Index: head/sys/contrib/pf/net/pf.c =================================================================== --- head/sys/contrib/pf/net/pf.c (revision 222487) +++ head/sys/contrib/pf/net/pf.c (revision 222488) @@ -1,7771 +1,7768 @@ /* $OpenBSD: pf.c,v 1.527 2007/02/22 15:23:23 pyr Exp $ */ /* add: $OpenBSD: pf.c,v 1.559 2007/09/18 18:45:59 markus Exp $ */ /* * Copyright (c) 2001 Daniel Hartmeier * Copyright (c) 2002,2003 Henning Brauer * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * * - Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * - 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 COPYRIGHT HOLDERS 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 * COPYRIGHT HOLDERS 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. * * Effort sponsored in part by the Defense Advanced Research Projects * Agency (DARPA) and Air Force Research Laboratory, Air Force * Materiel Command, USAF, under agreement number F30602-01-2-0537. * */ #ifdef __FreeBSD__ #include "opt_inet.h" #include "opt_inet6.h" #include __FBSDID("$FreeBSD$"); #endif #ifdef __FreeBSD__ #include "opt_bpf.h" #include "opt_pf.h" #ifdef DEV_BPF #define NBPFILTER DEV_BPF #else #define NBPFILTER 0 #endif #ifdef DEV_PFLOG #define NPFLOG DEV_PFLOG #else #define NPFLOG 0 #endif #ifdef DEV_PFSYNC #define NPFSYNC DEV_PFSYNC #else #define NPFSYNC 0 #endif #else #include "bpfilter.h" #include "pflog.h" #include "pfsync.h" #endif #include #include #include #include #include #include #include #include #ifdef __FreeBSD__ #include #include #else #include #endif #include #ifdef __FreeBSD__ #include #include #include #else #include #endif #include #include #include #include #ifndef __FreeBSD__ #include #endif #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifndef __FreeBSD__ #include #endif #include #include #if NPFSYNC > 0 #include #endif /* NPFSYNC > 0 */ #ifdef INET6 #include #include #include #include #ifdef __FreeBSD__ #include #include #endif #endif /* INET6 */ #ifdef __FreeBSD__ #include #include #include #include extern int ip_optcopy(struct ip *, struct ip *); extern int debug_pfugidhack; #endif #define DPFPRINTF(n, x) if (pf_status.debug >= (n)) printf x /* * Global variables */ struct pf_altqqueue pf_altqs[2]; struct pf_palist pf_pabuf; struct pf_altqqueue *pf_altqs_active; struct pf_altqqueue *pf_altqs_inactive; struct pf_status pf_status; u_int32_t ticket_altqs_active; u_int32_t ticket_altqs_inactive; int altqs_inactive_open; u_int32_t ticket_pabuf; struct pf_anchor_stackframe { struct pf_ruleset *rs; struct pf_rule *r; struct pf_anchor_node *parent; struct pf_anchor *child; } pf_anchor_stack[64]; #ifdef __FreeBSD__ uma_zone_t pf_src_tree_pl, pf_rule_pl; uma_zone_t pf_state_pl, pf_altq_pl, pf_pooladdr_pl; #else struct pool pf_src_tree_pl, pf_rule_pl; struct pool pf_state_pl, pf_altq_pl, pf_pooladdr_pl; #endif void pf_print_host(struct pf_addr *, u_int16_t, u_int8_t); void pf_init_threshold(struct pf_threshold *, u_int32_t, u_int32_t); void pf_add_threshold(struct pf_threshold *); int pf_check_threshold(struct pf_threshold *); void pf_change_ap(struct pf_addr *, u_int16_t *, u_int16_t *, u_int16_t *, struct pf_addr *, u_int16_t, u_int8_t, sa_family_t); int pf_modulate_sack(struct mbuf *, int, struct pf_pdesc *, struct tcphdr *, struct pf_state_peer *); #ifdef INET6 void pf_change_a6(struct pf_addr *, u_int16_t *, struct pf_addr *, u_int8_t); #endif /* INET6 */ void pf_change_icmp(struct pf_addr *, u_int16_t *, struct pf_addr *, struct pf_addr *, u_int16_t, u_int16_t *, u_int16_t *, u_int16_t *, u_int16_t *, u_int8_t, sa_family_t); #ifdef __FreeBSD__ void pf_send_tcp(struct mbuf *, const struct pf_rule *, sa_family_t, #else void pf_send_tcp(const struct pf_rule *, sa_family_t, #endif const struct pf_addr *, const struct pf_addr *, u_int16_t, u_int16_t, u_int32_t, u_int32_t, u_int8_t, u_int16_t, u_int16_t, u_int8_t, int, u_int16_t, struct ether_header *, struct ifnet *); void pf_send_icmp(struct mbuf *, u_int8_t, u_int8_t, sa_family_t, struct pf_rule *); struct pf_rule *pf_match_translation(struct pf_pdesc *, struct mbuf *, int, int, struct pfi_kif *, struct pf_addr *, u_int16_t, struct pf_addr *, u_int16_t, int); struct pf_rule *pf_get_translation(struct pf_pdesc *, struct mbuf *, int, int, struct pfi_kif *, struct pf_src_node **, struct pf_addr *, u_int16_t, struct pf_addr *, u_int16_t, struct pf_addr *, u_int16_t *); int pf_test_tcp(struct pf_rule **, struct pf_state **, int, struct pfi_kif *, struct mbuf *, int, void *, struct pf_pdesc *, struct pf_rule **, #ifdef __FreeBSD__ struct pf_ruleset **, struct ifqueue *, struct inpcb *); #else struct pf_ruleset **, struct ifqueue *); #endif int pf_test_udp(struct pf_rule **, struct pf_state **, int, struct pfi_kif *, struct mbuf *, int, void *, struct pf_pdesc *, struct pf_rule **, #ifdef __FreeBSD__ struct pf_ruleset **, struct ifqueue *, struct inpcb *); #else struct pf_ruleset **, struct ifqueue *); #endif int pf_test_icmp(struct pf_rule **, struct pf_state **, int, struct pfi_kif *, struct mbuf *, int, void *, struct pf_pdesc *, struct pf_rule **, struct pf_ruleset **, struct ifqueue *); int pf_test_other(struct pf_rule **, struct pf_state **, int, struct pfi_kif *, struct mbuf *, int, void *, struct pf_pdesc *, struct pf_rule **, struct pf_ruleset **, struct ifqueue *); int pf_test_fragment(struct pf_rule **, int, struct pfi_kif *, struct mbuf *, void *, struct pf_pdesc *, struct pf_rule **, struct pf_ruleset **); int pf_tcp_track_full(struct pf_state_peer *, struct pf_state_peer *, struct pf_state **, struct pfi_kif *, struct mbuf *, int, struct pf_pdesc *, u_short *, int *); int pf_tcp_track_sloppy(struct pf_state_peer *, struct pf_state_peer *, struct pf_state **, struct pf_pdesc *, u_short *); int pf_test_state_tcp(struct pf_state **, int, struct pfi_kif *, struct mbuf *, int, void *, struct pf_pdesc *, u_short *); int pf_test_state_udp(struct pf_state **, int, struct pfi_kif *, struct mbuf *, int, void *, struct pf_pdesc *); int pf_test_state_icmp(struct pf_state **, int, struct pfi_kif *, struct mbuf *, int, void *, struct pf_pdesc *, u_short *); int pf_test_state_other(struct pf_state **, int, struct pfi_kif *, struct pf_pdesc *); int pf_match_tag(struct mbuf *, struct pf_rule *, struct pf_mtag *, int *); int pf_step_out_of_anchor(int *, struct pf_ruleset **, int, struct pf_rule **, struct pf_rule **, int *); void pf_hash(struct pf_addr *, struct pf_addr *, struct pf_poolhashkey *, sa_family_t); int pf_map_addr(u_int8_t, struct pf_rule *, struct pf_addr *, struct pf_addr *, struct pf_addr *, struct pf_src_node **); int pf_get_sport(sa_family_t, u_int8_t, struct pf_rule *, struct pf_addr *, struct pf_addr *, u_int16_t, struct pf_addr *, u_int16_t*, u_int16_t, u_int16_t, struct pf_src_node **); void pf_route(struct mbuf **, struct pf_rule *, int, struct ifnet *, struct pf_state *, struct pf_pdesc *); void pf_route6(struct mbuf **, struct pf_rule *, int, struct ifnet *, struct pf_state *, struct pf_pdesc *); #ifdef __FreeBSD__ /* XXX: import */ #else int pf_socket_lookup(int, struct pf_pdesc *); #endif u_int8_t pf_get_wscale(struct mbuf *, int, u_int16_t, sa_family_t); u_int16_t pf_get_mss(struct mbuf *, int, u_int16_t, sa_family_t); u_int16_t pf_calc_mss(struct pf_addr *, sa_family_t, u_int16_t); void pf_set_rt_ifp(struct pf_state *, struct pf_addr *); int pf_check_proto_cksum(struct mbuf *, int, int, u_int8_t, sa_family_t); int pf_addr_wrap_neq(struct pf_addr_wrap *, struct pf_addr_wrap *); struct pf_state *pf_find_state_recurse(struct pfi_kif *, struct pf_state_cmp *, u_int8_t); int pf_src_connlimit(struct pf_state **); int pf_check_congestion(struct ifqueue *); #ifdef __FreeBSD__ int in4_cksum(struct mbuf *m, u_int8_t nxt, int off, int len); extern int pf_end_threads; struct pf_pool_limit pf_pool_limits[PF_LIMIT_MAX]; #else extern struct pool pfr_ktable_pl; extern struct pool pfr_kentry_pl; struct pf_pool_limit pf_pool_limits[PF_LIMIT_MAX] = { { &pf_state_pl, PFSTATE_HIWAT }, { &pf_src_tree_pl, PFSNODE_HIWAT }, { &pf_frent_pl, PFFRAG_FRENT_HIWAT }, { &pfr_ktable_pl, PFR_KTABLE_HIWAT }, { &pfr_kentry_pl, PFR_KENTRY_HIWAT } }; #endif #define STATE_LOOKUP() \ do { \ if (direction == PF_IN) \ *state = pf_find_state_recurse( \ kif, &key, PF_EXT_GWY); \ else \ *state = pf_find_state_recurse( \ kif, &key, PF_LAN_EXT); \ if (*state == NULL || (*state)->timeout == PFTM_PURGE) \ return (PF_DROP); \ if (direction == PF_OUT && \ (((*state)->rule.ptr->rt == PF_ROUTETO && \ (*state)->rule.ptr->direction == PF_OUT) || \ ((*state)->rule.ptr->rt == PF_REPLYTO && \ (*state)->rule.ptr->direction == PF_IN)) && \ (*state)->rt_kif != NULL && \ (*state)->rt_kif != kif) \ return (PF_PASS); \ } while (0) #define STATE_TRANSLATE(s) \ (s)->lan.addr.addr32[0] != (s)->gwy.addr.addr32[0] || \ ((s)->af == AF_INET6 && \ ((s)->lan.addr.addr32[1] != (s)->gwy.addr.addr32[1] || \ (s)->lan.addr.addr32[2] != (s)->gwy.addr.addr32[2] || \ (s)->lan.addr.addr32[3] != (s)->gwy.addr.addr32[3])) || \ (s)->lan.port != (s)->gwy.port #define BOUND_IFACE(r, k) \ ((r)->rule_flag & PFRULE_IFBOUND) ? (k) : pfi_all #define STATE_INC_COUNTERS(s) \ do { \ s->rule.ptr->states++; \ if (s->anchor.ptr != NULL) \ s->anchor.ptr->states++; \ if (s->nat_rule.ptr != NULL) \ s->nat_rule.ptr->states++; \ } while (0) #define STATE_DEC_COUNTERS(s) \ do { \ if (s->nat_rule.ptr != NULL) \ s->nat_rule.ptr->states--; \ if (s->anchor.ptr != NULL) \ s->anchor.ptr->states--; \ s->rule.ptr->states--; \ } while (0) struct pf_src_tree tree_src_tracking; struct pf_state_tree_id tree_id; struct pf_state_queue state_list; #ifdef __FreeBSD__ static int pf_src_compare(struct pf_src_node *, struct pf_src_node *); static int pf_state_compare_lan_ext(struct pf_state *, struct pf_state *); static int pf_state_compare_ext_gwy(struct pf_state *, struct pf_state *); static int pf_state_compare_id(struct pf_state *, struct pf_state *); #endif RB_GENERATE(pf_src_tree, pf_src_node, entry, pf_src_compare); RB_GENERATE(pf_state_tree_lan_ext, pf_state, u.s.entry_lan_ext, pf_state_compare_lan_ext); RB_GENERATE(pf_state_tree_ext_gwy, pf_state, u.s.entry_ext_gwy, pf_state_compare_ext_gwy); RB_GENERATE(pf_state_tree_id, pf_state, u.s.entry_id, pf_state_compare_id); #ifdef __FreeBSD__ static int #else static __inline int #endif pf_src_compare(struct pf_src_node *a, struct pf_src_node *b) { int diff; if (a->rule.ptr > b->rule.ptr) return (1); if (a->rule.ptr < b->rule.ptr) return (-1); if ((diff = a->af - b->af) != 0) return (diff); switch (a->af) { #ifdef INET case AF_INET: if (a->addr.addr32[0] > b->addr.addr32[0]) return (1); if (a->addr.addr32[0] < b->addr.addr32[0]) return (-1); break; #endif /* INET */ #ifdef INET6 case AF_INET6: if (a->addr.addr32[3] > b->addr.addr32[3]) return (1); if (a->addr.addr32[3] < b->addr.addr32[3]) return (-1); if (a->addr.addr32[2] > b->addr.addr32[2]) return (1); if (a->addr.addr32[2] < b->addr.addr32[2]) return (-1); if (a->addr.addr32[1] > b->addr.addr32[1]) return (1); if (a->addr.addr32[1] < b->addr.addr32[1]) return (-1); if (a->addr.addr32[0] > b->addr.addr32[0]) return (1); if (a->addr.addr32[0] < b->addr.addr32[0]) return (-1); break; #endif /* INET6 */ } return (0); } #ifdef __FreeBSD__ static int #else static __inline int #endif pf_state_compare_lan_ext(struct pf_state *a, struct pf_state *b) { int diff; if ((diff = a->proto - b->proto) != 0) return (diff); if ((diff = a->af - b->af) != 0) return (diff); switch (a->af) { #ifdef INET case AF_INET: if (a->lan.addr.addr32[0] > b->lan.addr.addr32[0]) return (1); if (a->lan.addr.addr32[0] < b->lan.addr.addr32[0]) return (-1); if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0]) return (1); if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0]) return (-1); break; #endif /* INET */ #ifdef INET6 case AF_INET6: if (a->lan.addr.addr32[3] > b->lan.addr.addr32[3]) return (1); if (a->lan.addr.addr32[3] < b->lan.addr.addr32[3]) return (-1); if (a->ext.addr.addr32[3] > b->ext.addr.addr32[3]) return (1); if (a->ext.addr.addr32[3] < b->ext.addr.addr32[3]) return (-1); if (a->lan.addr.addr32[2] > b->lan.addr.addr32[2]) return (1); if (a->lan.addr.addr32[2] < b->lan.addr.addr32[2]) return (-1); if (a->ext.addr.addr32[2] > b->ext.addr.addr32[2]) return (1); if (a->ext.addr.addr32[2] < b->ext.addr.addr32[2]) return (-1); if (a->lan.addr.addr32[1] > b->lan.addr.addr32[1]) return (1); if (a->lan.addr.addr32[1] < b->lan.addr.addr32[1]) return (-1); if (a->ext.addr.addr32[1] > b->ext.addr.addr32[1]) return (1); if (a->ext.addr.addr32[1] < b->ext.addr.addr32[1]) return (-1); if (a->lan.addr.addr32[0] > b->lan.addr.addr32[0]) return (1); if (a->lan.addr.addr32[0] < b->lan.addr.addr32[0]) return (-1); if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0]) return (1); if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0]) return (-1); break; #endif /* INET6 */ } if ((diff = a->lan.port - b->lan.port) != 0) return (diff); if ((diff = a->ext.port - b->ext.port) != 0) return (diff); return (0); } #ifdef __FreeBSD__ static int #else static __inline int #endif pf_state_compare_ext_gwy(struct pf_state *a, struct pf_state *b) { int diff; if ((diff = a->proto - b->proto) != 0) return (diff); if ((diff = a->af - b->af) != 0) return (diff); switch (a->af) { #ifdef INET case AF_INET: if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0]) return (1); if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0]) return (-1); if (a->gwy.addr.addr32[0] > b->gwy.addr.addr32[0]) return (1); if (a->gwy.addr.addr32[0] < b->gwy.addr.addr32[0]) return (-1); break; #endif /* INET */ #ifdef INET6 case AF_INET6: if (a->ext.addr.addr32[3] > b->ext.addr.addr32[3]) return (1); if (a->ext.addr.addr32[3] < b->ext.addr.addr32[3]) return (-1); if (a->gwy.addr.addr32[3] > b->gwy.addr.addr32[3]) return (1); if (a->gwy.addr.addr32[3] < b->gwy.addr.addr32[3]) return (-1); if (a->ext.addr.addr32[2] > b->ext.addr.addr32[2]) return (1); if (a->ext.addr.addr32[2] < b->ext.addr.addr32[2]) return (-1); if (a->gwy.addr.addr32[2] > b->gwy.addr.addr32[2]) return (1); if (a->gwy.addr.addr32[2] < b->gwy.addr.addr32[2]) return (-1); if (a->ext.addr.addr32[1] > b->ext.addr.addr32[1]) return (1); if (a->ext.addr.addr32[1] < b->ext.addr.addr32[1]) return (-1); if (a->gwy.addr.addr32[1] > b->gwy.addr.addr32[1]) return (1); if (a->gwy.addr.addr32[1] < b->gwy.addr.addr32[1]) return (-1); if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0]) return (1); if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0]) return (-1); if (a->gwy.addr.addr32[0] > b->gwy.addr.addr32[0]) return (1); if (a->gwy.addr.addr32[0] < b->gwy.addr.addr32[0]) return (-1); break; #endif /* INET6 */ } if ((diff = a->ext.port - b->ext.port) != 0) return (diff); if ((diff = a->gwy.port - b->gwy.port) != 0) return (diff); return (0); } #ifdef __FreeBSD__ static int #else static __inline int #endif pf_state_compare_id(struct pf_state *a, struct pf_state *b) { if (a->id > b->id) return (1); if (a->id < b->id) return (-1); if (a->creatorid > b->creatorid) return (1); if (a->creatorid < b->creatorid) return (-1); return (0); } #ifdef INET6 void pf_addrcpy(struct pf_addr *dst, struct pf_addr *src, sa_family_t af) { switch (af) { #ifdef INET case AF_INET: dst->addr32[0] = src->addr32[0]; break; #endif /* INET */ case AF_INET6: dst->addr32[0] = src->addr32[0]; dst->addr32[1] = src->addr32[1]; dst->addr32[2] = src->addr32[2]; dst->addr32[3] = src->addr32[3]; break; } } #endif /* INET6 */ struct pf_state * pf_find_state_byid(struct pf_state_cmp *key) { pf_status.fcounters[FCNT_STATE_SEARCH]++; return (RB_FIND(pf_state_tree_id, &tree_id, (struct pf_state *)key)); } struct pf_state * pf_find_state_recurse(struct pfi_kif *kif, struct pf_state_cmp *key, u_int8_t tree) { struct pf_state *s; pf_status.fcounters[FCNT_STATE_SEARCH]++; switch (tree) { case PF_LAN_EXT: if ((s = RB_FIND(pf_state_tree_lan_ext, &kif->pfik_lan_ext, (struct pf_state *)key)) != NULL) return (s); if ((s = RB_FIND(pf_state_tree_lan_ext, &pfi_all->pfik_lan_ext, (struct pf_state *)key)) != NULL) return (s); return (NULL); case PF_EXT_GWY: if ((s = RB_FIND(pf_state_tree_ext_gwy, &kif->pfik_ext_gwy, (struct pf_state *)key)) != NULL) return (s); if ((s = RB_FIND(pf_state_tree_ext_gwy, &pfi_all->pfik_ext_gwy, (struct pf_state *)key)) != NULL) return (s); return (NULL); default: panic("pf_find_state_recurse"); } } struct pf_state * pf_find_state_all(struct pf_state_cmp *key, u_int8_t tree, int *more) { struct pf_state *s, *ss = NULL; struct pfi_kif *kif; pf_status.fcounters[FCNT_STATE_SEARCH]++; switch (tree) { case PF_LAN_EXT: TAILQ_FOREACH(kif, &pfi_statehead, pfik_w_states) { s = RB_FIND(pf_state_tree_lan_ext, &kif->pfik_lan_ext, (struct pf_state *)key); if (s == NULL) continue; if (more == NULL) return (s); ss = s; (*more)++; } return (ss); case PF_EXT_GWY: TAILQ_FOREACH(kif, &pfi_statehead, pfik_w_states) { s = RB_FIND(pf_state_tree_ext_gwy, &kif->pfik_ext_gwy, (struct pf_state *)key); if (s == NULL) continue; if (more == NULL) return (s); ss = s; (*more)++; } return (ss); default: panic("pf_find_state_all"); } } void pf_init_threshold(struct pf_threshold *threshold, u_int32_t limit, u_int32_t seconds) { threshold->limit = limit * PF_THRESHOLD_MULT; threshold->seconds = seconds; threshold->count = 0; threshold->last = time_second; } void pf_add_threshold(struct pf_threshold *threshold) { u_int32_t t = time_second, diff = t - threshold->last; if (diff >= threshold->seconds) threshold->count = 0; else threshold->count -= threshold->count * diff / threshold->seconds; threshold->count += PF_THRESHOLD_MULT; threshold->last = t; } int pf_check_threshold(struct pf_threshold *threshold) { return (threshold->count > threshold->limit); } int pf_src_connlimit(struct pf_state **state) { struct pf_state *s; int bad = 0; (*state)->src_node->conn++; (*state)->src.tcp_est = 1; pf_add_threshold(&(*state)->src_node->conn_rate); if ((*state)->rule.ptr->max_src_conn && (*state)->rule.ptr->max_src_conn < (*state)->src_node->conn) { pf_status.lcounters[LCNT_SRCCONN]++; bad++; } if ((*state)->rule.ptr->max_src_conn_rate.limit && pf_check_threshold(&(*state)->src_node->conn_rate)) { pf_status.lcounters[LCNT_SRCCONNRATE]++; bad++; } if (!bad) return (0); if ((*state)->rule.ptr->overload_tbl) { struct pfr_addr p; u_int32_t killed = 0; pf_status.lcounters[LCNT_OVERLOAD_TABLE]++; if (pf_status.debug >= PF_DEBUG_MISC) { printf("pf_src_connlimit: blocking address "); pf_print_host(&(*state)->src_node->addr, 0, (*state)->af); } bzero(&p, sizeof(p)); p.pfra_af = (*state)->af; switch ((*state)->af) { #ifdef INET case AF_INET: p.pfra_net = 32; p.pfra_ip4addr = (*state)->src_node->addr.v4; break; #endif /* INET */ #ifdef INET6 case AF_INET6: p.pfra_net = 128; p.pfra_ip6addr = (*state)->src_node->addr.v6; break; #endif /* INET6 */ } pfr_insert_kentry((*state)->rule.ptr->overload_tbl, &p, time_second); /* kill existing states if that's required. */ if ((*state)->rule.ptr->flush) { pf_status.lcounters[LCNT_OVERLOAD_FLUSH]++; RB_FOREACH(s, pf_state_tree_id, &tree_id) { /* * Kill states from this source. (Only those * from the same rule if PF_FLUSH_GLOBAL is not * set) */ if (s->af == (*state)->af && (((*state)->direction == PF_OUT && PF_AEQ(&(*state)->src_node->addr, &s->lan.addr, s->af)) || ((*state)->direction == PF_IN && PF_AEQ(&(*state)->src_node->addr, &s->ext.addr, s->af))) && ((*state)->rule.ptr->flush & PF_FLUSH_GLOBAL || (*state)->rule.ptr == s->rule.ptr)) { s->timeout = PFTM_PURGE; s->src.state = s->dst.state = TCPS_CLOSED; killed++; } } if (pf_status.debug >= PF_DEBUG_MISC) printf(", %u states killed", killed); } if (pf_status.debug >= PF_DEBUG_MISC) printf("\n"); } /* kill this state */ (*state)->timeout = PFTM_PURGE; (*state)->src.state = (*state)->dst.state = TCPS_CLOSED; return (1); } int pf_insert_src_node(struct pf_src_node **sn, struct pf_rule *rule, struct pf_addr *src, sa_family_t af) { struct pf_src_node k; if (*sn == NULL) { k.af = af; PF_ACPY(&k.addr, src, af); if (rule->rule_flag & PFRULE_RULESRCTRACK || rule->rpool.opts & PF_POOL_STICKYADDR) k.rule.ptr = rule; else k.rule.ptr = NULL; pf_status.scounters[SCNT_SRC_NODE_SEARCH]++; *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k); } if (*sn == NULL) { if (!rule->max_src_nodes || rule->src_nodes < rule->max_src_nodes) (*sn) = pool_get(&pf_src_tree_pl, PR_NOWAIT); else pf_status.lcounters[LCNT_SRCNODES]++; if ((*sn) == NULL) return (-1); bzero(*sn, sizeof(struct pf_src_node)); pf_init_threshold(&(*sn)->conn_rate, rule->max_src_conn_rate.limit, rule->max_src_conn_rate.seconds); (*sn)->af = af; if (rule->rule_flag & PFRULE_RULESRCTRACK || rule->rpool.opts & PF_POOL_STICKYADDR) (*sn)->rule.ptr = rule; else (*sn)->rule.ptr = NULL; PF_ACPY(&(*sn)->addr, src, af); if (RB_INSERT(pf_src_tree, &tree_src_tracking, *sn) != NULL) { if (pf_status.debug >= PF_DEBUG_MISC) { printf("pf: src_tree insert failed: "); pf_print_host(&(*sn)->addr, 0, af); printf("\n"); } pool_put(&pf_src_tree_pl, *sn); return (-1); } (*sn)->creation = time_second; (*sn)->ruletype = rule->action; if ((*sn)->rule.ptr != NULL) (*sn)->rule.ptr->src_nodes++; pf_status.scounters[SCNT_SRC_NODE_INSERT]++; pf_status.src_nodes++; } else { if (rule->max_src_states && (*sn)->states >= rule->max_src_states) { pf_status.lcounters[LCNT_SRCSTATES]++; return (-1); } } return (0); } int pf_insert_state(struct pfi_kif *kif, struct pf_state *state) { /* Thou MUST NOT insert multiple duplicate keys */ state->u.s.kif = kif; if (RB_INSERT(pf_state_tree_lan_ext, &kif->pfik_lan_ext, state)) { if (pf_status.debug >= PF_DEBUG_MISC) { printf("pf: state insert failed: tree_lan_ext"); printf(" lan: "); pf_print_host(&state->lan.addr, state->lan.port, state->af); printf(" gwy: "); pf_print_host(&state->gwy.addr, state->gwy.port, state->af); printf(" ext: "); pf_print_host(&state->ext.addr, state->ext.port, state->af); if (state->sync_flags & PFSTATE_FROMSYNC) printf(" (from sync)"); printf("\n"); } return (-1); } if (RB_INSERT(pf_state_tree_ext_gwy, &kif->pfik_ext_gwy, state)) { if (pf_status.debug >= PF_DEBUG_MISC) { printf("pf: state insert failed: tree_ext_gwy"); printf(" lan: "); pf_print_host(&state->lan.addr, state->lan.port, state->af); printf(" gwy: "); pf_print_host(&state->gwy.addr, state->gwy.port, state->af); printf(" ext: "); pf_print_host(&state->ext.addr, state->ext.port, state->af); if (state->sync_flags & PFSTATE_FROMSYNC) printf(" (from sync)"); printf("\n"); } RB_REMOVE(pf_state_tree_lan_ext, &kif->pfik_lan_ext, state); return (-1); } if (state->id == 0 && state->creatorid == 0) { state->id = htobe64(pf_status.stateid++); state->creatorid = pf_status.hostid; } if (RB_INSERT(pf_state_tree_id, &tree_id, state) != NULL) { if (pf_status.debug >= PF_DEBUG_MISC) { #ifdef __FreeBSD__ printf("pf: state insert failed: " "id: %016llx creatorid: %08x", (long long)be64toh(state->id), ntohl(state->creatorid)); #else printf("pf: state insert failed: " "id: %016llx creatorid: %08x", betoh64(state->id), ntohl(state->creatorid)); #endif if (state->sync_flags & PFSTATE_FROMSYNC) printf(" (from sync)"); printf("\n"); } RB_REMOVE(pf_state_tree_lan_ext, &kif->pfik_lan_ext, state); RB_REMOVE(pf_state_tree_ext_gwy, &kif->pfik_ext_gwy, state); return (-1); } TAILQ_INSERT_TAIL(&state_list, state, u.s.entry_list); pf_status.fcounters[FCNT_STATE_INSERT]++; pf_status.states++; pfi_kif_ref(kif, PFI_KIF_REF_STATE); #if NPFSYNC pfsync_insert_state(state); #endif return (0); } void pf_purge_thread(void *v) { int nloops = 0, s; #ifdef __FreeBSD__ int locked; #endif for (;;) { tsleep(pf_purge_thread, PWAIT, "pftm", 1 * hz); #ifdef __FreeBSD__ sx_slock(&pf_consistency_lock); PF_LOCK(); locked = 0; if (pf_end_threads) { PF_UNLOCK(); sx_sunlock(&pf_consistency_lock); sx_xlock(&pf_consistency_lock); PF_LOCK(); pf_purge_expired_states(pf_status.states, 1); pf_purge_expired_fragments(); pf_purge_expired_src_nodes(1); pf_end_threads++; sx_xunlock(&pf_consistency_lock); PF_UNLOCK(); wakeup(pf_purge_thread); kproc_exit(0); } #endif s = splsoftnet(); /* process a fraction of the state table every second */ #ifdef __FreeBSD__ if(!pf_purge_expired_states(1 + (pf_status.states / pf_default_rule.timeout[PFTM_INTERVAL]), 0)) { PF_UNLOCK(); sx_sunlock(&pf_consistency_lock); sx_xlock(&pf_consistency_lock); PF_LOCK(); locked = 1; pf_purge_expired_states(1 + (pf_status.states / pf_default_rule.timeout[PFTM_INTERVAL]), 1); } #else pf_purge_expired_states(1 + (pf_status.states / pf_default_rule.timeout[PFTM_INTERVAL])); #endif /* purge other expired types every PFTM_INTERVAL seconds */ if (++nloops >= pf_default_rule.timeout[PFTM_INTERVAL]) { pf_purge_expired_fragments(); if (!pf_purge_expired_src_nodes(locked)) { PF_UNLOCK(); sx_sunlock(&pf_consistency_lock); sx_xlock(&pf_consistency_lock); PF_LOCK(); locked = 1; pf_purge_expired_src_nodes(1); } nloops = 0; } splx(s); #ifdef __FreeBSD__ PF_UNLOCK(); if (locked) sx_xunlock(&pf_consistency_lock); else sx_sunlock(&pf_consistency_lock); #endif } } u_int32_t pf_state_expires(const struct pf_state *state) { u_int32_t timeout; u_int32_t start; u_int32_t end; u_int32_t states; /* handle all PFTM_* > PFTM_MAX here */ if (state->timeout == PFTM_PURGE) return (time_second); if (state->timeout == PFTM_UNTIL_PACKET) return (0); #ifdef __FreeBSD__ KASSERT(state->timeout != PFTM_UNLINKED, ("pf_state_expires: timeout == PFTM_UNLINKED")); KASSERT((state->timeout < PFTM_MAX), ("pf_state_expires: timeout > PFTM_MAX")); #else KASSERT(state->timeout != PFTM_UNLINKED); KASSERT(state->timeout < PFTM_MAX); #endif timeout = state->rule.ptr->timeout[state->timeout]; if (!timeout) timeout = pf_default_rule.timeout[state->timeout]; start = state->rule.ptr->timeout[PFTM_ADAPTIVE_START]; if (start) { end = state->rule.ptr->timeout[PFTM_ADAPTIVE_END]; states = state->rule.ptr->states; } else { start = pf_default_rule.timeout[PFTM_ADAPTIVE_START]; end = pf_default_rule.timeout[PFTM_ADAPTIVE_END]; states = pf_status.states; } if (end && states > start && start < end) { if (states < end) return (state->expire + timeout * (end - states) / (end - start)); else return (time_second); } return (state->expire + timeout); } #ifdef __FreeBSD__ int pf_purge_expired_src_nodes(int waslocked) #else void pf_purge_expired_src_nodes(int waslocked) #endif { struct pf_src_node *cur, *next; int locked = waslocked; for (cur = RB_MIN(pf_src_tree, &tree_src_tracking); cur; cur = next) { next = RB_NEXT(pf_src_tree, &tree_src_tracking, cur); if (cur->states <= 0 && cur->expire <= time_second) { if (! locked) { #ifdef __FreeBSD__ if (!sx_try_upgrade(&pf_consistency_lock)) return (0); #else rw_enter_write(&pf_consistency_lock); #endif next = RB_NEXT(pf_src_tree, &tree_src_tracking, cur); locked = 1; } if (cur->rule.ptr != NULL) { cur->rule.ptr->src_nodes--; if (cur->rule.ptr->states <= 0 && cur->rule.ptr->max_src_nodes <= 0) pf_rm_rule(NULL, cur->rule.ptr); } RB_REMOVE(pf_src_tree, &tree_src_tracking, cur); pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; pf_status.src_nodes--; pool_put(&pf_src_tree_pl, cur); } } if (locked && !waslocked) #ifdef __FreeBSD__ sx_downgrade(&pf_consistency_lock); #else rw_exit_write(&pf_consistency_lock); #endif #ifdef __FreeBSD__ return (1); #endif } void pf_src_tree_remove_state(struct pf_state *s) { u_int32_t timeout; if (s->src_node != NULL) { if (s->proto == IPPROTO_TCP) { if (s->src.tcp_est) --s->src_node->conn; } if (--s->src_node->states <= 0) { timeout = s->rule.ptr->timeout[PFTM_SRC_NODE]; if (!timeout) timeout = pf_default_rule.timeout[PFTM_SRC_NODE]; s->src_node->expire = time_second + timeout; } } if (s->nat_src_node != s->src_node && s->nat_src_node != NULL) { if (--s->nat_src_node->states <= 0) { timeout = s->rule.ptr->timeout[PFTM_SRC_NODE]; if (!timeout) timeout = pf_default_rule.timeout[PFTM_SRC_NODE]; s->nat_src_node->expire = time_second + timeout; } } s->src_node = s->nat_src_node = NULL; } /* callers should be at splsoftnet */ void pf_unlink_state(struct pf_state *cur) { #ifdef __FreeBSD__ if (cur->local_flags & PFSTATE_EXPIRING) return; cur->local_flags |= PFSTATE_EXPIRING; #endif if (cur->src.state == PF_TCPS_PROXY_DST) { #ifdef __FreeBSD__ pf_send_tcp(NULL, cur->rule.ptr, cur->af, #else pf_send_tcp(cur->rule.ptr, cur->af, #endif &cur->ext.addr, &cur->lan.addr, cur->ext.port, cur->lan.port, cur->src.seqhi, cur->src.seqlo + 1, TH_RST|TH_ACK, 0, 0, 0, 1, cur->tag, NULL, NULL); } RB_REMOVE(pf_state_tree_ext_gwy, &cur->u.s.kif->pfik_ext_gwy, cur); RB_REMOVE(pf_state_tree_lan_ext, &cur->u.s.kif->pfik_lan_ext, cur); RB_REMOVE(pf_state_tree_id, &tree_id, cur); #if NPFSYNC if (cur->creatorid == pf_status.hostid) pfsync_delete_state(cur); #endif cur->timeout = PFTM_UNLINKED; pf_src_tree_remove_state(cur); } /* callers should be at splsoftnet and hold the * write_lock on pf_consistency_lock */ void pf_free_state(struct pf_state *cur) { #if NPFSYNC if (pfsyncif != NULL && (pfsyncif->sc_bulk_send_next == cur || pfsyncif->sc_bulk_terminator == cur)) return; #endif #ifdef __FreeBSD__ KASSERT(cur->timeout == PFTM_UNLINKED, ("pf_free_state: cur->timeout != PFTM_UNLINKED")); #else KASSERT(cur->timeout == PFTM_UNLINKED); #endif if (--cur->rule.ptr->states <= 0 && cur->rule.ptr->src_nodes <= 0) pf_rm_rule(NULL, cur->rule.ptr); if (cur->nat_rule.ptr != NULL) if (--cur->nat_rule.ptr->states <= 0 && cur->nat_rule.ptr->src_nodes <= 0) pf_rm_rule(NULL, cur->nat_rule.ptr); if (cur->anchor.ptr != NULL) if (--cur->anchor.ptr->states <= 0) pf_rm_rule(NULL, cur->anchor.ptr); pf_normalize_tcp_cleanup(cur); pfi_kif_unref(cur->u.s.kif, PFI_KIF_REF_STATE); TAILQ_REMOVE(&state_list, cur, u.s.entry_list); if (cur->tag) pf_tag_unref(cur->tag); pool_put(&pf_state_pl, cur); pf_status.fcounters[FCNT_STATE_REMOVALS]++; pf_status.states--; } #ifdef __FreeBSD__ int pf_purge_expired_states(u_int32_t maxcheck, int waslocked) #else void pf_purge_expired_states(u_int32_t maxcheck) #endif { static struct pf_state *cur = NULL; struct pf_state *next; #ifdef __FreeBSD__ int locked = waslocked; #else int locked = 0; #endif while (maxcheck--) { /* wrap to start of list when we hit the end */ if (cur == NULL) { cur = TAILQ_FIRST(&state_list); if (cur == NULL) break; /* list empty */ } /* get next state, as cur may get deleted */ next = TAILQ_NEXT(cur, u.s.entry_list); if (cur->timeout == PFTM_UNLINKED) { /* free unlinked state */ if (! locked) { #ifdef __FreeBSD__ if (!sx_try_upgrade(&pf_consistency_lock)) return (0); #else rw_enter_write(&pf_consistency_lock); #endif locked = 1; } pf_free_state(cur); } else if (pf_state_expires(cur) <= time_second) { /* unlink and free expired state */ pf_unlink_state(cur); if (! locked) { #ifdef __FreeBSD__ if (!sx_try_upgrade(&pf_consistency_lock)) return (0); #else rw_enter_write(&pf_consistency_lock); #endif locked = 1; } pf_free_state(cur); } cur = next; } #ifdef __FreeBSD__ if (!waslocked && locked) sx_downgrade(&pf_consistency_lock); return (1); #else if (locked) rw_exit_write(&pf_consistency_lock); #endif } int pf_tbladdr_setup(struct pf_ruleset *rs, struct pf_addr_wrap *aw) { if (aw->type != PF_ADDR_TABLE) return (0); if ((aw->p.tbl = pfr_attach_table(rs, aw->v.tblname)) == NULL) return (1); return (0); } void pf_tbladdr_remove(struct pf_addr_wrap *aw) { if (aw->type != PF_ADDR_TABLE || aw->p.tbl == NULL) return; pfr_detach_table(aw->p.tbl); aw->p.tbl = NULL; } void pf_tbladdr_copyout(struct pf_addr_wrap *aw) { struct pfr_ktable *kt = aw->p.tbl; if (aw->type != PF_ADDR_TABLE || kt == NULL) return; if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL) kt = kt->pfrkt_root; aw->p.tbl = NULL; aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ? kt->pfrkt_cnt : -1; } void pf_print_host(struct pf_addr *addr, u_int16_t p, sa_family_t af) { switch (af) { #ifdef INET case AF_INET: { u_int32_t a = ntohl(addr->addr32[0]); printf("%u.%u.%u.