proper s_port and d_port. * Following fragments will be recorded simply as IP packet with * ip_proto = ip->ip_p and s_port, d_port set to zero. * I know, it looks like bug. But I don't want to re-implement * ip packet assebmling here. Anyway, (in)famous trafd works this way - * and nobody complains yet :) */ if ((ip->ip_off & htons(IP_OFFMASK)) == 0) switch(r.r_ip_p) { case IPPROTO_TCP: { struct tcphdr *tcp; tcp = (struct tcphdr *)((caddr_t )ip + hlen); r.r_sport = tcp->th_sport; r.r_dport = tcp->th_dport; tcp_flags = tcp->th_flags; break; } case IPPROTO_UDP: r.r_ports = *(uint32_t *)((caddr_t )ip + hlen); break; } counter_u64_add(priv->nfinfo_packets, 1); counter_u64_add(priv->nfinfo_bytes, plen); /* Find hash slot. */ hsh = &priv->hash[ip_hash(&r)]; mtx_lock(&hsh->mtx); /* * Go through hash and find our entry. If we encounter an * entry, that should be expired, purge it. We do a reverse * search since most active entries are first, and most * searches are done on most active entries. */ TAILQ_FOREACH_REVERSE_SAFE(fle, &hsh->head, fhead, fle_hash, fle1) { if (bcmp(&r, &fle->f.r, sizeof(struct flow_rec)) == 0) break; if ((INACTIVE(fle) && SMALL(fle)) || AGED(fle)) { TAILQ_REMOVE(&hsh->head, fle, fle_hash); expire_flow(priv, priv_to_fib(priv, fle->f.r.fib), fle, NG_QUEUE); counter_u64_add(priv->nfinfo_act_exp, 1); } } if (fle) { /* An existent entry. */ fle->f.bytes += plen; fle->f.packets ++; fle->f.tcp_flags |= tcp_flags; fle->f.last = time_uptime; /* * We have the following reasons to expire flow in active way: * - it hit active timeout * - a TCP connection closed * - it is going to overflow counter */ if (tcp_flags & TH_FIN || tcp_flags & TH_RST || AGED(fle) || (fle->f.bytes >= (CNTR_MAX - IF_MAXMTU)) ) { TAILQ_REMOVE(&hsh->head, fle, fle_hash); expire_flow(priv, priv_to_fib(priv, fle->f.r.fib), fle, NG_QUEUE); counter_u64_add(priv->nfinfo_act_exp, 1); } else { /* * It is the newest, move it to the tail, * if it isn't there already. Next search will * locate it quicker. */ if (fle != TAILQ_LAST(&hsh->head, fhead)) { TAILQ_REMOVE(&hsh->head, fle, fle_hash); TAILQ_INSERT_TAIL(&hsh->head, fle, fle_hash); } } } else /* A new flow entry. */ error = hash_insert(priv, hsh, &r, plen, flags, tcp_flags); mtx_unlock(&hsh->mtx); return (error); } #ifdef INET6 /* Insert IPv6 packet from into flow cache. */ int ng_netflow_flow6_add(priv_p priv, fib_export_p fe, struct ip6_hdr *ip6, caddr_t upper_ptr, uint8_t upper_proto, uint8_t flags, unsigned int src_if_index) { struct flow_entry *fle = NULL, *fle1; struct flow6_entry *fle6; struct flow_hash_entry *hsh; struct flow6_rec r; int plen; int error = 0; uint8_t tcp_flags = 0; /* check version */ if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) return (EINVAL); bzero(&r, sizeof(r)); r.src.r_src6 = ip6->ip6_src; r.dst.r_dst6 = ip6->ip6_dst; r.fib = fe->fib; /* Assume L4 template by default */ r.flow_type = NETFLOW_V9_FLOW_V6_L4; plen = ntohs(ip6->ip6_plen) + sizeof(struct ip6_hdr); #if 0 /* XXX: set DSCP/CoS value */ r.r_tos = ip->ip_tos; #endif if ((flags & NG_NETFLOW_IS_FRAG) == 0) { switch(upper_proto) { case IPPROTO_TCP: { struct tcphdr *tcp; tcp = (struct tcphdr *)upper_ptr; r.r_ports = *(uint32_t *)upper_ptr; tcp_flags = tcp->th_flags; break; } case IPPROTO_UDP: case IPPROTO_SCTP: r.r_ports = *(uint32_t *)upper_ptr; break; } } r.r_ip_p = upper_proto; r.r_i_ifx = src_if_index; counter_u64_add(priv->nfinfo_packets6, 1); counter_u64_add(priv->nfinfo_bytes6, plen); /* Find hash slot. */ hsh = &priv->hash6[ip6_hash(&r)]; mtx_lock(&hsh->mtx); /* * Go through hash and find our entry. If we encounter an * entry, that should be expired, purge it. We do a reverse * search since most active entries are first, and most * searches are done on most active entries. */ TAILQ_FOREACH_REVERSE_SAFE(fle, &hsh->head, fhead, fle_hash, fle1) { if (fle->f.version != IP6VERSION) continue; fle6 = (struct flow6_entry *)fle; if (bcmp(&r, &fle6->f.r, sizeof(struct flow6_rec)) == 0) break; if ((INACTIVE(fle6) && SMALL(fle6)) || AGED(fle6)) { TAILQ_REMOVE(&hsh->head, fle, fle_hash); expire_flow(priv, priv_to_fib(priv, fle->f.r.fib), fle, NG_QUEUE); counter_u64_add(priv->nfinfo_act_exp, 1); } } if (fle != NULL) { /* An existent entry. */ fle6 = (struct flow6_entry *)fle; fle6->f.bytes += plen; fle6->f.packets ++; fle6->f.tcp_flags |= tcp_flags; fle6->f.last = time_uptime; /* * We have the following reasons to expire flow in active way: * - it hit active timeout * - a TCP connection closed * - it is going to overflow counter */ if (tcp_flags & TH_FIN || tcp_flags & TH_RST || AGED(fle6) || (fle6->f.bytes >= (CNTR_MAX - IF_MAXMTU)) ) { TAILQ_REMOVE(&hsh->head, fle, fle_hash); expire_flow(priv, priv_to_fib(priv, fle->f.r.fib), fle, NG_QUEUE); counter_u64_add(priv->nfinfo_act_exp, 1); } else { /* * It is the newest, move it to the tail, * if it isn't there already. Next search will * locate it quicker. */ if (fle != TAILQ_LAST(&hsh->head, fhead)) { TAILQ_REMOVE(&hsh->head, fle, fle_hash); TAILQ_INSERT_TAIL(&hsh->head, fle, fle_hash); } } } else /* A new flow entry. */ error = hash6_insert(priv, hsh, &r, plen, flags, tcp_flags); mtx_unlock(&hsh->mtx); return (error); } #endif /* * Return records from cache to userland. * * TODO: matching particular IP should be done in kernel, here. */ int ng_netflow_flow_show(priv_p priv, struct ngnf_show_header *req, struct ngnf_show_header *resp) { struct flow_hash_entry *hsh; struct flow_entry *fle; struct flow_entry_data *data = (struct flow_entry_data *)(resp + 1); #ifdef INET6 struct flow6_entry_data *data6 = (struct flow6_entry_data *)(resp + 1); #endif int i, max; i = req->hash_id; if (i > NBUCKETS-1) return (EINVAL); #ifdef INET6 if (req->version == 6) { resp->version = 6; hsh = priv->hash6 + i; max = NREC6_AT_ONCE; } else #endif if (req->version == 4) { resp->version = 4; hsh = priv->hash + i; max = NREC_AT_ONCE; } else return (EINVAL); /* * We will transfer not more than NREC_AT_ONCE. More data * will come in next message. * We send current hash index and current record number in list * to userland, and userland should return it back to us. * Then, we will restart with new entry. * * The resulting cache snapshot can be inaccurate if flow expiration * is taking place on hash item between userland data requests for * this hash item id. */ resp->nentries = 0; for (; i < NBUCKETS; hsh++, i++) { int list_id; if (mtx_trylock(&hsh->mtx) == 0) { /* * Requested hash index is not available, * relay decision to skip or re-request data * to userland. */ resp->hash_id = i; resp->list_id = 0; return (0); } list_id = 0; TAILQ_FOREACH(fle, &hsh->head, fle_hash) { if (hsh->mtx.mtx_lock & MTX_CONTESTED) { resp->hash_id = i; resp->list_id = list_id; mtx_unlock(&hsh->mtx); return (0); } list_id++; /* Search for particular record in list. */ if (req->list_id > 0) { if (list_id < req->list_id) continue; /* Requested list position found. */ req->list_id = 0; } #ifdef INET6 if (req->version == 6) { struct flow6_entry *fle6; fle6 = (struct flow6_entry *)fle; bcopy(&fle6->f, data6 + resp->nentries, sizeof(fle6->f)); } else #endif bcopy(&fle->f, data + resp->nentries, sizeof(fle->f)); resp->nentries++; if (resp->nentries == max) { resp->hash_id = i; /* * If it was the last item in list * we simply skip to next hash_id. */ resp->list_id = list_id + 1; mtx_unlock(&hsh->mtx); return (0); } } mtx_unlock(&hsh->mtx); } resp->hash_id = resp->list_id = 0; return (0); } /* We have full datagram in privdata. Send it to export hook. */ static int export_send(priv_p priv, fib_export_p fe, item_p item, int flags) { struct mbuf *m = NGI_M(item); struct netflow_v5_export_dgram *dgram = mtod(m, struct netflow_v5_export_dgram *); struct netflow_v5_header *header = &dgram->header; struct timespec ts; int error = 0; /* Fill mbuf header. */ m->m_len = m->m_pkthdr.len = sizeof(struct netflow_v5_record) * header->count + sizeof(struct netflow_v5_header); /* Fill export header. */ header->sys_uptime = htonl(MILLIUPTIME(time_uptime)); getnanotime(&ts); header->unix_secs = htonl(ts.tv_sec); header->unix_nsecs = htonl(ts.tv_nsec); header->engine_type = 0; header->engine_id = fe->domain_id; header->pad = 0; header->flow_seq = htonl(atomic_fetchadd_32(&fe->flow_seq, header->count)); header->count = htons(header->count); if (priv->export != NULL) NG_FWD_ITEM_HOOK_FLAGS(error, item, priv->export, flags); else NG_FREE_ITEM(item); return (error); } /* Add export record to dgram. */ static int export_add(item_p item, struct flow_entry *fle) { struct netflow_v5_export_dgram *dgram = mtod(NGI_M(item), struct netflow_v5_export_dgram *); struct netflow_v5_header *header = &dgram->header; struct netflow_v5_record *rec; rec = &dgram->r[header->count]; header->count ++; KASSERT(header->count <= NETFLOW_V5_MAX_RECORDS, ("ng_netflow: export too big")); /* Fill in export record. */ rec->src_addr = fle->f.r.r_src.s_addr; rec->dst_addr = fle->f.r.r_dst.s_addr; rec->next_hop = fle->f.next_hop.s_addr; rec->i_ifx = htons(fle->f.fle_i_ifx); rec->o_ifx = htons(fle->f.fle_o_ifx); rec->packets = htonl(fle->f.packets); rec->octets = htonl(fle->f.bytes); rec->first = htonl(MILLIUPTIME(fle->f.first)); rec->last = htonl(MILLIUPTIME(fle->f.last)); rec->s_port = fle->f.r.r_sport; rec->d_port = fle->f.r.r_dport; rec->flags = fle->f.tcp_flags; rec->prot = fle->f.r.r_ip_p; rec->tos = fle->f.r.r_tos; rec->dst_mask = fle->f.dst_mask; rec->src_mask = fle->f.src_mask; rec->pad1 = 0; rec->pad2 = 0; /* Not supported fields. */ rec->src_as = rec->dst_as = 0; if (header->count == NETFLOW_V5_MAX_RECORDS) return (1); /* end of datagram */ else return (0); } /* Periodic flow expiry run. */ void ng_netflow_expire(void *arg) { struct flow_entry *fle, *fle1; struct flow_hash_entry *hsh; priv_p priv = (priv_p )arg; int used, i; /* * Going through all the cache. */ used = uma_zone_get_cur(priv->zone); for (hsh = priv->hash, i = 0; i < NBUCKETS; hsh++, i++) { /* * Skip entries, that are already being worked on. */ if (mtx_trylock(&hsh->mtx) == 0) continue; TAILQ_FOREACH_SAFE(fle, &hsh->head, fle_hash, fle1) { /* * Interrupt thread wants this entry! * Quick! Quick! Bail out! */ if (hsh->mtx.mtx_lock & MTX_CONTESTED) break; /* * Don't expire aggressively while hash collision * ratio is predicted small. */ if (used <= (NBUCKETS*2) && !INACTIVE(fle)) break; if ((INACTIVE(fle) && (SMALL(fle) || (used > (NBUCKETS*2)))) || AGED(fle)) { TAILQ_REMOVE(&hsh->head, fle, fle_hash); expire_flow(priv, priv_to_fib(priv, fle->f.r.fib), fle, NG_NOFLAGS); used--; counter_u64_add(priv->nfinfo_inact_exp, 1); } } mtx_unlock(&hsh->mtx); } #ifdef INET6 used = uma_zone_get_cur(priv->zone6); for (hsh = priv->hash6, i = 0; i < NBUCKETS; hsh++, i++) { struct flow6_entry *fle6; /* * Skip entries, that are already being worked on. */ if (mtx_trylock(&hsh->mtx) == 0) continue; TAILQ_FOREACH_SAFE(fle, &hsh->head, fle_hash, fle1) { fle6 = (struct flow6_entry *)fle; /* * Interrupt thread wants this entry! * Quick! Quick! Bail out! */ if (hsh->mtx.mtx_lock & MTX_CONTESTED) break; /* * Don't expire aggressively while hash collision * ratio is predicted small. */ if (used <= (NBUCKETS*2) && !INACTIVE(fle6)) break; if ((INACTIVE(fle6) && (SMALL(fle6) || (used > (NBUCKETS*2)))) || AGED(fle6)) { TAILQ_REMOVE(&hsh->head, fle, fle_hash); expire_flow(priv, priv_to_fib(priv, fle->f.r.fib), fle, NG_NOFLAGS); used--; counter_u64_add(priv->nfinfo_inact_exp, 1); } } mtx_unlock(&hsh->mtx); } #endif /* Schedule next expire. */ callout_reset(&priv->exp_callout, (1*hz), &ng_netflow_expire, (void *)priv); }