%u", (a>>24)&255, (a>>16)&255, (a>>8)&255, a&255); if (p) { p = ntohs(p); printf(":%u", p); } break; } #endif /* INET */ #ifdef INET6 case AF_INET6: { u_int16_t b; u_int8_t i, curstart = 255, curend = 0, maxstart = 0, maxend = 0; for (i = 0; i < 8; i++) { if (!addr->addr16[i]) { if (curstart == 255) curstart = i; else curend = i; } else { if (curstart) { if ((curend - curstart) > (maxend - maxstart)) { maxstart = curstart; maxend = curend; curstart = 255; } } } } for (i = 0; i < 8; i++) { if (i >= maxstart && i <= maxend) { if (maxend != 7) { if (i == maxstart) printf(":"); } else { if (i == maxend) printf(":"); } } else { b = ntohs(addr->addr16[i]); printf("%x", b); if (i < 7) printf(":"); } } if (p) { p = ntohs(p); printf("[%u]", p); } break; } #endif /* INET6 */ } } void pf_print_state(struct pf_state *s) { switch (s->proto) { case IPPROTO_TCP: printf("TCP "); break; case IPPROTO_UDP: printf("UDP "); break; case IPPROTO_ICMP: printf("ICMP "); break; case IPPROTO_ICMPV6: printf("ICMPV6 "); break; default: printf("%u ", s->proto); break; } pf_print_host(&s->lan.addr, s->lan.port, s->af); printf(" "); pf_print_host(&s->gwy.addr, s->gwy.port, s->af); printf(" "); pf_print_host(&s->ext.addr, s->ext.port, s->af); printf(" [lo=%u high=%u win=%u modulator=%u", s->src.seqlo, s->src.seqhi, s->src.max_win, s->src.seqdiff); if (s->src.wscale && s->dst.wscale) printf(" wscale=%u", s->src.wscale & PF_WSCALE_MASK); printf("]"); printf(" [lo=%u high=%u win=%u modulator=%u", s->dst.seqlo, s->dst.seqhi, s->dst.max_win, s->dst.seqdiff); if (s->src.wscale && s->dst.wscale) printf(" wscale=%u", s->dst.wscale & PF_WSCALE_MASK); printf("]"); printf(" %u:%u", s->src.state, s->dst.state); } void pf_print_flags(u_int8_t f) { if (f) printf(" "); if (f & TH_FIN) printf("F"); if (f & TH_SYN) printf("S"); if (f & TH_RST) printf("R"); if (f & TH_PUSH) printf("P"); if (f & TH_ACK) printf("A"); if (f & TH_URG) printf("U"); if (f & TH_ECE) printf("E"); if (f & TH_CWR) printf("W"); } #define PF_SET_SKIP_STEPS(i) \ do { \ while (head[i] != cur) { \ head[i]->skip[i].ptr = cur; \ head[i] = TAILQ_NEXT(head[i], entries); \ } \ } while (0) void pf_calc_skip_steps(struct pf_rulequeue *rules) { struct pf_rule *cur, *prev, *head[PF_SKIP_COUNT]; int i; cur = TAILQ_FIRST(rules); prev = cur; for (i = 0; i < PF_SKIP_COUNT; ++i) head[i] = cur; while (cur != NULL) { if (cur->kif != prev->kif || cur->ifnot != prev->ifnot) PF_SET_SKIP_STEPS(PF_SKIP_IFP); if (cur->direction != prev->direction) PF_SET_SKIP_STEPS(PF_SKIP_DIR); if (cur->af != prev->af) PF_SET_SKIP_STEPS(PF_SKIP_AF); if (cur->proto != prev->proto) PF_SET_SKIP_STEPS(PF_SKIP_PROTO); if (cur->src.neg != prev->src.neg || pf_addr_wrap_neq(&cur->src.addr, &prev->src.addr)) PF_SET_SKIP_STEPS(PF_SKIP_SRC_ADDR); if (cur->src.port[0] != prev->src.port[0] || cur->src.port[1] != prev->src.port[1] || cur->src.port_op != prev->src.port_op) PF_SET_SKIP_STEPS(PF_SKIP_SRC_PORT); if (cur->dst.neg != prev->dst.neg || pf_addr_wrap_neq(&cur->dst.addr, &prev->dst.addr)) PF_SET_SKIP_STEPS(PF_SKIP_DST_ADDR); if (cur->dst.port[0] != prev->dst.port[0] || cur->dst.port[1] != prev->dst.port[1] || cur->dst.port_op != prev->dst.port_op) PF_SET_SKIP_STEPS(PF_SKIP_DST_PORT); prev = cur; cur = TAILQ_NEXT(cur, entries); } for (i = 0; i < PF_SKIP_COUNT; ++i) PF_SET_SKIP_STEPS(i); } int pf_addr_wrap_neq(struct pf_addr_wrap *aw1, struct pf_addr_wrap *aw2) { if (aw1->type != aw2->type) return (1); switch (aw1->type) { case PF_ADDR_ADDRMASK: if (PF_ANEQ(&aw1->v.a.addr, &aw2->v.a.addr, 0)) return (1); if (PF_ANEQ(&aw1->v.a.mask, &aw2->v.a.mask, 0)) return (1); return (0); case PF_ADDR_DYNIFTL: return (aw1->p.dyn->pfid_kt != aw2->p.dyn->pfid_kt); case PF_ADDR_NOROUTE: case PF_ADDR_URPFFAILED: return (0); case PF_ADDR_TABLE: return (aw1->p.tbl != aw2->p.tbl); case PF_ADDR_RTLABEL: return (aw1->v.rtlabel != aw2->v.rtlabel); default: printf("invalid address type: %d\n", aw1->type); return (1); } } u_int16_t pf_cksum_fixup(u_int16_t cksum, u_int16_t old, u_int16_t new, u_int8_t udp) { u_int32_t l; if (udp && !cksum) return (0x0000); l = cksum + old - new; l = (l >> 16) + (l & 65535); l = l & 65535; if (udp && !l) return (0xFFFF); return (l); } void pf_change_ap(struct pf_addr *a, u_int16_t *p, u_int16_t *ic, u_int16_t *pc, struct pf_addr *an, u_int16_t pn, u_int8_t u, sa_family_t af) { struct pf_addr ao; u_int16_t po = *p; PF_ACPY(&ao, a, af); PF_ACPY(a, an, af); *p = pn; switch (af) { #ifdef INET case AF_INET: *ic = pf_cksum_fixup(pf_cksum_fixup(*ic, ao.addr16[0], an->addr16[0], 0), ao.addr16[1], an->addr16[1], 0); *p = pn; *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc, ao.addr16[0], an->addr16[0], u), ao.addr16[1], an->addr16[1], u), po, pn, u); break; #endif /* INET */ #ifdef INET6 case AF_INET6: *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc, ao.addr16[0], an->addr16[0], u), ao.addr16[1], an->addr16[1], u), ao.addr16[2], an->addr16[2], u), ao.addr16[3], an->addr16[3], u), ao.addr16[4], an->addr16[4], u), ao.addr16[5], an->addr16[5], u), ao.addr16[6], an->addr16[6], u), ao.addr16[7], an->addr16[7], u), po, pn, u); break; #endif /* INET6 */ } } /* Changes a u_int32_t. Uses a void * so there are no align restrictions */ void pf_change_a(void *a, u_int16_t *c, u_int32_t an, u_int8_t u) { u_int32_t ao; memcpy(&ao, a, sizeof(ao)); memcpy(a, &an, sizeof(u_int32_t)); *c = pf_cksum_fixup(pf_cksum_fixup(*c, ao / 65536, an / 65536, u), ao % 65536, an % 65536, u); } #ifdef INET6 void pf_change_a6(struct pf_addr *a, u_int16_t *c, struct pf_addr *an, u_int8_t u) { struct pf_addr ao; PF_ACPY(&ao, a, AF_INET6); PF_ACPY(a, an, AF_INET6); *c = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( pf_cksum_fixup(pf_cksum_fixup(*c, ao.addr16[0], an->addr16[0], u), ao.addr16[1], an->addr16[1], u), ao.addr16[2], an->addr16[2], u), ao.addr16[3], an->addr16[3], u), ao.addr16[4], an->addr16[4], u), ao.addr16[5], an->addr16[5], u), ao.addr16[6], an->addr16[6], u), ao.addr16[7], an->addr16[7], u); } #endif /* INET6 */ void pf_change_icmp(struct pf_addr *ia, u_int16_t *ip, struct pf_addr *oa, struct pf_addr *na, u_int16_t np, u_int16_t *pc, u_int16_t *h2c, u_int16_t *ic, u_int16_t *hc, u_int8_t u, sa_family_t af) { struct pf_addr oia, ooa; PF_ACPY(&oia, ia, af); PF_ACPY(&ooa, oa, af); /* Change inner protocol port, fix inner protocol checksum. */ if (ip != NULL) { u_int16_t oip = *ip; u_int32_t opc = 0; /* make the compiler happy */ if (pc != NULL) opc = *pc; *ip = np; if (pc != NULL) *pc = pf_cksum_fixup(*pc, oip, *ip, u); *ic = pf_cksum_fixup(*ic, oip, *ip, 0); if (pc != NULL) *ic = pf_cksum_fixup(*ic, opc, *pc, 0); } /* Change inner ip address, fix inner ip and icmp checksums. */ PF_ACPY(ia, na, af); switch (af) { #ifdef INET case AF_INET: { u_int32_t oh2c = *h2c; *h2c = pf_cksum_fixup(pf_cksum_fixup(*h2c, oia.addr16[0], ia->addr16[0], 0), oia.addr16[1], ia->addr16[1], 0); *ic = pf_cksum_fixup(pf_cksum_fixup(*ic, oia.addr16[0], ia->addr16[0], 0), oia.addr16[1], ia->addr16[1], 0); *ic = pf_cksum_fixup(*ic, oh2c, *h2c, 0); break; } #endif /* INET */ #ifdef INET6 case AF_INET6: *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( pf_cksum_fixup(pf_cksum_fixup(*ic, oia.addr16[0], ia->addr16[0], u), oia.addr16[1], ia->addr16[1], u), oia.addr16[2], ia->addr16[2], u), oia.addr16[3], ia->addr16[3], u), oia.addr16[4], ia->addr16[4], u), oia.addr16[5], ia->addr16[5], u), oia.addr16[6], ia->addr16[6], u), oia.addr16[7], ia->addr16[7], u); break; #endif /* INET6 */ } /* Change outer ip address, fix outer ip or icmpv6 checksum. */ PF_ACPY(oa, na, af); switch (af) { #ifdef INET case AF_INET: *hc = pf_cksum_fixup(pf_cksum_fixup(*hc, ooa.addr16[0], oa->addr16[0], 0), ooa.addr16[1], oa->addr16[1], 0); break; #endif /* INET */ #ifdef INET6 case AF_INET6: *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( pf_cksum_fixup(pf_cksum_fixup(*ic, ooa.addr16[0], oa->addr16[0], u), ooa.addr16[1], oa->addr16[1], u), ooa.addr16[2], oa->addr16[2], u), ooa.addr16[3], oa->addr16[3], u), ooa.addr16[4], oa->addr16[4], u), ooa.addr16[5], oa->addr16[5], u), ooa.addr16[6], oa->addr16[6], u), ooa.addr16[7], oa->addr16[7], u); break; #endif /* INET6 */ } } /* * Need to modulate the sequence numbers in the TCP SACK option * (credits to Krzysztof Pfaff for report and patch) */ int pf_modulate_sack(struct mbuf *m, int off, struct pf_pdesc *pd, struct tcphdr *th, struct pf_state_peer *dst) { int hlen = (th->th_off << 2) - sizeof(*th), thoptlen = hlen; #ifdef __FreeBSD__ u_int8_t opts[TCP_MAXOLEN], *opt = opts; #else u_int8_t opts[MAX_TCPOPTLEN], *opt = opts; #endif int copyback = 0, i, olen; struct sackblk sack; #define TCPOLEN_SACKLEN (TCPOLEN_SACK + 2) if (hlen < TCPOLEN_SACKLEN || !pf_pull_hdr(m, off + sizeof(*th), opts, hlen, NULL, NULL, pd->af)) return 0; while (hlen >= TCPOLEN_SACKLEN) { olen = opt[1]; switch (*opt) { case TCPOPT_EOL: /* FALLTHROUGH */ case TCPOPT_NOP: opt++; hlen--; break; case TCPOPT_SACK: if (olen > hlen) olen = hlen; if (olen >= TCPOLEN_SACKLEN) { for (i = 2; i + TCPOLEN_SACK <= olen; i += TCPOLEN_SACK) { memcpy(&sack, &opt[i], sizeof(sack)); pf_change_a(&sack.start, &th->th_sum, htonl(ntohl(sack.start) - dst->seqdiff), 0); pf_change_a(&sack.end, &th->th_sum, htonl(ntohl(sack.end) - dst->seqdiff), 0); memcpy(&opt[i], &sack, sizeof(sack)); } copyback = 1; } /* FALLTHROUGH */ default: if (olen < 2) olen = 2; hlen -= olen; opt += olen; } } if (copyback) #ifdef __FreeBSD__ m_copyback(m, off + sizeof(*th), thoptlen, (caddr_t)opts); #else m_copyback(m, off + sizeof(*th), thoptlen, opts); #endif return (copyback); } void #ifdef __FreeBSD__ pf_send_tcp(struct mbuf *replyto, const struct pf_rule *r, sa_family_t af, #else pf_send_tcp(const struct pf_rule *r, sa_family_t af, #endif const struct pf_addr *saddr, const struct pf_addr *daddr, u_int16_t sport, u_int16_t dport, u_int32_t seq, u_int32_t ack, u_int8_t flags, u_int16_t win, u_int16_t mss, u_int8_t ttl, int tag, u_int16_t rtag, struct ether_header *eh, struct ifnet *ifp) { struct mbuf *m; int len, tlen; #ifdef INET struct ip *h; #endif /* INET */ #ifdef INET6 struct ip6_hdr *h6; #endif /* INET6 */ struct tcphdr *th; char *opt; struct pf_mtag *pf_mtag; #ifdef __FreeBSD__ KASSERT( #ifdef INET af == AF_INET #else 0 #endif || #ifdef INET6 af == AF_INET6 #else 0 #endif , ("Unsupported AF %d", af)); len = 0; th = NULL; #ifdef INET h = NULL; #endif #ifdef INET6 h6 = NULL; #endif #endif /* maximum segment size tcp option */ tlen = sizeof(struct tcphdr); if (mss) tlen += 4; switch (af) { #ifdef INET case AF_INET: len = sizeof(struct ip) + tlen; break; #endif /* INET */ #ifdef INET6 case AF_INET6: len = sizeof(struct ip6_hdr) + tlen; break; #endif /* INET6 */ } /* create outgoing mbuf */ m = m_gethdr(M_DONTWAIT, MT_HEADER); if (m == NULL) return; #ifdef __FreeBSD__ #ifdef MAC if (replyto) mac_netinet_firewall_reply(replyto, m); else mac_netinet_firewall_send(m); #else (void)replyto; #endif #endif if ((pf_mtag = pf_get_mtag(m)) == NULL) { m_freem(m); return; } if (tag) #ifdef __FreeBSD__ m->m_flags |= M_SKIP_FIREWALL; #else pf_mtag->flags |= PF_TAG_GENERATED; #endif pf_mtag->tag = rtag; if (r != NULL && r->rtableid >= 0) #ifdef __FreeBSD__ { M_SETFIB(m, r->rtableid); #endif pf_mtag->rtableid = r->rtableid; #ifdef __FreeBSD__ } #endif #ifdef ALTQ if (r != NULL && r->qid) { pf_mtag->qid = r->qid; /* add hints for ecn */ pf_mtag->af = af; pf_mtag->hdr = mtod(m, struct ip *); } #endif /* ALTQ */ m->m_data += max_linkhdr; m->m_pkthdr.len = m->m_len = len; m->m_pkthdr.rcvif = NULL; bzero(m->m_data, len); switch (af) { #ifdef INET case AF_INET: h = mtod(m, struct ip *); /* IP header fields included in the TCP checksum */ h->ip_p = IPPROTO_TCP; h->ip_len = htons(tlen); h->ip_src.s_addr = saddr->v4.s_addr; h->ip_dst.s_addr = daddr->v4.s_addr; th = (struct tcphdr *)((caddr_t)h + sizeof(struct ip)); break; #endif /* INET */ #ifdef INET6 case AF_INET6: h6 = mtod(m, struct ip6_hdr *); /* IP header fields included in the TCP checksum */ h6->ip6_nxt = IPPROTO_TCP; h6->ip6_plen = htons(tlen); memcpy(&h6->ip6_src, &saddr->v6, sizeof(struct in6_addr)); memcpy(&h6->ip6_dst, &daddr->v6, sizeof(struct in6_addr)); th = (struct tcphdr *)((caddr_t)h6 + sizeof(struct ip6_hdr)); break; #endif /* INET6 */ } /* TCP header */ th->th_sport = sport; th->th_dport = dport; th->th_seq = htonl(seq); th->th_ack = htonl(ack); th->th_off = tlen >> 2; th->th_flags = flags; th->th_win = htons(win); if (mss) { opt = (char *)(th + 1); opt[0] = TCPOPT_MAXSEG; opt[1] = 4; HTONS(mss); bcopy((caddr_t)&mss, (caddr_t)(opt + 2), 2); } switch (af) { #ifdef INET case AF_INET: /* TCP checksum */ th->th_sum = in_cksum(m, len); /* Finish the IP header */ h->ip_v = 4; h->ip_hl = sizeof(*h) >> 2; h->ip_tos = IPTOS_LOWDELAY; #ifdef __FreeBSD__ h->ip_off = V_path_mtu_discovery ? IP_DF : 0; h->ip_len = len; #else h->ip_off = htons(ip_mtudisc ? IP_DF : 0); h->ip_len = htons(len); #endif h->ip_ttl = ttl ? ttl : V_ip_defttl; h->ip_sum = 0; if (eh == NULL) { #ifdef __FreeBSD__ PF_UNLOCK(); ip_output(m, (void *)NULL, (void *)NULL, 0, (void *)NULL, (void *)NULL); PF_LOCK(); #else /* ! __FreeBSD__ */ ip_output(m, (void *)NULL, (void *)NULL, 0, (void *)NULL, (void *)NULL); #endif } else { struct route ro; struct rtentry rt; struct ether_header *e = (void *)ro.ro_dst.sa_data; if (ifp == NULL) { m_freem(m); return; } rt.rt_ifp = ifp; ro.ro_rt = &rt; ro.ro_dst.sa_len = sizeof(ro.ro_dst); ro.ro_dst.sa_family = pseudo_AF_HDRCMPLT; bcopy(eh->ether_dhost, e->ether_shost, ETHER_ADDR_LEN); bcopy(eh->ether_shost, e->ether_dhost, ETHER_ADDR_LEN); e->ether_type = eh->ether_type; #ifdef __FreeBSD__ PF_UNLOCK(); /* XXX_IMPORT: later */ ip_output(m, (void *)NULL, &ro, 0, (void *)NULL, (void *)NULL); PF_LOCK(); #else /* ! __FreeBSD__ */ ip_output(m, (void *)NULL, &ro, IP_ROUTETOETHER, (void *)NULL, (void *)NULL); #endif } break; #endif /* INET */ #ifdef INET6 case AF_INET6: /* TCP checksum */ th->th_sum = in6_cksum(m, IPPROTO_TCP, sizeof(struct ip6_hdr), tlen); h6->ip6_vfc |= IPV6_VERSION; h6->ip6_hlim = IPV6_DEFHLIM; #ifdef __FreeBSD__ PF_UNLOCK(); ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL); PF_LOCK(); #else ip6_output(m, NULL, NULL, 0, NULL, NULL); #endif break; #endif /* INET6 */ } } void pf_send_icmp(struct mbuf *m, u_int8_t type, u_int8_t code, sa_family_t af, struct pf_rule *r) { struct pf_mtag *pf_mtag; struct mbuf *m0; #ifdef __FreeBSD__ #ifdef INET struct ip *ip; #endif #endif #ifdef __FreeBSD__ m0 = m_copypacket(m, M_DONTWAIT); if (m0 == NULL) return; #else m0 = m_copy(m, 0, M_COPYALL); #endif if ((pf_mtag = pf_get_mtag(m0)) == NULL) return; #ifdef __FreeBSD__ /* XXX: revisit */ m0->m_flags |= M_SKIP_FIREWALL; #else pf_mtag->flags |= PF_TAG_GENERATED; #endif if (r->rtableid >= 0) #ifdef __FreeBSD__ { M_SETFIB(m0, r->rtableid); #endif pf_mtag->rtableid = r->rtableid; #ifdef __FreeBSD__ } #endif #ifdef ALTQ if (r->qid) { pf_mtag->qid = r->qid; /* add hints for ecn */ pf_mtag->af = af; pf_mtag->hdr = mtod(m0, struct ip *); } #endif /* ALTQ */ switch (af) { #ifdef INET case AF_INET: #ifdef __FreeBSD__ /* icmp_error() expects host byte ordering */ ip = mtod(m0, struct ip *); NTOHS(ip->ip_len); NTOHS(ip->ip_off); PF_UNLOCK(); icmp_error(m0, type, code, 0, 0); PF_LOCK(); #else icmp_error(m0, type, code, 0, 0); #endif break; #endif /* INET */ #ifdef INET6 case AF_INET6: #ifdef __FreeBSD__ PF_UNLOCK(); #endif icmp6_error(m0, type, code, 0); #ifdef __FreeBSD__ PF_LOCK(); #endif break; #endif /* INET6 */ } } /* * Return 1 if the addresses a and b match (with mask m), otherwise return 0. * If n is 0, they match if they are equal. If n is != 0, they match if they * are different. */ int pf_match_addr(u_int8_t n, struct pf_addr *a, struct pf_addr *m, struct pf_addr *b, sa_family_t af) { int match = 0; switch (af) { #ifdef INET case AF_INET: if ((a->addr32[0] & m->addr32[0]) == (b->addr32[0] & m->addr32[0])) match++; break; #endif /* INET */ #ifdef INET6 case AF_INET6: if (((a->addr32[0] & m->addr32[0]) == (b->addr32[0] & m->addr32[0])) && ((a->addr32[1] & m->addr32[1]) == (b->addr32[1] & m->addr32[1])) && ((a->addr32[2] & m->addr32[2]) == (b->addr32[2] & m->addr32[2])) && ((a->addr32[3] & m->addr32[3]) == (b->addr32[3] & m->addr32[3]))) match++; break; #endif /* INET6 */ } if (match) { if (n) return (0); else return (1); } else { if (n) return (1); else return (0); } } int pf_match(u_int8_t op, u_int32_t a1, u_int32_t a2, u_int32_t p) { switch (op) { case PF_OP_IRG: return ((p > a1) && (p < a2)); case PF_OP_XRG: return ((p < a1) || (p > a2)); case PF_OP_RRG: return ((p >= a1) && (p <= a2)); case PF_OP_EQ: return (p == a1); case PF_OP_NE: return (p != a1); case PF_OP_LT: return (p < a1); case PF_OP_LE: return (p <= a1); case PF_OP_GT: return (p > a1); case PF_OP_GE: return (p >= a1); } return (0); /* never reached */ } int pf_match_port(u_int8_t op, u_int16_t a1, u_int16_t a2, u_int16_t p) { NTOHS(a1); NTOHS(a2); NTOHS(p); return (pf_match(op, a1, a2, p)); } int pf_match_uid(u_int8_t op, uid_t a1, uid_t a2, uid_t u) { if (u == UID_MAX && op != PF_OP_EQ && op != PF_OP_NE) return (0); return (pf_match(op, a1, a2, u)); } int pf_match_gid(u_int8_t op, gid_t a1, gid_t a2, gid_t g) { if (g == GID_MAX && op != PF_OP_EQ && op != PF_OP_NE) return (0); return (pf_match(op, a1, a2, g)); } #ifndef __FreeBSD__ struct pf_mtag * pf_find_mtag(struct mbuf *m) { struct m_tag *mtag; if ((mtag = m_tag_find(m, PACKET_TAG_PF, NULL)) == NULL) return (NULL); return ((struct pf_mtag *)(mtag + 1)); } struct pf_mtag * pf_get_mtag(struct mbuf *m) { struct m_tag *mtag; if ((mtag = m_tag_find(m, PACKET_TAG_PF, NULL)) == NULL) { mtag = m_tag_get(PACKET_TAG_PF, sizeof(struct pf_mtag), M_NOWAIT); if (mtag == NULL) return (NULL); bzero(mtag + 1, sizeof(struct pf_mtag)); m_tag_prepend(m, mtag); } return ((struct pf_mtag *)(mtag + 1)); } #endif int pf_match_tag(struct mbuf *m, struct pf_rule *r, struct pf_mtag *pf_mtag, int *tag) { if (*tag == -1) *tag = pf_mtag->tag; return ((!r->match_tag_not && r->match_tag == *tag) || (r->match_tag_not && r->match_tag != *tag)); } int pf_tag_packet(struct mbuf *m, struct pf_mtag *pf_mtag, int tag, int rtableid) { if (tag <= 0 && rtableid < 0) return (0); if (pf_mtag == NULL) if ((pf_mtag = pf_get_mtag(m)) == NULL) return (1); if (tag > 0) pf_mtag->tag = tag; if (rtableid >= 0) #ifdef __FreeBSD__ { M_SETFIB(m, rtableid); #endif pf_mtag->rtableid = rtableid; #ifdef __FreeBSD__ } #endif return (0); } static void pf_step_into_anchor(int *depth, struct pf_ruleset **rs, int n, struct pf_rule **r, struct pf_rule **a, int *match) { struct pf_anchor_stackframe *f; (*r)->anchor->match = 0; if (match) *match = 0; if (*depth >= sizeof(pf_anchor_stack) / sizeof(pf_anchor_stack[0])) { printf("pf_step_into_anchor: stack overflow\n"); *r = TAILQ_NEXT(*r, entries); return; } else if (*depth == 0 && a != NULL) *a = *r; f = pf_anchor_stack + (*depth)++; f->rs = *rs; f->r = *r; if ((*r)->anchor_wildcard) { f->parent = &(*r)->anchor->children; if ((f->child = RB_MIN(pf_anchor_node, f->parent)) == NULL) { *r = NULL; return; } *rs = &f->child->ruleset; } else { f->parent = NULL; f->child = NULL; *rs = &(*r)->anchor->ruleset; } *r = TAILQ_FIRST((*rs)->rules[n].active.ptr); } int pf_step_out_of_anchor(int *depth, struct pf_ruleset **rs, int n, struct pf_rule **r, struct pf_rule **a, int *match) { struct pf_anchor_stackframe *f; int quick = 0; do { if (*depth <= 0) break; f = pf_anchor_stack + *depth - 1; if (f->parent != NULL && f->child != NULL) { if (f->child->match || (match != NULL && *match)) { f->r->anchor->match = 1; *match = 0; } f->child = RB_NEXT(pf_anchor_node, f->parent, f->child); if (f->child != NULL) { *rs = &f->child->ruleset; *r = TAILQ_FIRST((*rs)->rules[n].active.ptr); if (*r == NULL) continue; else break; } } (*depth)--; if (*depth == 0 && a != NULL) *a = NULL; *rs = f->rs; if (f->r->anchor->match || (match != NULL && *match)) quick = f->r->quick; *r = TAILQ_NEXT(f->r, entries); } while (*r == NULL); return (quick); } #ifdef INET6 void pf_poolmask(struct pf_addr *naddr, struct pf_addr *raddr, struct pf_addr *rmask, struct pf_addr *saddr, sa_family_t af) { switch (af) { #ifdef INET case AF_INET: naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) | ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]); break; #endif /* INET */ case AF_INET6: naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) | ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]); naddr->addr32[1] = (raddr->addr32[1] & rmask->addr32[1]) | ((rmask->addr32[1] ^ 0xffffffff ) & saddr->addr32[1]); naddr->addr32[2] = (raddr->addr32[2] & rmask->addr32[2]) | ((rmask->addr32[2] ^ 0xffffffff ) & saddr->addr32[2]); naddr->addr32[3] = (raddr->addr32[3] & rmask->addr32[3]) | ((rmask->addr32[3] ^ 0xffffffff ) & saddr->addr32[3]); break; } } void pf_addr_inc(struct pf_addr *addr, sa_family_t af) { switch (af) { #ifdef INET case AF_INET: addr->addr32[0] = htonl(ntohl(addr->addr32[0]) + 1); break; #endif /* INET */ case AF_INET6: if (addr->addr32[3] == 0xffffffff) { addr->addr32[3] = 0; if (addr->addr32[2] == 0xffffffff) { addr->addr32[2] = 0; if (addr->addr32[1] == 0xffffffff) { addr->addr32[1] = 0; addr->addr32[0] = htonl(ntohl(addr->addr32[0]) + 1); } else addr->addr32[1] = htonl(ntohl(addr->addr32[1]) + 1); } else addr->addr32[2] = htonl(ntohl(addr->addr32[2]) + 1); } else addr->addr32[3] = htonl(ntohl(addr->addr32[3]) + 1); break; } } #endif /* INET6 */ #define mix(a,b,c) \ do { \ a -= b; a -= c; a ^= (c >> 13); \ b -= c; b -= a; b ^= (a << 8); \ c -= a; c -= b; c ^= (b >> 13); \ a -= b; a -= c; a ^= (c >> 12); \ b -= c; b -= a; b ^= (a << 16); \ c -= a; c -= b; c ^= (b >> 5); \ a -= b; a -= c; a ^= (c >> 3); \ b -= c; b -= a; b ^= (a << 10); \ c -= a; c -= b; c ^= (b >> 15); \ } while (0) /* * hash function based on bridge_hash in if_bridge.c */ void pf_hash(struct pf_addr *inaddr, struct pf_addr *hash, struct pf_poolhashkey *key, sa_family_t af) { u_int32_t a = 0x9e3779b9, b = 0x9e3779b9, c = key->key32[0]; switch (af) { #ifdef INET case AF_INET: a += inaddr->addr32[0]; b += key->key32[1]; mix(a, b, c); hash->addr32[0] = c + key->key32[2]; break; #endif /* INET */ #ifdef INET6 case AF_INET6: a += inaddr->addr32[0]; b += inaddr->addr32[2]; mix(a, b, c); hash->addr32[0] = c; a += inaddr->addr32[1]; b += inaddr->addr32[3]; c += key->key32[1]; mix(a, b, c); hash->addr32[1] = c; a += inaddr->addr32[2]; b += inaddr->addr32[1]; c += key->key32[2]; mix(a, b, c); hash->addr32[2] = c; a += inaddr->addr32[3]; b += inaddr->addr32[0]; c += key->key32[3]; mix(a, b, c); hash->addr32[3] = c; break; #endif /* INET6 */ } } int pf_map_addr(sa_family_t af, struct pf_rule *r, struct pf_addr *saddr, struct pf_addr *naddr, struct pf_addr *init_addr, struct pf_src_node **sn) { unsigned char hash[16]; struct pf_pool *rpool = &r->rpool; struct pf_addr *raddr = &rpool->cur->addr.v.a.addr; struct pf_addr *rmask = &rpool->cur->addr.v.a.mask; struct pf_pooladdr *acur = rpool->cur; struct pf_src_node k; if (*sn == NULL && r->rpool.opts & PF_POOL_STICKYADDR && (r->rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) { k.af = af; PF_ACPY(&k.addr, saddr, af); if (r->rule_flag & PFRULE_RULESRCTRACK || r->rpool.opts & PF_POOL_STICKYADDR) k.rule.ptr = r; else k.rule.ptr = NULL; pf_status.scounters[SCNT_SRC_NODE_SEARCH]++; *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k); if (*sn != NULL && !PF_AZERO(&(*sn)->raddr, af)) { PF_ACPY(naddr, &(*sn)->raddr, af); if (pf_status.debug >= PF_DEBUG_MISC) { printf("pf_map_addr: src tracking maps "); pf_print_host(&k.addr, 0, af); printf(" to "); pf_print_host(naddr, 0, af); printf("\n"); } return (0); } } if (rpool->cur->addr.type == PF_ADDR_NOROUTE) return (1); if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) { switch (af) { #ifdef INET case AF_INET: if (rpool->cur->addr.p.dyn->pfid_acnt4 < 1 && (rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN) return (1); raddr = &rpool->cur->addr.p.dyn->pfid_addr4; rmask = &rpool->cur->addr.p.dyn->pfid_mask4; break; #endif /* INET */ #ifdef INET6 case AF_INET6: if (rpool->cur->addr.p.dyn->pfid_acnt6 < 1 && (rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN) return (1); raddr = &rpool->cur->addr.p.dyn->pfid_addr6; rmask = &rpool->cur->addr.p.dyn->pfid_mask6; break; #endif /* INET6 */ } } else if (rpool->cur->addr.type == PF_ADDR_TABLE) { if ((rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN) return (1); /* unsupported */ } else { raddr = &rpool->cur->addr.v.a.addr; rmask = &rpool->cur->addr.v.a.mask; } switch (rpool->opts & PF_POOL_TYPEMASK) { case PF_POOL_NONE: PF_ACPY(naddr, raddr, af); break; case PF_POOL_BITMASK: PF_POOLMASK(naddr, raddr, rmask, saddr, af); break; case PF_POOL_RANDOM: if (init_addr != NULL && PF_AZERO(init_addr, af)) { switch (af) { #ifdef INET case AF_INET: rpool->counter.addr32[0] = htonl(arc4random()); break; #endif /* INET */ #ifdef INET6 case AF_INET6: if (rmask->addr32[3] != 0xffffffff) rpool->counter.addr32[3] = htonl(arc4random()); else break; if (rmask->addr32[2] != 0xffffffff) rpool->counter.addr32[2] = htonl(arc4random()); else break; if (rmask->addr32[1] != 0xffffffff) rpool->counter.addr32[1] = htonl(arc4random()); else break; if (rmask->addr32[0] != 0xffffffff) rpool->counter.addr32[0] = htonl(arc4random()); break; #endif /* INET6 */ } PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af); PF_ACPY(init_addr, naddr, af); } else { PF_AINC(&rpool->counter, af); PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af); } break; case PF_POOL_SRCHASH: pf_hash(saddr, (struct pf_addr *)&hash, &rpool->key, af); PF_POOLMASK(naddr, raddr, rmask, (struct pf_addr *)&hash, af); break; case PF_POOL_ROUNDROBIN: if (rpool->cur->addr.type == PF_ADDR_TABLE) { if (!pfr_pool_get(rpool->cur->addr.p.tbl, &rpool->tblidx, &rpool->counter, &raddr, &rmask, af)) goto get_addr; } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) { if (!pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt, &rpool->tblidx, &rpool->counter, &raddr, &rmask, af)) goto get_addr; } else if (pf_match_addr(0, raddr, rmask, &rpool->counter, af)) goto get_addr; try_next: if ((rpool->cur = TAILQ_NEXT(rpool->cur, entries)) == NULL) rpool->cur = TAILQ_FIRST(&rpool->list); if (rpool->cur->addr.type == PF_ADDR_TABLE) { rpool->tblidx = -1; if (pfr_pool_get(rpool->cur->addr.p.tbl, &rpool->tblidx, &rpool->counter, &raddr, &rmask, af)) { /* table contains no address of type 'af' */ if (rpool->cur != acur) goto try_next; return (1); } } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) { rpool->tblidx = -1; if (pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt, &rpool->tblidx, &rpool->counter, &raddr, &rmask, af)) { /* table contains no address of type 'af' */ if (rpool->cur != acur) goto try_next; return (1); } } else { raddr = &rpool->cur->addr.v.a.addr; rmask = &rpool->cur->addr.v.a.mask; PF_ACPY(&rpool->counter, raddr, af); } get_addr: PF_ACPY(naddr, &rpool->counter, af); if (init_addr != NULL && PF_AZERO(init_addr, af)) PF_ACPY(init_addr, naddr, af); PF_AINC(&rpool->counter, af); break; } if (*sn != NULL) PF_ACPY(&(*sn)->raddr, naddr, af); if (pf_status.debug >= PF_DEBUG_MISC && (rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) { printf("pf_map_addr: selected address "); pf_print_host(naddr, 0, af); printf("\n"); } return (0); } int pf_get_sport(sa_family_t af, u_int8_t proto, struct pf_rule *r, struct pf_addr *saddr, struct pf_addr *daddr, u_int16_t dport, struct pf_addr *naddr, u_int16_t *nport, u_int16_t low, u_int16_t high, struct pf_src_node **sn) { struct pf_state_cmp key; struct pf_addr init_addr; u_int16_t cut; bzero(&init_addr, sizeof(init_addr)); if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn)) return (1); if (proto == IPPROTO_ICMP) { low = 1; high = 65535; } do { key.af = af; key.proto = proto; PF_ACPY(&key.ext.addr, daddr, key.af); PF_ACPY(&key.gwy.addr, naddr, key.af); key.ext.port = dport; /* * port search; start random, step; * similar 2 portloop in in_pcbbind */ if (!(proto == IPPROTO_TCP || proto == IPPROTO_UDP || proto == IPPROTO_ICMP)) { key.gwy.port = dport; if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL) return (0); } else if (low == 0 && high == 0) { key.gwy.port = *nport; if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL) return (0); } else if (low == high) { key.gwy.port = htons(low); if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL) { *nport = htons(low); return (0); } } else { u_int16_t tmp; if (low > high) { tmp = low; low = high; high = tmp; } /* low < high */ cut = htonl(arc4random()) % (1 + high - low) + low; /* low <= cut <= high */ for (tmp = cut; tmp <= high; ++(tmp)) { key.gwy.port = htons(tmp); if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL) { *nport = htons(tmp); return (0); } } for (tmp = cut - 1; tmp >= low; --(tmp)) { key.gwy.port = htons(tmp); if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL) { *nport = htons(tmp); return (0); } } } switch (r->rpool.opts & PF_POOL_TYPEMASK) { case PF_POOL_RANDOM: case PF_POOL_ROUNDROBIN: if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn)) return (1); break; case PF_POOL_NONE: case PF_POOL_SRCHASH: case PF_POOL_BITMASK: default: return (1); } } while (! PF_AEQ(&init_addr, naddr, af) ); return (1); /* none available */ } struct pf_rule * pf_match_translation(struct pf_pdesc *pd, struct mbuf *m, int off, int direction, struct pfi_kif *kif, struct pf_addr *saddr, u_int16_t sport, struct pf_addr *daddr, u_int16_t dport, int rs_num) { struct pf_rule *r, *rm = NULL; struct pf_ruleset *ruleset = NULL; int tag = -1; int rtableid = -1; int asd = 0; r = TAILQ_FIRST(pf_main_ruleset.rules[rs_num].active.ptr); while (r && rm == NULL) { struct pf_rule_addr *src = NULL, *dst = NULL; struct pf_addr_wrap *xdst = NULL; if (r->action == PF_BINAT && direction == PF_IN) { src = &r->dst; if (r->rpool.cur != NULL) xdst = &r->rpool.cur->addr; } else { src = &r->src; dst = &r->dst; } r->evaluations++; if (pfi_kif_match(r->kif, kif) == r->ifnot) r = r->skip[PF_SKIP_IFP].ptr; else if (r->direction && r->direction != direction) r = r->skip[PF_SKIP_DIR].ptr; else if (r->af && r->af != pd->af) r = r->skip[PF_SKIP_AF].ptr; else if (r->proto && r->proto != pd->proto) r = r->skip[PF_SKIP_PROTO].ptr; else if (PF_MISMATCHAW(&src->addr, saddr, pd->af, src->neg, kif)) r = r->skip[src == &r->src ? PF_SKIP_SRC_ADDR : PF_SKIP_DST_ADDR].ptr; else if (src->port_op && !pf_match_port(src->port_op, src->port[0], src->port[1], sport)) r = r->skip[src == &r->src ? PF_SKIP_SRC_PORT : PF_SKIP_DST_PORT].ptr; else if (dst != NULL && PF_MISMATCHAW(&dst->addr, daddr, pd->af, dst->neg, NULL)) r = r->skip[PF_SKIP_DST_ADDR].ptr; else if (xdst != NULL && PF_MISMATCHAW(xdst, daddr, pd->af, 0, NULL)) r = TAILQ_NEXT(r, entries); else if (dst != NULL && dst->port_op && !pf_match_port(dst->port_op, dst->port[0], dst->port[1], dport)) r = r->skip[PF_SKIP_DST_PORT].ptr; else if (r->match_tag && !pf_match_tag(m, r, pd->pf_mtag, &tag)) r = TAILQ_NEXT(r, entries); else if (r->os_fingerprint != PF_OSFP_ANY && (pd->proto != IPPROTO_TCP || !pf_osfp_match(pf_osfp_fingerprint(pd, m, off, pd->hdr.tcp), r->os_fingerprint))) r = TAILQ_NEXT(r, entries); else { if (r->tag) tag = r->tag; if (r->rtableid >= 0) rtableid = r->rtableid; if (r->anchor == NULL) { rm = r; } else pf_step_into_anchor(&asd, &ruleset, rs_num, &r, NULL, NULL); } if (r == NULL) pf_step_out_of_anchor(&asd, &ruleset, rs_num, &r, NULL, NULL); } if (pf_tag_packet(m, pd->pf_mtag, tag, rtableid)) return (NULL); if (rm != NULL && (rm->action == PF_NONAT || rm->action == PF_NORDR || rm->action == PF_NOBINAT)) return (NULL); return (rm); } struct pf_rule * pf_get_translation(struct pf_pdesc *pd, struct mbuf *m, int off, int direction, struct pfi_kif *kif, struct pf_src_node **sn, struct pf_addr *saddr, u_int16_t sport, struct pf_addr *daddr, u_int16_t dport, struct pf_addr *naddr, u_int16_t *nport) { struct pf_rule *r = NULL; if (direction == PF_OUT) { r = pf_match_translation(pd, m, off, direction, kif, saddr, sport, daddr, dport, PF_RULESET_BINAT); if (r == NULL) r = pf_match_translation(pd, m, off, direction, kif, saddr, sport, daddr, dport, PF_RULESET_NAT); } else { r = pf_match_translation(pd, m, off, direction, kif, saddr, sport, daddr, dport, PF_RULESET_RDR); if (r == NULL) r = pf_match_translation(pd, m, off, direction, kif, saddr, sport, daddr, dport, PF_RULESET_BINAT); } if (r != NULL) { switch (r->action) { case PF_NONAT: case PF_NOBINAT: case PF_NORDR: return (NULL); case PF_NAT: if (pf_get_sport(pd->af, pd->proto, r, saddr, daddr, dport, naddr, nport, r->rpool.proxy_port[0], r->rpool.proxy_port[1], sn)) { DPFPRINTF(PF_DEBUG_MISC, ("pf: NAT proxy port allocation " "(%u-%u) failed\n", r->rpool.proxy_port[0], r->rpool.proxy_port[1])); return (NULL); } break; case PF_BINAT: switch (direction) { case PF_OUT: if (r->rpool.cur->addr.type == PF_ADDR_DYNIFTL){ switch (pd->af) { #ifdef INET case AF_INET: if (r->rpool.cur->addr.p.dyn-> pfid_acnt4 < 1) return (NULL); PF_POOLMASK(naddr, &r->rpool.cur->addr.p.dyn-> pfid_addr4, &r->rpool.cur->addr.p.dyn-> pfid_mask4, saddr, AF_INET); break; #endif /* INET */ #ifdef INET6 case AF_INET6: if (r->rpool.cur->addr.p.dyn-> pfid_acnt6 < 1) return (NULL); PF_POOLMASK(naddr, &r->rpool.cur->addr.p.dyn-> pfid_addr6, &r->rpool.cur->addr.p.dyn-> pfid_mask6, saddr, AF_INET6); break; #endif /* INET6 */ } } else PF_POOLMASK(naddr, &r->rpool.cur->addr.v.a.addr, &r->rpool.cur->addr.v.a.mask, saddr, pd->af); break; case PF_IN: if (r->src.addr.type == PF_ADDR_DYNIFTL) { switch (pd->af) { #ifdef INET case AF_INET: if (r->src.addr.p.dyn-> pfid_acnt4 < 1) return (NULL); PF_POOLMASK(naddr, &r->src.addr.p.dyn-> pfid_addr4, &r->src.addr.p.dyn-> pfid_mask4, daddr, AF_INET); break; #endif /* INET */ #ifdef INET6 case AF_INET6: if (r->src.addr.p.dyn-> pfid_acnt6 < 1) return (NULL); PF_POOLMASK(naddr, &r->src.addr.p.dyn-> pfid_addr6, &r->src.addr.p.dyn-> pfid_mask6, daddr, AF_INET6); break; #endif /* INET6 */ } } else PF_POOLMASK(naddr, &r->src.addr.v.a.addr, &r->src.addr.v.a.mask, daddr, pd->af); break; } break; case PF_RDR: { if (pf_map_addr(pd->af, r, saddr, naddr, NULL, sn)) return (NULL); if ((r->rpool.opts & PF_POOL_TYPEMASK) == PF_POOL_BITMASK) PF_POOLMASK(naddr, naddr, &r->rpool.cur->addr.v.a.mask, daddr, pd->af); if (r->rpool.proxy_port[1]) { u_int32_t tmp_nport; tmp_nport = ((ntohs(dport) - ntohs(r->dst.port[0])) % (r->rpool.proxy_port[1] - r->rpool.proxy_port[0] + 1)) + r->rpool.proxy_port[0]; /* wrap around if necessary */ if (tmp_nport > 65535) tmp_nport -= 65535; *nport = htons((u_int16_t)tmp_nport); } else if (r->rpool.proxy_port[0]) *nport = htons(r->rpool.proxy_port[0]); break; } default: return (NULL); } } return (r); } int #ifdef __FreeBSD__ pf_socket_lookup(int direction, struct pf_pdesc *pd, struct inpcb *inp_arg) #else pf_socket_lookup(int direction, struct pf_pdesc *pd) #endif { struct pf_addr *saddr, *daddr; u_int16_t sport, dport; #ifdef __FreeBSD__ struct inpcbinfo *pi; #else struct inpcbtable *tb; #endif struct inpcb *inp; if (pd == NULL) return (-1); pd->lookup.uid = UID_MAX; pd->lookup.gid = GID_MAX; pd->lookup.pid = NO_PID; /* XXX: revisit */ #ifdef __FreeBSD__ if (inp_arg != NULL) { INP_LOCK_ASSERT(inp_arg); pd->lookup.uid = inp_arg->inp_cred->cr_uid; pd->lookup.gid = inp_arg->inp_cred->cr_groups[0]; return (1); } #endif switch (pd->proto) { case IPPROTO_TCP: if (pd->hdr.tcp == NULL) return (-1); sport = pd->hdr.tcp->th_sport; dport = pd->hdr.tcp->th_dport; #ifdef __FreeBSD__ pi = &V_tcbinfo; #else tb = &tcbtable; #endif break; case IPPROTO_UDP: if (pd->hdr.udp == NULL) return (-1); sport = pd->hdr.udp->uh_sport; dport = pd->hdr.udp->uh_dport; #ifdef __FreeBSD__ pi = &V_udbinfo; #else tb = &udbtable; #endif break; default: return (-1); } if (direction == PF_IN) { saddr = pd->src; daddr = pd->dst; } else { u_int16_t p; p = sport; sport = dport; dport = p; saddr = pd->dst; daddr = pd->src; } switch (pd->af) { #ifdef INET case AF_INET: #ifdef __FreeBSD__ - INP_INFO_RLOCK(pi); /* XXX LOR */ - inp = in_pcblookup_hash(pi, saddr->v4, sport, daddr->v4, - dport, 0, NULL); + inp = in_pcblookup(pi, saddr->v4, sport, daddr->v4, + dport, INPLOOKUP_RLOCKPCB, NULL); if (inp == NULL) { - inp = in_pcblookup_hash(pi, saddr->v4, sport, - daddr->v4, dport, INPLOOKUP_WILDCARD, NULL); - if(inp == NULL) { - INP_INFO_RUNLOCK(pi); + inp = in_pcblookup(pi, saddr->v4, sport, + daddr->v4, dport, INPLOOKUP_WILDCARD | + INPLOOKUP_RLOCKPCB, NULL); + if (inp == NULL) return (-1); - } } #else inp = in_pcbhashlookup(tb, saddr->v4, sport, daddr->v4, dport); if (inp == NULL) { inp = in_pcblookup_listen(tb, daddr->v4, dport, 0); if (inp == NULL) return (-1); } #endif break; #endif /* INET */ #ifdef INET6 case AF_INET6: #ifdef __FreeBSD__ - INP_INFO_RLOCK(pi); - inp = in6_pcblookup_hash(pi, &saddr->v6, sport, - &daddr->v6, dport, 0, NULL); + inp = in6_pcblookup(pi, &saddr->v6, sport, + &daddr->v6, dport, INPLOOKUP_RLOCKPCB, NULL); if (inp == NULL) { - inp = in6_pcblookup_hash(pi, &saddr->v6, sport, - &daddr->v6, dport, INPLOOKUP_WILDCARD, NULL); - if (inp == NULL) { - INP_INFO_RUNLOCK(pi); + inp = in6_pcblookup(pi, &saddr->v6, sport, + &daddr->v6, dport, INPLOOKUP_WILDCARD | + INPLOOKUP_RLOCKPCB, NULL); + if (inp == NULL) return (-1); - } } #else inp = in6_pcbhashlookup(tb, &saddr->v6, sport, &daddr->v6, dport); if (inp == NULL) { inp = in6_pcblookup_listen(tb, &daddr->v6, dport, 0); if (inp == NULL) return (-1); } #endif break; #endif /* INET6 */ default: return (-1); } #ifdef __FreeBSD__ + INP_RLOCK_ASSERT(inp); pd->lookup.uid = inp->inp_cred->cr_uid; pd->lookup.gid = inp->inp_cred->cr_groups[0]; - INP_INFO_RUNLOCK(pi); + INP_RUNLOCK(inp); #else pd->lookup.uid = inp->inp_socket->so_euid; pd->lookup.gid = inp->inp_socket->so_egid; pd->lookup.pid = inp->inp_socket->so_cpid; #endif return (1); } u_int8_t pf_get_wscale(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af) { int hlen; u_int8_t hdr[60]; u_int8_t *opt, optlen; u_int8_t wscale = 0; hlen = th_off << 2; /* hlen <= sizeof(hdr) */ if (hlen <= sizeof(struct tcphdr)) return (0); if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af)) return (0); opt = hdr + sizeof(struct tcphdr); hlen -= sizeof(struct tcphdr); while (hlen >= 3) { switch (*opt) { case TCPOPT_EOL: case TCPOPT_NOP: ++opt; --hlen; break; case TCPOPT_WINDOW: wscale = opt[2]; if (wscale > TCP_MAX_WINSHIFT) wscale = TCP_MAX_WINSHIFT; wscale |= PF_WSCALE_FLAG; /* FALLTHROUGH */ default: optlen = opt[1]; if (optlen < 2) optlen = 2; hlen -= optlen; opt += optlen; break; } } return (wscale); } u_int16_t pf_get_mss(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af) { int hlen; u_int8_t hdr[60]; u_int8_t *opt, optlen; u_int16_t mss = V_tcp_mssdflt; hlen = th_off << 2; /* hlen <= sizeof(hdr) */ if (hlen <= sizeof(struct tcphdr)) return (0); if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af)) return (0); opt = hdr + sizeof(struct tcphdr); hlen -= sizeof(struct tcphdr); while (hlen >= TCPOLEN_MAXSEG) { switch (*opt) { case TCPOPT_EOL: case TCPOPT_NOP: ++opt; --hlen; break; case TCPOPT_MAXSEG: bcopy((caddr_t)(opt + 2), (caddr_t)&mss, 2); NTOHS(mss); /* FALLTHROUGH */ default: optlen = opt[1]; if (optlen < 2) optlen = 2; hlen -= optlen; opt += optlen; break; } } return (mss); } u_int16_t pf_calc_mss(struct pf_addr *addr, sa_family_t af, u_int16_t offer) { #ifdef INET struct sockaddr_in *dst; struct route ro; #endif /* INET */ #ifdef INET6 struct sockaddr_in6 *dst6; struct route_in6 ro6; #endif /* INET6 */ struct rtentry *rt = NULL; int hlen = 0; /* make the compiler happy */ u_int16_t mss = V_tcp_mssdflt; switch (af) { #ifdef INET case AF_INET: hlen = sizeof(struct ip); bzero(&ro, sizeof(ro)); dst = (struct sockaddr_in *)&ro.ro_dst; dst->sin_family = AF_INET; dst->sin_len = sizeof(*dst); dst->sin_addr = addr->v4; #ifdef __FreeBSD__ #ifdef RTF_PRCLONING rtalloc_ign(&ro, (RTF_CLONING | RTF_PRCLONING)); #else /* !RTF_PRCLONING */ in_rtalloc_ign(&ro, 0, 0); #endif #else /* ! __FreeBSD__ */ rtalloc_noclone(&ro, NO_CLONING); #endif rt = ro.ro_rt; break; #endif /* INET */ #ifdef INET6 case AF_INET6: hlen = sizeof(struct ip6_hdr); bzero(&ro6, sizeof(ro6)); dst6 = (struct sockaddr_in6 *)&ro6.ro_dst; dst6->sin6_family = AF_INET6; dst6->sin6_len = sizeof(*dst6); dst6->sin6_addr = addr->v6; #ifdef __FreeBSD__ #ifdef RTF_PRCLONING rtalloc_ign((struct route *)&ro6, (RTF_CLONING | RTF_PRCLONING)); #else /* !RTF_PRCLONING */ rtalloc_ign((struct route *)&ro6, 0); #endif #else /* ! __FreeBSD__ */ rtalloc_noclone((struct route *)&ro6, NO_CLONING); #endif rt = ro6.ro_rt; break; #endif /* INET6 */ } if (rt && rt->rt_ifp) { mss = rt->rt_ifp->if_mtu - hlen - sizeof(struct tcphdr); mss = max(V_tcp_mssdflt, mss); RTFREE(rt); } mss = min(mss, offer); mss = max(mss, 64); /* sanity - at least max opt space */ return (mss); } void pf_set_rt_ifp(struct pf_state *s, struct pf_addr *saddr) { struct pf_rule *r = s->rule.ptr; s->rt_kif = NULL; if (!r->rt || r->rt == PF_FASTROUTE) return; switch (s->af) { #ifdef INET case AF_INET: pf_map_addr(AF_INET, r, saddr, &s->rt_addr, NULL, &s->nat_src_node); s->rt_kif = r->rpool.cur->kif; break; #endif /* INET */ #ifdef INET6 case AF_INET6: pf_map_addr(AF_INET6, r, saddr, &s->rt_addr, NULL, &s->nat_src_node); s->rt_kif = r->rpool.cur->kif; break; #endif /* INET6 */ } } int pf_test_tcp(struct pf_rule **rm, struct pf_state **sm, int direction, struct pfi_kif *kif, struct mbuf *m, int off, void *h, #ifdef __FreeBSD__ struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm, struct ifqueue *ifq, struct inpcb *inp) #else struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm, struct ifqueue *ifq) #endif { struct pf_rule *nr = NULL; struct pf_addr *saddr = pd->src, *daddr = pd->dst; struct tcphdr *th = pd->hdr.tcp; u_int16_t bport, nport = 0; sa_family_t af = pd->af; struct pf_rule *r, *a = NULL; struct pf_ruleset *ruleset = NULL; struct pf_src_node *nsn = NULL; u_short reason; int rewrite = 0; int tag = -1, rtableid = -1; u_int16_t mss = V_tcp_mssdflt; int asd = 0; int match = 0; if (pf_check_congestion(ifq)) { REASON_SET(&reason, PFRES_CONGEST); return (PF_DROP); } #ifdef __FreeBSD__ if (inp != NULL) pd->lookup.done = pf_socket_lookup(direction, pd, inp); else if (debug_pfugidhack) { PF_UNLOCK(); DPFPRINTF(PF_DEBUG_MISC, ("pf: unlocked lookup\n")); pd->lookup.done = pf_socket_lookup(direction, pd, inp); PF_LOCK(); } #endif r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr); if (direction == PF_OUT) { bport = nport = th->th_sport; /* check outgoing packet for BINAT/NAT */ if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn, saddr, th->th_sport, daddr, th->th_dport, &pd->naddr, &nport)) != NULL) { PF_ACPY(&pd->baddr, saddr, af); pf_change_ap(saddr, &th->th_sport, pd->ip_sum, &th->th_sum, &pd->naddr, nport, 0, af); rewrite++; if (nr->natpass) r = NULL; pd->nat_rule = nr; } } else { bport = nport = th->th_dport; /* check incoming packet for BINAT/RDR */ if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn, saddr, th->th_sport, daddr, th->th_dport, &pd->naddr, &nport)) != NULL) { PF_ACPY(&pd->baddr, daddr, af); pf_change_ap(daddr, &th->th_dport, pd->ip_sum, &th->th_sum, &pd->naddr, nport, 0, af); rewrite++; if (nr->natpass) r = NULL; pd->nat_rule = nr; } } while (r != NULL) { r->evaluations++; if (pfi_kif_match(r->kif, kif) == r->ifnot) r = r->skip[PF_SKIP_IFP].ptr; else if (r->direction && r->direction != direction) r = r->skip[PF_SKIP_DIR].ptr; else if (r->af && r->af != af) r = r->skip[PF_SKIP_AF].ptr; else if (r->proto && r->proto != IPPROTO_TCP) r = r->skip[PF_SKIP_PROTO].ptr; else if (PF_MISMATCHAW(&r->src.addr, saddr, af, r->src.neg, kif)) r = r->skip[PF_SKIP_SRC_ADDR].ptr; else if (r->src.port_op && !pf_match_port(r->src.port_op, r->src.port[0], r->src.port[1], th->th_sport)) r = r->skip[PF_SKIP_SRC_PORT].ptr; else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, r->dst.neg, NULL)) r = r->skip[PF_SKIP_DST_ADDR].ptr; else if (r->dst.port_op && !pf_match_port(r->dst.port_op, r->dst.port[0], r->dst.port[1], th->th_dport)) r = r->skip[PF_SKIP_DST_PORT].ptr; else if (r->tos && !(r->tos == pd->tos)) r = TAILQ_NEXT(r, entries); else if (r->rule_flag & PFRULE_FRAGMENT) r = TAILQ_NEXT(r, entries); else if ((r->flagset & th->th_flags) != r->flags) r = TAILQ_NEXT(r, entries); else if (r->uid.op && (pd->lookup.done || (pd->lookup.done = #ifdef __FreeBSD__ pf_socket_lookup(direction, pd, inp), 1)) && #else pf_socket_lookup(direction, pd), 1)) && #endif !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1], pd->lookup.uid)) r = TAILQ_NEXT(r, entries); else if (r->gid.op && (pd->lookup.done || (pd->lookup.done = #ifdef __FreeBSD__ pf_socket_lookup(direction, pd, inp), 1)) && #else pf_socket_lookup(direction, pd), 1)) && #endif !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1], pd->lookup.gid)) r = TAILQ_NEXT(r, entries); else if (r->prob && r->prob <= arc4random()) r = TAILQ_NEXT(r, entries); else if (r->match_tag && !pf_match_tag(m, r, pd->pf_mtag, &tag)) r = TAILQ_NEXT(r, entries); else if (r->os_fingerprint != PF_OSFP_ANY && !pf_osfp_match( pf_osfp_fingerprint(pd, m, off, th), r->os_fingerprint)) r = TAILQ_NEXT(r, entries); else { if (r->tag) tag = r->tag; if (r->rtableid >= 0) rtableid = r->rtableid; if (r->anchor == NULL) { match = 1; *rm = r; *am = a; *rsm = ruleset; if ((*rm)->quick) break; r = TAILQ_NEXT(r, entries); } else pf_step_into_anchor(&asd, &ruleset, PF_RULESET_FILTER, &r, &a, &match); } if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset, PF_RULESET_FILTER, &r, &a, &match)) break; } r = *rm; a = *am; ruleset = *rsm; REASON_SET(&reason, PFRES_MATCH); if (r->log || (nr != NULL && nr->natpass && nr->log)) { if (rewrite) #ifdef __FreeBSD__ m_copyback(m, off, sizeof(*th), (caddr_t)th); #else m_copyback(m, off, sizeof(*th), th); #endif PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr, a, ruleset, pd); } if ((r->action == PF_DROP) && ((r->rule_flag & PFRULE_RETURNRST) || (r->rule_flag & PFRULE_RETURNICMP) || (r->rule_flag & PFRULE_RETURN))) { /* undo NAT changes, if they have taken place */ if (nr != NULL) { if (direction == PF_OUT) { pf_change_ap(saddr, &th->th_sport, pd->ip_sum, &th->th_sum, &pd->baddr, bport, 0, af); rewrite++; } else { pf_change_ap(daddr, &th->th_dport, pd->ip_sum, &th->th_sum, &pd->baddr, bport, 0, af); rewrite++; } } if (((r->rule_flag & PFRULE_RETURNRST) || (r->rule_flag & PFRULE_RETURN)) && !(th->th_flags & TH_RST)) { u_int32_t ack = ntohl(th->th_seq) + pd->p_len; if (th->th_flags & TH_SYN) ack++; if (th->th_flags & TH_FIN) ack++; #ifdef __FreeBSD__ pf_send_tcp(m, r, af, pd->dst, #else pf_send_tcp(r, af, pd->dst, #endif pd->src, th->th_dport, th->th_sport, ntohl(th->th_ack), ack, TH_RST|TH_ACK, 0, 0, r->return_ttl, 1, 0, pd->eh, kif->pfik_ifp); } else if ((af == AF_INET) && r->return_icmp) pf_send_icmp(m, r->return_icmp >> 8, r->return_icmp & 255, af, r); else if ((af == AF_INET6) && r->return_icmp6) pf_send_icmp(m, r->return_icmp6 >> 8, r->return_icmp6 & 255, af, r); } if (r->action == PF_DROP) return (PF_DROP); if (pf_tag_packet(m, pd->pf_mtag, tag, rtableid)) { REASON_SET(&reason, PFRES_MEMORY); return (PF_DROP); } if (r->keep_state || nr != NULL || (pd->flags & PFDESC_TCP_NORM)) { /* create new state */ u_int16_t len; struct pf_state *s = NULL; struct pf_src_node *sn = NULL; len = pd->tot_len - off - (th->th_off << 2); /* check maximums */ if (r->max_states && (r->states >= r->max_states)) { pf_status.lcounters[LCNT_STATES]++; REASON_SET(&reason, PFRES_MAXSTATES); goto cleanup; } /* src node for filter rule */ if ((r->rule_flag & PFRULE_SRCTRACK || r->rpool.opts & PF_POOL_STICKYADDR) && pf_insert_src_node(&sn, r, saddr, af) != 0) { REASON_SET(&reason, PFRES_SRCLIMIT); goto cleanup; } /* src node for translation rule */ if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) && ((direction == PF_OUT && pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) || (pf_insert_src_node(&nsn, nr, saddr, af) != 0))) { REASON_SET(&reason, PFRES_SRCLIMIT); goto cleanup; } s = pool_get(&pf_state_pl, PR_NOWAIT); if (s == NULL) { REASON_SET(&reason, PFRES_MEMORY); cleanup: if (sn != NULL && sn->states == 0 && sn->expire == 0) { RB_REMOVE(pf_src_tree, &tree_src_tracking, sn); pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; pf_status.src_nodes--; pool_put(&pf_src_tree_pl, sn); } if (nsn != sn && nsn != NULL && nsn->states == 0 && nsn->expire == 0) { RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn); pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; pf_status.src_nodes--; pool_put(&pf_src_tree_pl, nsn); } return (PF_DROP); } bzero(s, sizeof(*s)); s->rule.ptr = r; s->nat_rule.ptr = nr; s->anchor.ptr = a; STATE_INC_COUNTERS(s); if (r->allow_opts) s->state_flags |= PFSTATE_ALLOWOPTS; if (r->rule_flag & PFRULE_STATESLOPPY) s->state_flags |= PFSTATE_SLOPPY; s->log = r->log & PF_LOG_ALL; if (nr != NULL) s->log |= nr->log & PF_LOG_ALL; s->proto = IPPROTO_TCP; s->direction = direction; s->af = af; if (direction == PF_OUT) { PF_ACPY(&s->gwy.addr, saddr, af); s->gwy.port = th->th_sport; /* sport */ PF_ACPY(&s->ext.addr, daddr, af); s->ext.port = th->th_dport; if (nr != NULL) { PF_ACPY(&s->lan.addr, &pd->baddr, af); s->lan.port = bport; } else { PF_ACPY(&s->lan.addr, &s->gwy.addr, af); s->lan.port = s->gwy.port; } } else { PF_ACPY(&s->lan.addr, daddr, af); s->lan.port = th->th_dport; PF_ACPY(&s->ext.addr, saddr, af); s->ext.port = th->th_sport; if (nr != NULL) { PF_ACPY(&s->gwy.addr, &pd->baddr, af); s->gwy.port = bport; } else { PF_ACPY(&s->gwy.addr, &s->lan.addr, af); s->gwy.port = s->lan.port; } } s->src.seqlo = ntohl(th->th_seq); s->src.seqhi = s->src.seqlo + len + 1; if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN && r->keep_state == PF_STATE_MODULATE) { /* Generate sequence number modulator */ #ifdef __FreeBSD__ while ((s->src.seqdiff = pf_new_isn(s) - s->src.seqlo) == 0) ; #else while ((s->src.seqdiff = tcp_rndiss_next() - s->src.seqlo) == 0) ; #endif pf_change_a(&th->th_seq, &th->th_sum, htonl(s->src.seqlo + s->src.seqdiff), 0); rewrite = 1; } else s->src.seqdiff = 0; if (th->th_flags & TH_SYN) { s->src.seqhi++; s->src.wscale = pf_get_wscale(m, off, th->th_off, af); } s->src.max_win = MAX(ntohs(th->th_win), 1); if (s->src.wscale & PF_WSCALE_MASK) { /* Remove scale factor from initial window */ int win = s->src.max_win; win += 1 << (s->src.wscale & PF_WSCALE_MASK); s->src.max_win = (win - 1) >> (s->src.wscale & PF_WSCALE_MASK); } if (th->th_flags & TH_FIN) s->src.seqhi++; s->dst.seqhi = 1; s->dst.max_win = 1; s->src.state = TCPS_SYN_SENT; s->dst.state = TCPS_CLOSED; s->creation = time_second; s->expire = time_second; s->timeout = PFTM_TCP_FIRST_PACKET; pf_set_rt_ifp(s, saddr); if (sn != NULL) { s->src_node = sn; s->src_node->states++; } if (nsn != NULL) { PF_ACPY(&nsn->raddr, &pd->naddr, af); s->nat_src_node = nsn; s->nat_src_node->states++; } if ((pd->flags & PFDESC_TCP_NORM) && pf_normalize_tcp_init(m, off, pd, th, &s->src, &s->dst)) { REASON_SET(&reason, PFRES_MEMORY); pf_src_tree_remove_state(s); STATE_DEC_COUNTERS(s); pool_put(&pf_state_pl, s); return (PF_DROP); } if ((pd->flags & PFDESC_TCP_NORM) && s->src.scrub && pf_normalize_tcp_stateful(m, off, pd, &reason, th, s, &s->src, &s->dst, &rewrite)) { /* This really shouldn't happen!!! */ DPFPRINTF(PF_DEBUG_URGENT, ("pf_normalize_tcp_stateful failed on first pkt")); pf_normalize_tcp_cleanup(s); pf_src_tree_remove_state(s); STATE_DEC_COUNTERS(s); pool_put(&pf_state_pl, s); return (PF_DROP); } if (pf_insert_state(BOUND_IFACE(r, kif), s)) { pf_normalize_tcp_cleanup(s); REASON_SET(&reason, PFRES_STATEINS); pf_src_tree_remove_state(s); STATE_DEC_COUNTERS(s); pool_put(&pf_state_pl, s); return (PF_DROP); } else *sm = s; if (tag > 0) { pf_tag_ref(tag); s->tag = tag; } if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN && r->keep_state == PF_STATE_SYNPROXY) { s->src.state = PF_TCPS_PROXY_SRC; if (nr != NULL) { if (direction == PF_OUT) { pf_change_ap(saddr, &th->th_sport, pd->ip_sum, &th->th_sum, &pd->baddr, bport, 0, af); } else { pf_change_ap(daddr, &th->th_dport, pd->ip_sum, &th->th_sum, &pd->baddr, bport, 0, af); } } s->src.seqhi = htonl(arc4random()); /* Find mss option */ mss = pf_get_mss(m, off, th->th_off, af); mss = pf_calc_mss(saddr, af, mss); mss = pf_calc_mss(daddr, af, mss); s->src.mss = mss; #ifdef __FreeBSD__ pf_send_tcp(NULL, r, af, daddr, saddr, th->th_dport, #else pf_send_tcp(r, af, daddr, saddr, th->th_dport, #endif th->th_sport, s->src.seqhi, ntohl(th->th_seq) + 1, TH_SYN|TH_ACK, 0, s->src.mss, 0, 1, 0, NULL, NULL); REASON_SET(&reason, PFRES_SYNPROXY); return (PF_SYNPROXY_DROP); } } /* copy back packet headers if we performed NAT operations */ if (rewrite) m_copyback(m, off, sizeof(*th), (caddr_t)th); return (PF_PASS); } int pf_test_udp(struct pf_rule **rm, struct pf_state **sm, int direction, struct pfi_kif *kif, struct mbuf *m, int off, void *h, #ifdef __FreeBSD__ struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm, struct ifqueue *ifq, struct inpcb *inp) #else struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm, struct ifqueue *ifq) #endif { struct pf_rule *nr = NULL; struct pf_addr *saddr = pd->src, *daddr = pd->dst; struct udphdr *uh = pd->hdr.udp; u_int16_t bport, nport = 0; sa_family_t af = pd->af; struct pf_rule *r, *a = NULL; struct pf_ruleset *ruleset = NULL; struct pf_src_node *nsn = NULL; u_short reason; int rewrite = 0; int tag = -1, rtableid = -1; int asd = 0; int match = 0; if (pf_check_congestion(ifq)) { REASON_SET(&reason, PFRES_CONGEST); return (PF_DROP); } #ifdef __FreeBSD__ if (inp != NULL) pd->lookup.done = pf_socket_lookup(direction, pd, inp); else if (debug_pfugidhack) { PF_UNLOCK(); DPFPRINTF(PF_DEBUG_MISC, ("pf: unlocked lookup\n")); pd->lookup.done = pf_socket_lookup(direction, pd, inp); PF_LOCK(); } #endif r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr); if (direction == PF_OUT) { bport = nport = uh->uh_sport; /* check outgoing packet for BINAT/NAT */ if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn, saddr, uh->uh_sport, daddr, uh->uh_dport, &pd->naddr, &nport)) != NULL) { PF_ACPY(&pd->baddr, saddr, af); pf_change_ap(saddr, &uh->uh_sport, pd->ip_sum, &uh->uh_sum, &pd->naddr, nport, 1, af); rewrite++; if (nr->natpass) r = NULL; pd->nat_rule = nr; } } else { bport = nport = uh->uh_dport; /* check incoming packet for BINAT/RDR */ if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn, saddr, uh->uh_sport, daddr, uh->uh_dport, &pd->naddr, &nport)) != NULL) { PF_ACPY(&pd->baddr, daddr, af); pf_change_ap(daddr, &uh->uh_dport, pd->ip_sum, &uh->uh_sum, &pd->naddr, nport, 1, af); rewrite++; if (nr->natpass) r = NULL; pd->nat_rule = nr; } } while (r != NULL) { r->evaluations++; if (pfi_kif_match(r->kif, kif) == r->ifnot) r = r->skip[PF_SKIP_IFP].ptr; else if (r->direction && r->direction != direction) r = r->skip[PF_SKIP_DIR].ptr; else if (r->af && r->af != af) r = r->skip[PF_SKIP_AF].ptr; else if (r->proto && r->proto != IPPROTO_UDP) r = r->skip[PF_SKIP_PROTO].ptr; else if (PF_MISMATCHAW(&r->src.addr, saddr, af, r->src.neg, kif)) r = r->skip[PF_SKIP_SRC_ADDR].ptr; else if (r->src.port_op && !pf_match_port(r->src.port_op, r->src.port[0], r->src.port[1], uh->uh_sport)) r = r->skip[PF_SKIP_SRC_PORT].ptr; else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, r->dst.neg, NULL)) r = r->skip[PF_SKIP_DST_ADDR].ptr; else if (r->dst.port_op && !pf_match_port(r->dst.port_op, r->dst.port[0], r->dst.port[1], uh->uh_dport)) r = r->skip[PF_SKIP_DST_PORT].ptr; else if (r->tos && !(r->tos == pd->tos)) r = TAILQ_NEXT(r, entries); else if (r->rule_flag & PFRULE_FRAGMENT) r = TAILQ_NEXT(r, entries); else if (r->uid.op && (pd->lookup.done || (pd->lookup.done = #ifdef __FreeBSD__ pf_socket_lookup(direction, pd, inp), 1)) && #else pf_socket_lookup(direction, pd), 1)) && #endif !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1], pd->lookup.uid)) r = TAILQ_NEXT(r, entries); else if (r->gid.op && (pd->lookup.done || (pd->lookup.done = #ifdef __FreeBSD__ pf_socket_lookup(direction, pd, inp), 1)) && #else pf_socket_lookup(direction, pd), 1)) && #endif !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1], pd->lookup.gid)) r = TAILQ_NEXT(r, entries); else if (r->prob && r->prob <= arc4random()) r = TAILQ_NEXT(r, entries); else if (r->match_tag && !pf_match_tag(m, r, pd->pf_mtag, &tag)) r = TAILQ_NEXT(r, entries); else if (r->os_fingerprint != PF_OSFP_ANY) r = TAILQ_NEXT(r, entries); else { if (r->tag) tag = r->tag; if (r->rtableid >= 0) rtableid = r->rtableid; if (r->anchor == NULL) { match = 1; *rm = r; *am = a; *rsm = ruleset; if ((*rm)->quick) break; r = TAILQ_NEXT(r, entries); } else pf_step_into_anchor(&asd, &ruleset, PF_RULESET_FILTER, &r, &a, &match); } if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset, PF_RULESET_FILTER, &r, &a, &match)) break; } r = *rm; a = *am; ruleset = *rsm; REASON_SET(&reason, PFRES_MATCH); if (r->log || (nr != NULL && nr->natpass && nr->log)) { if (rewrite) #ifdef __FreeBSD__ m_copyback(m, off, sizeof(*uh), (caddr_t)uh); #else m_copyback(m, off, sizeof(*uh), uh); #endif PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr, a, ruleset, pd); } if ((r->action == PF_DROP) && ((r->rule_flag & PFRULE_RETURNICMP) || (r->rule_flag & PFRULE_RETURN))) { /* undo NAT changes, if they have taken place */ if (nr != NULL) { if (direction == PF_OUT) { pf_change_ap(saddr, &uh->uh_sport, pd->ip_sum, &uh->uh_sum, &pd->baddr, bport, 1, af); rewrite++; } else { pf_change_ap(daddr, &uh->uh_dport, pd->ip_sum, &uh->uh_sum, &pd->baddr, bport, 1, af); rewrite++; } } if ((af == AF_INET) && r->return_icmp) pf_send_icmp(m, r->return_icmp >> 8, r->return_icmp & 255, af, r); else if ((af == AF_INET6) && r->return_icmp6) pf_send_icmp(m, r->return_icmp6 >> 8, r->return_icmp6 & 255, af, r); } if (r->action == PF_DROP) return (PF_DROP); if (pf_tag_packet(m, pd->pf_mtag, tag, rtableid)) { REASON_SET(&reason, PFRES_MEMORY); return (PF_DROP); } if (r->keep_state || nr != NULL) { /* create new state */ struct pf_state *s = NULL; struct pf_src_node *sn = NULL; /* check maximums */ if (r->max_states && (r->states >= r->max_states)) { pf_status.lcounters[LCNT_STATES]++; REASON_SET(&reason, PFRES_MAXSTATES); goto cleanup; } /* src node for filter rule */ if ((r->rule_flag & PFRULE_SRCTRACK || r->rpool.opts & PF_POOL_STICKYADDR) && pf_insert_src_node(&sn, r, saddr, af) != 0) { REASON_SET(&reason, PFRES_SRCLIMIT); goto cleanup; } /* src node for translation rule */ if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) && ((direction == PF_OUT && pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) || (pf_insert_src_node(&nsn, nr, saddr, af) != 0))) { REASON_SET(&reason, PFRES_SRCLIMIT); goto cleanup; } s = pool_get(&pf_state_pl, PR_NOWAIT); if (s == NULL) { REASON_SET(&reason, PFRES_MEMORY); cleanup: if (sn != NULL && sn->states == 0 && sn->expire == 0) { RB_REMOVE(pf_src_tree, &tree_src_tracking, sn); pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; pf_status.src_nodes--; pool_put(&pf_src_tree_pl, sn); } if (nsn != sn && nsn != NULL && nsn->states == 0 && nsn->expire == 0) { RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn); pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; pf_status.src_nodes--; pool_put(&pf_src_tree_pl, nsn); } return (PF_DROP); } bzero(s, sizeof(*s)); s->rule.ptr = r; s->nat_rule.ptr = nr; s->anchor.ptr = a; STATE_INC_COUNTERS(s); if (r->allow_opts) s->state_flags |= PFSTATE_ALLOWOPTS; if (r->rule_flag & PFRULE_STATESLOPPY) s->state_flags |= PFSTATE_SLOPPY; s->log = r->log & PF_LOG_ALL; if (nr != NULL) s->log |= nr->log & PF_LOG_ALL; s->proto = IPPROTO_UDP; s->direction = direction; s->af = af; if (direction == PF_OUT) { PF_ACPY(&s->gwy.addr, saddr, af); s->gwy.port = uh->uh_sport; PF_ACPY(&s->ext.addr, daddr, af); s->ext.port = uh->uh_dport; if (nr != NULL) { PF_ACPY(&s->lan.addr, &pd->baddr, af); s->lan.port = bport; } else { PF_ACPY(&s->lan.addr, &s->gwy.addr, af); s->lan.port = s->gwy.port; } } else { PF_ACPY(&s->lan.addr, daddr, af); s->lan.port = uh->uh_dport; PF_ACPY(&s->ext.addr, saddr, af); s->ext.port = uh->uh_sport; if (nr != NULL) { PF_ACPY(&s->gwy.addr, &pd->baddr, af); s->gwy.port = bport; } else { PF_ACPY(&s->gwy.addr, &s->lan.addr, af); s->gwy.port = s->lan.port; } } s->src.state = PFUDPS_SINGLE; s->dst.state = PFUDPS_NO_TRAFFIC; s->creation = time_second; s->expire = time_second; s->timeout = PFTM_UDP_FIRST_PACKET; pf_set_rt_ifp(s, saddr); if (sn != NULL) { s->src_node = sn; s->src_node->states++; } if (nsn != NULL) { PF_ACPY(&nsn->raddr, &pd->naddr, af); s->nat_src_node = nsn; s->nat_src_node->states++; } if (pf_insert_state(BOUND_IFACE(r, kif), s)) { REASON_SET(&reason, PFRES_STATEINS); pf_src_tree_remove_state(s); STATE_DEC_COUNTERS(s); pool_put(&pf_state_pl, s); return (PF_DROP); } else *sm = s; if (tag > 0) { pf_tag_ref(tag); s->tag = tag; } } /* copy back packet headers if we performed NAT operations */ if (rewrite) m_copyback(m, off, sizeof(*uh), (caddr_t)uh); return (PF_PASS); } int pf_test_icmp(struct pf_rule **rm, struct pf_state **sm, int direction, struct pfi_kif *kif, struct mbuf *m, int off, void *h, struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm, struct ifqueue *ifq) { struct pf_rule *nr = NULL; struct pf_addr *saddr = pd->src, *daddr = pd->dst; struct pf_rule *r, *a = NULL; struct pf_ruleset *ruleset = NULL; struct pf_src_node *nsn = NULL; u_short reason; u_int16_t icmpid = 0, bport, nport = 0; sa_family_t af = pd->af; u_int8_t icmptype = 0; /* make the compiler happy */ u_int8_t icmpcode = 0; /* make the compiler happy */ int state_icmp = 0; int tag = -1, rtableid = -1; #ifdef INET6 int rewrite = 0; #endif /* INET6 */ int asd = 0; int match = 0; if (pf_check_congestion(ifq)) { REASON_SET(&reason, PFRES_CONGEST); return (PF_DROP); } switch (pd->proto) { #ifdef INET case IPPROTO_ICMP: icmptype = pd->hdr.icmp->icmp_type; icmpcode = pd->hdr.icmp->icmp_code; icmpid = pd->hdr.icmp->icmp_id; if (icmptype == ICMP_UNREACH || icmptype == ICMP_SOURCEQUENCH || icmptype == ICMP_REDIRECT || icmptype == ICMP_TIMXCEED || icmptype == ICMP_PARAMPROB) state_icmp++; break; #endif /* INET */ #ifdef INET6 case IPPROTO_ICMPV6: icmptype = pd->hdr.icmp6->icmp6_type; icmpcode = pd->hdr.icmp6->icmp6_code; icmpid = pd->hdr.icmp6->icmp6_id; if (icmptype == ICMP6_DST_UNREACH || icmptype == ICMP6_PACKET_TOO_BIG || icmptype == ICMP6_TIME_EXCEEDED || icmptype == ICMP6_PARAM_PROB) state_icmp++; break; #endif /* INET6 */ } r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr); if (direction == PF_OUT) { bport = nport = icmpid; /* check outgoing packet for BINAT/NAT */ if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn, saddr, icmpid, daddr, icmpid, &pd->naddr, &nport)) != NULL) { PF_ACPY(&pd->baddr, saddr, af); switch (af) { #ifdef INET case AF_INET: pf_change_a(&saddr->v4.s_addr, pd->ip_sum, pd->naddr.v4.s_addr, 0); pd->hdr.icmp->icmp_cksum = pf_cksum_fixup( pd->hdr.icmp->icmp_cksum, icmpid, nport, 0); pd->hdr.icmp->icmp_id = nport; m_copyback(m, off, ICMP_MINLEN, (caddr_t)pd->hdr.icmp); break; #endif /* INET */ #ifdef INET6 case AF_INET6: pf_change_a6(saddr, &pd->hdr.icmp6->icmp6_cksum, &pd->naddr, 0); rewrite++; break; #endif /* INET6 */ } if (nr->natpass) r = NULL; pd->nat_rule = nr; } } else { bport = nport = icmpid; /* check incoming packet for BINAT/RDR */ if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn, saddr, icmpid, daddr, icmpid, &pd->naddr, &nport)) != NULL) { PF_ACPY(&pd->baddr, daddr, af); switch (af) { #ifdef INET case AF_INET: pf_change_a(&daddr->v4.s_addr, pd->ip_sum, pd->naddr.v4.s_addr, 0); break; #endif /* INET */ #ifdef INET6 case AF_INET6: pf_change_a6(daddr, &pd->hdr.icmp6->icmp6_cksum, &pd->naddr, 0); rewrite++; break; #endif /* INET6 */ } if (nr->natpass) r = NULL; pd->nat_rule = nr; } } while (r != NULL) { r->evaluations++; if (pfi_kif_match(r->kif, kif) == r->ifnot) r = r->skip[PF_SKIP_IFP].ptr; else if (r->direction && r->direction != direction) r = r->skip[PF_SKIP_DIR].ptr; else if (r->af && r->af != af) r = r->skip[PF_SKIP_AF].ptr; else if (r->proto && r->proto != pd->proto) r = r->skip[PF_SKIP_PROTO].ptr; else if (PF_MISMATCHAW(&r->src.addr, saddr, af, r->src.neg, kif)) r = r->skip[PF_SKIP_SRC_ADDR].ptr; else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, r->dst.neg, NULL)) r = r->skip[PF_SKIP_DST_ADDR].ptr; else if (r->type && r->type != icmptype + 1) r = TAILQ_NEXT(r, entries); else if (r->code && r->code != icmpcode + 1) r = TAILQ_NEXT(r, entries); else if (r->tos && !(r->tos == pd->tos)) r = TAILQ_NEXT(r, entries); else if (r->rule_flag & PFRULE_FRAGMENT) r = TAILQ_NEXT(r, entries); else if (r->prob && r->prob <= arc4random()) r = TAILQ_NEXT(r, entries); else if (r->match_tag && !pf_match_tag(m, r, pd->pf_mtag, &tag)) r = TAILQ_NEXT(r, entries); else if (r->os_fingerprint != PF_OSFP_ANY) r = TAILQ_NEXT(r, entries); else { if (r->tag) tag = r->tag; if (r->rtableid >= 0) rtableid = r->rtableid; if (r->anchor == NULL) { match = 1; *rm = r; *am = a; *rsm = ruleset; if ((*rm)->quick) break; r = TAILQ_NEXT(r, entries); } else pf_step_into_anchor(&asd, &ruleset, PF_RULESET_FILTER, &r, &a, &match); } if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset, PF_RULESET_FILTER, &r, &a, &match)) break; } r = *rm; a = *am; ruleset = *rsm; REASON_SET(&reason, PFRES_MATCH); if (r->log || (nr != NULL && nr->natpass && nr->log)) { #ifdef INET6 if (rewrite) m_copyback(m, off, sizeof(struct icmp6_hdr), (caddr_t)pd->hdr.icmp6); #endif /* INET6 */ PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr, a, ruleset, pd); } if (r->action != PF_PASS) return (PF_DROP); if (pf_tag_packet(m, pd->pf_mtag, tag, rtableid)) { REASON_SET(&reason, PFRES_MEMORY); return (PF_DROP); } if (!state_icmp && (r->keep_state || nr != NULL)) { /* create new state */ struct pf_state *s = NULL; struct pf_src_node *sn = NULL; /* check maximums */ if (r->max_states && (r->states >= r->max_states)) { pf_status.lcounters[LCNT_STATES]++; REASON_SET(&reason, PFRES_MAXSTATES); goto cleanup; } /* src node for filter rule */ if ((r->rule_flag & PFRULE_SRCTRACK || r->rpool.opts & PF_POOL_STICKYADDR) && pf_insert_src_node(&sn, r, saddr, af) != 0) { REASON_SET(&reason, PFRES_SRCLIMIT); goto cleanup; } /* src node for translation rule */ if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) && ((direction == PF_OUT && pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) || (pf_insert_src_node(&nsn, nr, saddr, af) != 0))) { REASON_SET(&reason, PFRES_SRCLIMIT); goto cleanup; } s = pool_get(&pf_state_pl, PR_NOWAIT); if (s == NULL) { REASON_SET(&reason, PFRES_MEMORY); cleanup: if (sn != NULL && sn->states == 0 && sn->expire == 0) { RB_REMOVE(pf_src_tree, &tree_src_tracking, sn); pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; pf_status.src_nodes--; pool_put(&pf_src_tree_pl, sn); } if (nsn != sn && nsn != NULL && nsn->states == 0 && nsn->expire == 0) { RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn); pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; pf_status.src_nodes--; pool_put(&pf_src_tree_pl, nsn); } return (PF_DROP); } bzero(s, sizeof(*s)); s->rule.ptr = r; s->nat_rule.ptr = nr; s->anchor.ptr = a; STATE_INC_COUNTERS(s); if (r->allow_opts) s->state_flags |= PFSTATE_ALLOWOPTS; if (r->rule_flag & PFRULE_STATESLOPPY) s->state_flags |= PFSTATE_SLOPPY; s->log = r->log & PF_LOG_ALL; if (nr != NULL) s->log |= nr->log & PF_LOG_ALL; s->proto = pd->proto; s->direction = direction; s->af = af; if (direction == PF_OUT) { PF_ACPY(&s->gwy.addr, saddr, af); s->gwy.port = nport; PF_ACPY(&s->ext.addr, daddr, af); s->ext.port = 0; if (nr != NULL) { PF_ACPY(&s->lan.addr, &pd->baddr, af); s->lan.port = bport; } else { PF_ACPY(&s->lan.addr, &s->gwy.addr, af); s->lan.port = s->gwy.port; } } else { PF_ACPY(&s->lan.addr, daddr, af); s->lan.port = nport; PF_ACPY(&s->ext.addr, saddr, af); s->ext.port = 0; if (nr != NULL) { PF_ACPY(&s->gwy.addr, &pd->baddr, af); s->gwy.port = bport; } else { PF_ACPY(&s->gwy.addr, &s->lan.addr, af); s->gwy.port = s->lan.port; } } s->creation = time_second; s->expire = time_second; s->timeout = PFTM_ICMP_FIRST_PACKET; pf_set_rt_ifp(s, saddr); if (sn != NULL) { s->src_node = sn; s->src_node->states++; } if (nsn != NULL) { PF_ACPY(&nsn->raddr, &pd->naddr, af); s->nat_src_node = nsn; s->nat_src_node->states++; } if (pf_insert_state(BOUND_IFACE(r, kif), s)) { REASON_SET(&reason, PFRES_STATEINS); pf_src_tree_remove_state(s); STATE_DEC_COUNTERS(s); pool_put(&pf_state_pl, s); return (PF_DROP); } else *sm = s; if (tag > 0) { pf_tag_ref(tag); s->tag = tag; } } #ifdef INET6 /* copy back packet headers if we performed IPv6 NAT operations */ if (rewrite) m_copyback(m, off, sizeof(struct icmp6_hdr), (caddr_t)pd->hdr.icmp6); #endif /* INET6 */ return (PF_PASS); } int pf_test_other(struct pf_rule **rm, struct pf_state **sm, int direction, struct pfi_kif *kif, struct mbuf *m, int off, void *h, struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm, struct ifqueue *ifq) { struct pf_rule *nr = NULL; struct pf_rule *r, *a = NULL; struct pf_ruleset *ruleset = NULL; struct pf_src_node *nsn = NULL; struct pf_addr *saddr = pd->src, *daddr = pd->dst; sa_family_t af = pd->af; u_short reason; int tag = -1, rtableid = -1; int asd = 0; int match = 0; if (pf_check_congestion(ifq)) { REASON_SET(&reason, PFRES_CONGEST); return (PF_DROP); } r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr); if (direction == PF_OUT) { /* check outgoing packet for BINAT/NAT */ if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn, saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) { PF_ACPY(&pd->baddr, saddr, af); switch (af) { #ifdef INET case AF_INET: pf_change_a(&saddr->v4.s_addr, pd->ip_sum, pd->naddr.v4.s_addr, 0); break; #endif /* INET */ #ifdef INET6 case AF_INET6: PF_ACPY(saddr, &pd->naddr, af); break; #endif /* INET6 */ } if (nr->natpass) r = NULL; pd->nat_rule = nr; } } else { /* check incoming packet for BINAT/RDR */ if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn, saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) { PF_ACPY(&pd->baddr, daddr, af); switch (af) { #ifdef INET case AF_INET: pf_change_a(&daddr->v4.s_addr, pd->ip_sum, pd->naddr.v4.s_addr, 0); break; #endif /* INET */ #ifdef INET6 case AF_INET6: PF_ACPY(daddr, &pd->naddr, af); break; #endif /* INET6 */ } if (nr->natpass) r = NULL; pd->nat_rule = nr; } } while (r != NULL) { r->evaluations++; if (pfi_kif_match(r->kif, kif) == r->ifnot) r = r->skip[PF_SKIP_IFP].ptr; else if (r->direction && r->direction != direction) r = r->skip[PF_SKIP_DIR].ptr; else if (r->af && r->af != af) r = r->skip[PF_SKIP_AF].ptr; else if (r->proto && r->proto != pd->proto) r = r->skip[PF_SKIP_PROTO].ptr; else if (PF_MISMATCHAW(&r->src.addr, pd->src, af, r->src.neg, kif)) r = r->skip[PF_SKIP_SRC_ADDR].ptr; else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af, r->dst.neg, NULL)) r = r->skip[PF_SKIP_DST_ADDR].ptr; else if (r->tos && !(r->tos == pd->tos)) r = TAILQ_NEXT(r, entries); else if (r->rule_flag & PFRULE_FRAGMENT) r = TAILQ_NEXT(r, entries); else if (r->prob && r->prob <= arc4random()) r = TAILQ_NEXT(r, entries); else if (r->match_tag && !pf_match_tag(m, r, pd->pf_mtag, &tag)) r = TAILQ_NEXT(r, entries); else if (r->os_fingerprint != PF_OSFP_ANY) r = TAILQ_NEXT(r, entries); else { if (r->tag) tag = r->tag; if (r->rtableid >= 0) rtableid = r->rtableid; if (r->anchor == NULL) { match = 1; *rm = r; *am = a; *rsm = ruleset; if ((*rm)->quick) break; r = TAILQ_NEXT(r, entries); } else pf_step_into_anchor(&asd, &ruleset, PF_RULESET_FILTER, &r, &a, &match); } if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset, PF_RULESET_FILTER, &r, &a, &match)) break; } r = *rm; a = *am; ruleset = *rsm; REASON_SET(&reason, PFRES_MATCH); if (r->log || (nr != NULL && nr->natpass && nr->log)) PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr, a, ruleset, pd); if ((r->action == PF_DROP) && ((r->rule_flag & PFRULE_RETURNICMP) || (r->rule_flag & PFRULE_RETURN))) { struct pf_addr *a = NULL; if (nr != NULL) { if (direction == PF_OUT) a = saddr; else a = daddr; } if (a != NULL) { switch (af) { #ifdef INET case AF_INET: pf_change_a(&a->v4.s_addr, pd->ip_sum, pd->baddr.v4.s_addr, 0); break; #endif /* INET */ #ifdef INET6 case AF_INET6: PF_ACPY(a, &pd->baddr, af); break; #endif /* INET6 */ } } if ((af == AF_INET) && r->return_icmp) pf_send_icmp(m, r->return_icmp >> 8, r->return_icmp & 255, af, r); else if ((af == AF_INET6) && r->return_icmp6) pf_send_icmp(m, r->return_icmp6 >> 8, r->return_icmp6 & 255, af, r); } if (r->action != PF_PASS) return (PF_DROP); if (pf_tag_packet(m, pd->pf_mtag, tag, rtableid)) { REASON_SET(&reason, PFRES_MEMORY); return (PF_DROP); } if (r->keep_state || nr != NULL) { /* create new state */ struct pf_state *s = NULL; struct pf_src_node *sn = NULL; /* check maximums */ if (r->max_states && (r->states >= r->max_states)) { pf_status.lcounters[LCNT_STATES]++; REASON_SET(&reason, PFRES_MAXSTATES); goto cleanup; } /* src node for filter rule */ if ((r->rule_flag & PFRULE_SRCTRACK || r->rpool.opts & PF_POOL_STICKYADDR) && pf_insert_src_node(&sn, r, saddr, af) != 0) { REASON_SET(&reason, PFRES_SRCLIMIT); goto cleanup; } /* src node for translation rule */ if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) && ((direction == PF_OUT && pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) || (pf_insert_src_node(&nsn, nr, saddr, af) != 0))) { REASON_SET(&reason, PFRES_SRCLIMIT); goto cleanup; } s = pool_get(&pf_state_pl, PR_NOWAIT); if (s == NULL) { REASON_SET(&reason, PFRES_MEMORY); cleanup: if (sn != NULL && sn->states == 0 && sn->expire == 0) { RB_REMOVE(pf_src_tree, &tree_src_tracking, sn); pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; pf_status.src_nodes--; pool_put(&pf_src_tree_pl, sn); } if (nsn != sn && nsn != NULL && nsn->states == 0 && nsn->expire == 0) { RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn); pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; pf_status.src_nodes--; pool_put(&pf_src_tree_pl, nsn); } return (PF_DROP); } bzero(s, sizeof(*s)); s->rule.ptr = r; s->nat_rule.ptr = nr; s->anchor.ptr = a; STATE_INC_COUNTERS(s); if (r->allow_opts) s->state_flags |= PFSTATE_ALLOWOPTS; if (r->rule_flag & PFRULE_STATESLOPPY) s->state_flags |= PFSTATE_SLOPPY; s->log = r->log & PF_LOG_ALL; if (nr != NULL) s->log |= nr->log & PF_LOG_ALL; s->proto = pd->proto; s->direction = direction; s->af = af; if (direction == PF_OUT) { PF_ACPY(&s->gwy.addr, saddr, af); PF_ACPY(&s->ext.addr, daddr, af); if (nr != NULL) PF_ACPY(&s->lan.addr, &pd->baddr, af); else PF_ACPY(&s->lan.addr, &s->gwy.addr, af); } else { PF_ACPY(&s->lan.addr, daddr, af); PF_ACPY(&s->ext.addr, saddr, af); if (nr != NULL) PF_ACPY(&s->gwy.addr, &pd->baddr, af); else PF_ACPY(&s->gwy.addr, &s->lan.addr, af); } s->src.state = PFOTHERS_SINGLE; s->dst.state = PFOTHERS_NO_TRAFFIC; s->creation = time_second; s->expire = time_second; s->timeout = PFTM_OTHER_FIRST_PACKET; pf_set_rt_ifp(s, saddr); if (sn != NULL) { s->src_node = sn; s->src_node->states++; } if (nsn != NULL) { PF_ACPY(&nsn->raddr, &pd->naddr, af); s->nat_src_node = nsn; s->nat_src_node->states++; } if (pf_insert_state(BOUND_IFACE(r, kif), s)) { REASON_SET(&reason, PFRES_STATEINS); pf_src_tree_remove_state(s); STATE_DEC_COUNTERS(s); pool_put(&pf_state_pl, s); return (PF_DROP); } else *sm = s; if (tag > 0) { pf_tag_ref(tag); s->tag = tag; } } return (PF_PASS); } int pf_test_fragment(struct pf_rule **rm, int direction, struct pfi_kif *kif, struct mbuf *m, void *h, struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm) { struct pf_rule *r, *a = NULL; struct pf_ruleset *ruleset = NULL; sa_family_t af = pd->af; u_short reason; int tag = -1; int asd = 0; int match = 0; r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr); while (r != NULL) { r->evaluations++; if (pfi_kif_match(r->kif, kif) == r->ifnot) r = r->skip[PF_SKIP_IFP].ptr; else if (r->direction && r->direction != direction) r = r->skip[PF_SKIP_DIR].ptr; else if (r->af && r->af != af) r = r->skip[PF_SKIP_AF].ptr; else if (r->proto && r->proto != pd->proto) r = r->skip[PF_SKIP_PROTO].ptr; else if (PF_MISMATCHAW(&r->src.addr, pd->src, af, r->src.neg, kif)) r = r->skip[PF_SKIP_SRC_ADDR].ptr; else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af, r->dst.neg, NULL)) r = r->skip[PF_SKIP_DST_ADDR].ptr; else if (r->tos && !(r->tos == pd->tos)) r = TAILQ_NEXT(r, entries); else if (r->os_fingerprint != PF_OSFP_ANY) r = TAILQ_NEXT(r, entries); else if (pd->proto == IPPROTO_UDP && (r->src.port_op || r->dst.port_op)) r = TAILQ_NEXT(r, entries); else if (pd->proto == IPPROTO_TCP && (r->src.port_op || r->dst.port_op || r->flagset)) r = TAILQ_NEXT(r, entries); else if ((pd->proto == IPPROTO_ICMP || pd->proto == IPPROTO_ICMPV6) && (r->type || r->code)) r = TAILQ_NEXT(r, entries); else if (r->prob && r->prob <= arc4random()) r = TAILQ_NEXT(r, entries); else if (r->match_tag && !pf_match_tag(m, r, pd->pf_mtag, &tag)) r = TAILQ_NEXT(r, entries); else { if (r->anchor == NULL) { match = 1; *rm = r; *am = a; *rsm = ruleset; if ((*rm)->quick) break; r = TAILQ_NEXT(r, entries); } else pf_step_into_anchor(&asd, &ruleset, PF_RULESET_FILTER, &r, &a, &match); } if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset, PF_RULESET_FILTER, &r, &a, &match)) break; } r = *rm; a = *am; ruleset = *rsm; REASON_SET(&reason, PFRES_MATCH); if (r->log) PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset, pd); if (r->action != PF_PASS) return (PF_DROP); if (pf_tag_packet(m, pd->pf_mtag, tag, -1)) { REASON_SET(&reason, PFRES_MEMORY); return (PF_DROP); } return (PF_PASS); } int pf_tcp_track_full(struct pf_state_peer *src, struct pf_state_peer *dst, struct pf_state **state, struct pfi_kif *kif, struct mbuf *m, int off, struct pf_pdesc *pd, u_short *reason, int *copyback) { struct tcphdr *th = pd->hdr.tcp; u_int16_t win = ntohs(th->th_win); u_int32_t ack, end, seq, orig_seq; u_int8_t sws, dws; int ackskew; if (src->wscale && dst->wscale && !(th->th_flags & TH_SYN)) { sws = src->wscale & PF_WSCALE_MASK; dws = dst->wscale & PF_WSCALE_MASK; } else sws = dws = 0; /* * Sequence tracking algorithm from Guido van Rooij's paper: * http://www.madison-gurkha.com/publications/tcp_filtering/ * tcp_filtering.ps */ orig_seq = seq = ntohl(th->th_seq); if (src->seqlo == 0) { /* First packet from this end. Set its state */ if ((pd->flags & PFDESC_TCP_NORM || dst->scrub) && src->scrub == NULL) { if (pf_normalize_tcp_init(m, off, pd, th, src, dst)) { REASON_SET(reason, PFRES_MEMORY); return (PF_DROP); } } /* Deferred generation of sequence number modulator */ if (dst->seqdiff && !src->seqdiff) { #ifdef __FreeBSD__ while ((src->seqdiff = pf_new_isn(*state) - seq) == 0) ; #else while ((src->seqdiff = tcp_rndiss_next() - seq) == 0) ; #endif ack = ntohl(th->th_ack) - dst->seqdiff; pf_change_a(&th->th_seq, &th->th_sum, htonl(seq + src->seqdiff), 0); pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0); *copyback = 1; } else { ack = ntohl(th->th_ack); } end = seq + pd->p_len; if (th->th_flags & TH_SYN) { end++; if (dst->wscale & PF_WSCALE_FLAG) { src->wscale = pf_get_wscale(m, off, th->th_off, pd->af); if (src->wscale & PF_WSCALE_FLAG) { /* Remove scale factor from initial * window */ sws = src->wscale & PF_WSCALE_MASK; win = ((u_int32_t)win + (1 << sws) - 1) >> sws; dws = dst->wscale & PF_WSCALE_MASK; } else { /* fixup other window */ dst->max_win <<= dst->wscale & PF_WSCALE_MASK; /* in case of a retrans SYN|ACK */ dst->wscale = 0; } } } if (th->th_flags & TH_FIN) end++; src->seqlo = seq; if (src->state < TCPS_SYN_SENT) src->state = TCPS_SYN_SENT; /* * May need to slide the window (seqhi may have been set by * the crappy stack check or if we picked up the connection * after establishment) */ if (src->seqhi == 1 || SEQ_GEQ(end + MAX(1, dst->max_win << dws), src->seqhi)) src->seqhi = end + MAX(1, dst->max_win << dws); if (win > src->max_win) src->max_win = win; } else { ack = ntohl(th->th_ack) - dst->seqdiff; if (src->seqdiff) { /* Modulate sequence numbers */ pf_change_a(&th->th_seq, &th->th_sum, htonl(seq + src->seqdiff), 0); pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0); *copyback = 1; } end = seq + pd->p_len; if (th->th_flags & TH_SYN) end++; if (th->th_flags & TH_FIN) end++; } if ((th->th_flags & TH_ACK) == 0) { /* Let it pass through the ack skew check */ ack = dst->seqlo; } else if ((ack == 0 && (th->th_flags & (TH_ACK|TH_RST)) == (TH_ACK|TH_RST)) || /* broken tcp stacks do not set ack */ (dst->state < TCPS_SYN_SENT)) { /* * Many stacks (ours included) will set the ACK number in an * FIN|ACK if the SYN times out -- no sequence to ACK. */ ack = dst->seqlo; } if (seq == end) { /* Ease sequencing restrictions on no data packets */ seq = src->seqlo; end = seq; } ackskew = dst->seqlo - ack; /* * Need to demodulate the sequence numbers in any TCP SACK options * (Selective ACK). We could optionally validate the SACK values * against the current ACK window, either forwards or backwards, but * I'm not confident that SACK has been implemented properly * everywhere. It wouldn't surprise me if several stacks accidently * SACK too far backwards of previously ACKed data. There really aren't * any security implications of bad SACKing unless the target stack * doesn't validate the option length correctly. Someone trying to * spoof into a TCP connection won't bother blindly sending SACK * options anyway. */ if (dst->seqdiff && (th->th_off << 2) > sizeof(struct tcphdr)) { if (pf_modulate_sack(m, off, pd, th, dst)) *copyback = 1; } #define MAXACKWINDOW (0xffff + 1500) /* 1500 is an arbitrary fudge factor */ if (SEQ_GEQ(src->seqhi, end) && /* Last octet inside other's window space */ SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) && /* Retrans: not more than one window back */ (ackskew >= -MAXACKWINDOW) && /* Acking not more than one reassembled fragment backwards */ (ackskew <= (MAXACKWINDOW << sws)) && /* Acking not more than one window forward */ ((th->th_flags & TH_RST) == 0 || orig_seq == src->seqlo || (orig_seq == src->seqlo + 1) || (pd->flags & PFDESC_IP_REAS) == 0)) { /* Require an exact/+1 sequence match on resets when possible */ if (dst->scrub || src->scrub) { if (pf_normalize_tcp_stateful(m, off, pd, reason, th, *state, src, dst, copyback)) return (PF_DROP); } /* update max window */ if (src->max_win < win) src->max_win = win; /* synchronize sequencing */ if (SEQ_GT(end, src->seqlo)) src->seqlo = end; /* slide the window of what the other end can send */ if (SEQ_GEQ(ack + (win << sws), dst->seqhi)) dst->seqhi = ack + MAX((win << sws), 1); /* update states */ if (th->th_flags & TH_SYN) if (src->state < TCPS_SYN_SENT) src->state = TCPS_SYN_SENT; if (th->th_flags & TH_FIN) if (src->state < TCPS_CLOSING) src->state = TCPS_CLOSING; if (th->th_flags & TH_ACK) { if (dst->state == TCPS_SYN_SENT) { dst->state = TCPS_ESTABLISHED; if (src->state == TCPS_ESTABLISHED && (*state)->src_node != NULL && pf_src_connlimit(state)) { REASON_SET(reason, PFRES_SRCLIMIT); return (PF_DROP); } } else if (dst->state == TCPS_CLOSING) dst->state = TCPS_FIN_WAIT_2; } if (th->th_flags & TH_RST) src->state = dst->state = TCPS_TIME_WAIT; /* update expire time */ (*state)->expire = time_second; if (src->state >= TCPS_FIN_WAIT_2 && dst->state >= TCPS_FIN_WAIT_2) (*state)->timeout = PFTM_TCP_CLOSED; else if (src->state >= TCPS_CLOSING && dst->state >= TCPS_CLOSING) (*state)->timeout = PFTM_TCP_FIN_WAIT; else if (src->state < TCPS_ESTABLISHED || dst->state < TCPS_ESTABLISHED) (*state)->timeout = PFTM_TCP_OPENING; else if (src->state >= TCPS_CLOSING || dst->state >= TCPS_CLOSING) (*state)->timeout = PFTM_TCP_CLOSING; else (*state)->timeout = PFTM_TCP_ESTABLISHED; /* Fall through to PASS packet */ } else if ((dst->state < TCPS_SYN_SENT || dst->state >= TCPS_FIN_WAIT_2 || src->state >= TCPS_FIN_WAIT_2) && SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) && /* Within a window forward of the originating packet */ SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW)) { /* Within a window backward of the originating packet */ /* * This currently handles three situations: * 1) Stupid stacks will shotgun SYNs before their peer * replies. * 2) When PF catches an already established stream (the * firewall rebooted, the state table was flushed, routes * changed...) * 3) Packets get funky immediately after the connection * closes (this should catch Solaris spurious ACK|FINs * that web servers like to spew after a close) * * This must be a little more careful than the above code * since packet floods will also be caught here. We don't * update the TTL here to mitigate the damage of a packet * flood and so the same code can handle awkward establishment * and a loosened connection close. * In the establishment case, a correct peer response will * validate the connection, go through the normal state code * and keep updating the state TTL. */ if (pf_status.debug >= PF_DEBUG_MISC) { printf("pf: loose state match: "); pf_print_state(*state); pf_print_flags(th->th_flags); printf(" seq=%u (%u) ack=%u len=%u ackskew=%d " "pkts=%llu:%llu\n", seq, orig_seq, ack, pd->p_len, #ifdef __FreeBSD__ ackskew, (unsigned long long)(*state)->packets[0], (unsigned long long)(*state)->packets[1]); #else ackskew, (*state)->packets[0], (*state)->packets[1]); #endif } if (dst->scrub || src->scrub) { if (pf_normalize_tcp_stateful(m, off, pd, reason, th, *state, src, dst, copyback)) return (PF_DROP); } /* update max window */ if (src->max_win < win) src->max_win = win; /* synchronize sequencing */ if (SEQ_GT(end, src->seqlo)) src->seqlo = end; /* slide the window of what the other end can send */ if (SEQ_GEQ(ack + (win << sws), dst->seqhi)) dst->seqhi = ack + MAX((win << sws), 1); /* * Cannot set dst->seqhi here since this could be a shotgunned * SYN and not an already established connection. */ if (th->th_flags & TH_FIN) if (src->state < TCPS_CLOSING) src->state = TCPS_CLOSING; if (th->th_flags & TH_RST) src->state = dst->state = TCPS_TIME_WAIT; /* Fall through to PASS packet */ } else { if ((*state)->dst.state == TCPS_SYN_SENT && (*state)->src.state == TCPS_SYN_SENT) { /* Send RST for state mismatches during handshake */ if (!(th->th_flags & TH_RST)) #ifdef __FreeBSD__ pf_send_tcp(m, (*state)->rule.ptr, pd->af, #else pf_send_tcp((*state)->rule.ptr, pd->af, #endif pd->dst, pd->src, th->th_dport, th->th_sport, ntohl(th->th_ack), 0, TH_RST, 0, 0, (*state)->rule.ptr->return_ttl, 1, 0, pd->eh, kif->pfik_ifp); src->seqlo = 0; src->seqhi = 1; src->max_win = 1; } else if (pf_status.debug >= PF_DEBUG_MISC) { printf("pf: BAD state: "); pf_print_state(*state); pf_print_flags(th->th_flags); printf(" seq=%u (%u) ack=%u len=%u ackskew=%d " #ifdef notyet "pkts=%llu:%llu dir=%s,%s\n", #else "pkts=%llu:%llu%s\n", #endif seq, orig_seq, ack, pd->p_len, ackskew, #ifdef __FreeBSD__ (unsigned long long)(*state)->packets[0], (unsigned long long)(*state)->packets[1], #else (*state)->packets[0], (*state)->packets[1], #endif #ifdef notyet direction == PF_IN ? "in" : "out", direction == (*state)->direction ? "fwd" : "rev"); #else ""); #endif printf("pf: State failure on: %c %c %c %c | %c %c\n", SEQ_GEQ(src->seqhi, end) ? ' ' : '1', SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) ? ' ': '2', (ackskew >= -MAXACKWINDOW) ? ' ' : '3', (ackskew <= (MAXACKWINDOW << sws)) ? ' ' : '4', SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) ?' ' :'5', SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW) ?' ' :'6'); } REASON_SET(reason, PFRES_BADSTATE); return (PF_DROP); } /* Any packets which have gotten here are to be passed */ return (PF_PASS); } int pf_tcp_track_sloppy(struct pf_state_peer *src, struct pf_state_peer *dst, struct pf_state **state, struct pf_pdesc *pd, u_short *reason) { struct tcphdr *th = pd->hdr.tcp; if (th->th_flags & TH_SYN) if (src->state < TCPS_SYN_SENT) src->state = TCPS_SYN_SENT; if (th->th_flags & TH_FIN) if (src->state < TCPS_CLOSING) src->state = TCPS_CLOSING; if (th->th_flags & TH_ACK) { if (dst->state == TCPS_SYN_SENT) { dst->state = TCPS_ESTABLISHED; if (src->state == TCPS_ESTABLISHED && (*state)->src_node != NULL && pf_src_connlimit(state)) { REASON_SET(reason, PFRES_SRCLIMIT); return (PF_DROP); } } else if (dst->state == TCPS_CLOSING) { dst->state = TCPS_FIN_WAIT_2; } else if (src->state == TCPS_SYN_SENT && dst->state < TCPS_SYN_SENT) { /* * Handle a special sloppy case where we only see one * half of the connection. If there is a ACK after * the initial SYN without ever seeing a packet from * the destination, set the connection to established. */ dst->state = src->state = TCPS_ESTABLISHED; if ((*state)->src_node != NULL && pf_src_connlimit(state)) { REASON_SET(reason, PFRES_SRCLIMIT); return (PF_DROP); } } else if (src->state == TCPS_CLOSING && dst->state == TCPS_ESTABLISHED && dst->seqlo == 0) { /* * Handle the closing of half connections where we * don't see the full bidirectional FIN/ACK+ACK * handshake. */ dst->state = TCPS_CLOSING; } } if (th->th_flags & TH_RST) src->state = dst->state = TCPS_TIME_WAIT; /* update expire time */ (*state)->expire = time_second; if (src->state >= TCPS_FIN_WAIT_2 && dst->state >= TCPS_FIN_WAIT_2) (*state)->timeout = PFTM_TCP_CLOSED; else if (src->state >= TCPS_CLOSING && dst->state >= TCPS_CLOSING) (*state)->timeout = PFTM_TCP_FIN_WAIT; else if (src->state < TCPS_ESTABLISHED || dst->state < TCPS_ESTABLISHED) (*state)->timeout = PFTM_TCP_OPENING; else if (src->state >= TCPS_CLOSING || dst->state >= TCPS_CLOSING) (*state)->timeout = PFTM_TCP_CLOSING; else (*state)->timeout = PFTM_TCP_ESTABLISHED; return (PF_PASS); } int pf_test_state_tcp(struct pf_state **state, int direction, struct pfi_kif *kif, struct mbuf *m, int off, void *h, struct pf_pdesc *pd, u_short *reason) { struct pf_state_cmp key; struct tcphdr *th = pd->hdr.tcp; int copyback = 0; struct pf_state_peer *src, *dst; key.af = pd->af; key.proto = IPPROTO_TCP; if (direction == PF_IN) { PF_ACPY(&key.ext.addr, pd->src, key.af); PF_ACPY(&key.gwy.addr, pd->dst, key.af); key.ext.port = th->th_sport; key.gwy.port = th->th_dport; } else { PF_ACPY(&key.lan.addr, pd->src, key.af); PF_ACPY(&key.ext.addr, pd->dst, key.af); key.lan.port = th->th_sport; key.ext.port = th->th_dport; } STATE_LOOKUP(); if (direction == (*state)->direction) { src = &(*state)->src; dst = &(*state)->dst; } else { src = &(*state)->dst; dst = &(*state)->src; } if ((*state)->src.state == PF_TCPS_PROXY_SRC) { if (direction != (*state)->direction) { REASON_SET(reason, PFRES_SYNPROXY); return (PF_SYNPROXY_DROP); } if (th->th_flags & TH_SYN) { if (ntohl(th->th_seq) != (*state)->src.seqlo) { REASON_SET(reason, PFRES_SYNPROXY); return (PF_DROP); } #ifdef __FreeBSD__ pf_send_tcp(NULL, (*state)->rule.ptr, pd->af, pd->dst, #else pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst, #endif pd->src, th->th_dport, th->th_sport, (*state)->src.seqhi, ntohl(th->th_seq) + 1, TH_SYN|TH_ACK, 0, (*state)->src.mss, 0, 1, 0, NULL, NULL); REASON_SET(reason, PFRES_SYNPROXY); return (PF_SYNPROXY_DROP); } else if (!(th->th_flags & TH_ACK) || (ntohl(th->th_ack) != (*state)->src.seqhi + 1) || (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) { REASON_SET(reason, PFRES_SYNPROXY); return (PF_DROP); } else if ((*state)->src_node != NULL && pf_src_connlimit(state)) { REASON_SET(reason, PFRES_SRCLIMIT); return (PF_DROP); } else (*state)->src.state = PF_TCPS_PROXY_DST; } if ((*state)->src.state == PF_TCPS_PROXY_DST) { struct pf_state_host *src, *dst; if (direction == PF_OUT) { src = &(*state)->gwy; dst = &(*state)->ext; } else { src = &(*state)->ext; dst = &(*state)->lan; } if (direction == (*state)->direction) { if (((th->th_flags & (TH_SYN|TH_ACK)) != TH_ACK) || (ntohl(th->th_ack) != (*state)->src.seqhi + 1) || (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) { REASON_SET(reason, PFRES_SYNPROXY); return (PF_DROP); } (*state)->src.max_win = MAX(ntohs(th->th_win), 1); if ((*state)->dst.seqhi == 1) (*state)->dst.seqhi = htonl(arc4random()); #ifdef __FreeBSD__ pf_send_tcp(NULL, (*state)->rule.ptr, pd->af, &src->addr, #else pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr, #endif &dst->addr, src->port, dst->port, (*state)->dst.seqhi, 0, TH_SYN, 0, (*state)->src.mss, 0, 0, (*state)->tag, NULL, NULL); REASON_SET(reason, PFRES_SYNPROXY); return (PF_SYNPROXY_DROP); } else if (((th->th_flags & (TH_SYN|TH_ACK)) != (TH_SYN|TH_ACK)) || (ntohl(th->th_ack) != (*state)->dst.seqhi + 1)) { REASON_SET(reason, PFRES_SYNPROXY); return (PF_DROP); } else { (*state)->dst.max_win = MAX(ntohs(th->th_win), 1); (*state)->dst.seqlo = ntohl(th->th_seq); #ifdef __FreeBSD__ pf_send_tcp(NULL, (*state)->rule.ptr, pd->af, pd->dst, #else pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst, #endif pd->src, th->th_dport, th->th_sport, ntohl(th->th_ack), ntohl(th->th_seq) + 1, TH_ACK, (*state)->src.max_win, 0, 0, 0, (*state)->tag, NULL, NULL); #ifdef __FreeBSD__ pf_send_tcp(NULL, (*state)->rule.ptr, pd->af, &src->addr, #else pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr, #endif &dst->addr, src->port, dst->port, (*state)->src.seqhi + 1, (*state)->src.seqlo + 1, TH_ACK, (*state)->dst.max_win, 0, 0, 1, 0, NULL, NULL); (*state)->src.seqdiff = (*state)->dst.seqhi - (*state)->src.seqlo; (*state)->dst.seqdiff = (*state)->src.seqhi - (*state)->dst.seqlo; (*state)->src.seqhi = (*state)->src.seqlo + (*state)->dst.max_win; (*state)->dst.seqhi = (*state)->dst.seqlo + (*state)->src.max_win; (*state)->src.wscale = (*state)->dst.wscale = 0; (*state)->src.state = (*state)->dst.state = TCPS_ESTABLISHED; REASON_SET(reason, PFRES_SYNPROXY); return (PF_SYNPROXY_DROP); } } if (((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN) && dst->state >= TCPS_FIN_WAIT_2 && src->state >= TCPS_FIN_WAIT_2) { if (pf_status.debug >= PF_DEBUG_MISC) { printf("pf: state reuse "); pf_print_state(*state); pf_print_flags(th->th_flags); printf("\n"); } /* XXX make sure it's the same direction ?? */ (*state)->src.state = (*state)->dst.state = TCPS_CLOSED; pf_unlink_state(*state); *state = NULL; return (PF_DROP); } if ((*state)->state_flags & PFSTATE_SLOPPY) { if (pf_tcp_track_sloppy(src, dst, state, pd, reason) == PF_DROP) return (PF_DROP); } else { if (pf_tcp_track_full(src, dst, state, kif, m, off, pd, reason, ©back) == PF_DROP) return (PF_DROP); } /* translate source/destination address, if necessary */ if (STATE_TRANSLATE(*state)) { if (direction == PF_OUT) pf_change_ap(pd->src, &th->th_sport, pd->ip_sum, &th->th_sum, &(*state)->gwy.addr, (*state)->gwy.port, 0, pd->af); else pf_change_ap(pd->dst, &th->th_dport, pd->ip_sum, &th->th_sum, &(*state)->lan.addr, (*state)->lan.port, 0, pd->af); m_copyback(m, off, sizeof(*th), (caddr_t)th); } else if (copyback) { /* Copyback sequence modulation or stateful scrub changes */ m_copyback(m, off, sizeof(*th), (caddr_t)th); } return (PF_PASS); } int pf_test_state_udp(struct pf_state **state, int direction, struct pfi_kif *kif, struct mbuf *m, int off, void *h, struct pf_pdesc *pd) { struct pf_state_peer *src, *dst; struct pf_state_cmp key; struct udphdr *uh = pd->hdr.udp; key.af = pd->af; key.proto = IPPROTO_UDP; if (direction == PF_IN) { PF_ACPY(&key.ext.addr, pd->src, key.af); PF_ACPY(&key.gwy.addr, pd->dst, key.af); key.ext.port = uh->uh_sport; key.gwy.port = uh->uh_dport; } else { PF_ACPY(&key.lan.addr, pd->src, key.af); PF_ACPY(&key.ext.addr, pd->dst, key.af); key.lan.port = uh->uh_sport; key.ext.port = uh->uh_dport; } STATE_LOOKUP(); if (direction == (*state)->direction) { src = &(*state)->src; dst = &(*state)->dst; } else { src = &(*state)->dst; dst = &(*state)->src; } /* update states */ if (src->state < PFUDPS_SINGLE) src->state = PFUDPS_SINGLE; if (dst->state == PFUDPS_SINGLE) dst->state = PFUDPS_MULTIPLE; /* update expire time */ (*state)->expire = time_second; if (src->state == PFUDPS_MULTIPLE && dst->state == PFUDPS_MULTIPLE) (*state)->timeout = PFTM_UDP_MULTIPLE; else (*state)->timeout = PFTM_UDP_SINGLE; /* translate source/destination address, if necessary */ if (STATE_TRANSLATE(*state)) { if (direction == PF_OUT) pf_change_ap(pd->src, &uh->uh_sport, pd->ip_sum, &uh->uh_sum, &(*state)->gwy.addr, (*state)->gwy.port, 1, pd->af); else pf_change_ap(pd->dst, &uh->uh_dport, pd->ip_sum, &uh->uh_sum, &(*state)->lan.addr, (*state)->lan.port, 1, pd->af); m_copyback(m, off, sizeof(*uh), (caddr_t)uh); } return (PF_PASS); } int pf_test_state_icmp(struct pf_state **state, int direction, struct pfi_kif *kif, struct mbuf *m, int off, void *h, struct pf_pdesc *pd, u_short *reason) { struct pf_addr *saddr = pd->src, *daddr = pd->dst; u_int16_t icmpid = 0; /* make the compiler happy */ u_int16_t *icmpsum = NULL; /* make the compiler happy */ u_int8_t icmptype = 0; /* make the compiler happy */ int state_icmp = 0; struct pf_state_cmp key; switch (pd->proto) { #ifdef INET case IPPROTO_ICMP: icmptype = pd->hdr.icmp->icmp_type; icmpid = pd->hdr.icmp->icmp_id; icmpsum = &pd->hdr.icmp->icmp_cksum; if (icmptype == ICMP_UNREACH || icmptype == ICMP_SOURCEQUENCH || icmptype == ICMP_REDIRECT || icmptype == ICMP_TIMXCEED || icmptype == ICMP_PARAMPROB) state_icmp++; break; #endif /* INET */ #ifdef INET6 case IPPROTO_ICMPV6: icmptype = pd->hdr.icmp6->icmp6_type; icmpid = pd->hdr.icmp6->icmp6_id; icmpsum = &pd->hdr.icmp6->icmp6_cksum; if (icmptype == ICMP6_DST_UNREACH || icmptype == ICMP6_PACKET_TOO_BIG || icmptype == ICMP6_TIME_EXCEEDED || icmptype == ICMP6_PARAM_PROB) state_icmp++; break; #endif /* INET6 */ } if (!state_icmp) { /* * ICMP query/reply message not related to a TCP/UDP packet. * Search for an ICMP state. */ key.af = pd->af; key.proto = pd->proto; if (direction == PF_IN) { PF_ACPY(&key.ext.addr, pd->src, key.af); PF_ACPY(&key.gwy.addr, pd->dst, key.af); key.ext.port = 0; key.gwy.port = icmpid; } else { PF_ACPY(&key.lan.addr, pd->src, key.af); PF_ACPY(&key.ext.addr, pd->dst, key.af); key.lan.port = icmpid; key.ext.port = 0; } STATE_LOOKUP(); (*state)->expire = time_second; (*state)->timeout = PFTM_ICMP_ERROR_REPLY; /* translate source/destination address, if necessary */ if (STATE_TRANSLATE(*state)) { if (direction == PF_OUT) { switch (pd->af) { #ifdef INET case AF_INET: pf_change_a(&saddr->v4.s_addr, pd->ip_sum, (*state)->gwy.addr.v4.s_addr, 0); pd->hdr.icmp->icmp_cksum = pf_cksum_fixup( pd->hdr.icmp->icmp_cksum, icmpid, (*state)->gwy.port, 0); pd->hdr.icmp->icmp_id = (*state)->gwy.port; m_copyback(m, off, ICMP_MINLEN, (caddr_t)pd->hdr.icmp); break; #endif /* INET */ #ifdef INET6 case AF_INET6: pf_change_a6(saddr, &pd->hdr.icmp6->icmp6_cksum, &(*state)->gwy.addr, 0); m_copyback(m, off, sizeof(struct icmp6_hdr), (caddr_t)pd->hdr.icmp6); break; #endif /* INET6 */ } } else { switch (pd->af) { #ifdef INET case AF_INET: pf_change_a(&daddr->v4.s_addr, pd->ip_sum, (*state)->lan.addr.v4.s_addr, 0); pd->hdr.icmp->icmp_cksum = pf_cksum_fixup( pd->hdr.icmp->icmp_cksum, icmpid, (*state)->lan.port, 0); pd->hdr.icmp->icmp_id = (*state)->lan.port; m_copyback(m, off, ICMP_MINLEN, (caddr_t)pd->hdr.icmp); break; #endif /* INET */ #ifdef INET6 case AF_INET6: pf_change_a6(daddr, &pd->hdr.icmp6->icmp6_cksum, &(*state)->lan.addr, 0); m_copyback(m, off, sizeof(struct icmp6_hdr), (caddr_t)pd->hdr.icmp6); break; #endif /* INET6 */ } } } return (PF_PASS); } else { /* * ICMP error message in response to a TCP/UDP packet. * Extract the inner TCP/UDP header and search for that state. */ struct pf_pdesc pd2; #ifdef INET struct ip h2; #endif /* INET */ #ifdef INET6 struct ip6_hdr h2_6; int terminal = 0; #endif /* INET6 */ int ipoff2 = 0; /* make the compiler happy */ int off2 = 0; /* make the compiler happy */ pd2.af = pd->af; switch (pd->af) { #ifdef INET case AF_INET: /* offset of h2 in mbuf chain */ ipoff2 = off + ICMP_MINLEN; if (!pf_pull_hdr(m, ipoff2, &h2, sizeof(h2), NULL, reason, pd2.af)) { DPFPRINTF(PF_DEBUG_MISC, ("pf: ICMP error message too short " "(ip)\n")); return (PF_DROP); } /* * ICMP error messages don't refer to non-first * fragments */ if (h2.ip_off & htons(IP_OFFMASK)) { REASON_SET(reason, PFRES_FRAG); return (PF_DROP); } /* offset of protocol header that follows h2 */ off2 = ipoff2 + (h2.ip_hl << 2); pd2.proto = h2.ip_p; pd2.src = (struct pf_addr *)&h2.ip_src; pd2.dst = (struct pf_addr *)&h2.ip_dst; pd2.ip_sum = &h2.ip_sum; break; #endif /* INET */ #ifdef INET6 case AF_INET6: ipoff2 = off + sizeof(struct icmp6_hdr); if (!pf_pull_hdr(m, ipoff2, &h2_6, sizeof(h2_6), NULL, reason, pd2.af)) { DPFPRINTF(PF_DEBUG_MISC, ("pf: ICMP error message too short " "(ip6)\n")); return (PF_DROP); } pd2.proto = h2_6.ip6_nxt; pd2.src = (struct pf_addr *)&h2_6.ip6_src; pd2.dst = (struct pf_addr *)&h2_6.ip6_dst; pd2.ip_sum = NULL; off2 = ipoff2 + sizeof(h2_6); do { switch (pd2.proto) { case IPPROTO_FRAGMENT: /* * ICMPv6 error messages for * non-first fragments */ REASON_SET(reason, PFRES_FRAG); return (PF_DROP); case IPPROTO_AH: case IPPROTO_HOPOPTS: case IPPROTO_ROUTING: case IPPROTO_DSTOPTS: { /* get next header and header length */ struct ip6_ext opt6; if (!pf_pull_hdr(m, off2, &opt6, sizeof(opt6), NULL, reason, pd2.af)) { DPFPRINTF(PF_DEBUG_MISC, ("pf: ICMPv6 short opt\n")); return (PF_DROP); } if (pd2.proto == IPPROTO_AH) off2 += (opt6.ip6e_len + 2) * 4; else off2 += (opt6.ip6e_len + 1) * 8; pd2.proto = opt6.ip6e_nxt; /* goto the next header */ break; } default: terminal++; break; } } while (!terminal); break; #endif /* INET6 */ #ifdef __FreeBSD__ default: panic("AF not supported: %d", pd->af); #endif } switch (pd2.proto) { case IPPROTO_TCP: { struct tcphdr th; u_int32_t seq; struct pf_state_peer *src, *dst; u_int8_t dws; int copyback = 0; /* * Only the first 8 bytes of the TCP header can be * expected. Don't access any TCP header fields after * th_seq, an ackskew test is not possible. */ if (!pf_pull_hdr(m, off2, &th, 8, NULL, reason, pd2.af)) { DPFPRINTF(PF_DEBUG_MISC, ("pf: ICMP error message too short " "(tcp)\n")); return (PF_DROP); } key.af = pd2.af; key.proto = IPPROTO_TCP; if (direction == PF_IN) { PF_ACPY(&key.ext.addr, pd2.dst, key.af); PF_ACPY(&key.gwy.addr, pd2.src, key.af); key.ext.port = th.th_dport; key.gwy.port = th.th_sport; } else { PF_ACPY(&key.lan.addr, pd2.dst, key.af); PF_ACPY(&key.ext.addr, pd2.src, key.af); key.lan.port = th.th_dport; key.ext.port = th.th_sport; } STATE_LOOKUP(); if (direction == (*state)->direction) { src = &(*state)->dst; dst = &(*state)->src; } else { src = &(*state)->src; dst = &(*state)->dst; } if (src->wscale && dst->wscale) dws = dst->wscale & PF_WSCALE_MASK; else dws = 0; /* Demodulate sequence number */ seq = ntohl(th.th_seq) - src->seqdiff; if (src->seqdiff) { pf_change_a(&th.th_seq, icmpsum, htonl(seq), 0); copyback = 1; } if (!((*state)->state_flags & PFSTATE_SLOPPY) && (!SEQ_GEQ(src->seqhi, seq) || !SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)))) { if (pf_status.debug >= PF_DEBUG_MISC) { printf("pf: BAD ICMP %d:%d ", icmptype, pd->hdr.icmp->icmp_code); pf_print_host(pd->src, 0, pd->af); printf(" -> "); pf_print_host(pd->dst, 0, pd->af); printf(" state: "); pf_print_state(*state); printf(" seq=%u\n", seq); } REASON_SET(reason, PFRES_BADSTATE); return (PF_DROP); } if (STATE_TRANSLATE(*state)) { if (direction == PF_IN) { pf_change_icmp(pd2.src, &th.th_sport, daddr, &(*state)->lan.addr, (*state)->lan.port, NULL, pd2.ip_sum, icmpsum, pd->ip_sum, 0, pd2.af); } else { pf_change_icmp(pd2.dst, &th.th_dport, saddr, &(*state)->gwy.addr, (*state)->gwy.port, NULL, pd2.ip_sum, icmpsum, pd->ip_sum, 0, pd2.af); } copyback = 1; } if (copyback) { switch (pd2.af) { #ifdef INET case AF_INET: m_copyback(m, off, ICMP_MINLEN, (caddr_t)pd->hdr.icmp); m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2); break; #endif /* INET */ #ifdef INET6 case AF_INET6: m_copyback(m, off, sizeof(struct icmp6_hdr), (caddr_t)pd->hdr.icmp6); m_copyback(m, ipoff2, sizeof(h2_6), (caddr_t)&h2_6); break; #endif /* INET6 */ } m_copyback(m, off2, 8, (caddr_t)&th); } return (PF_PASS); break; } case IPPROTO_UDP: { struct udphdr uh; if (!pf_pull_hdr(m, off2, &uh, sizeof(uh), NULL, reason, pd2.af)) { DPFPRINTF(PF_DEBUG_MISC, ("pf: ICMP error message too short " "(udp)\n")); return (PF_DROP); } key.af = pd2.af; key.proto = IPPROTO_UDP; if (direction == PF_IN) { PF_ACPY(&key.ext.addr, pd2.dst, key.af); PF_ACPY(&key.gwy.addr, pd2.src, key.af); key.ext.port = uh.uh_dport; key.gwy.port = uh.uh_sport; } else { PF_ACPY(&key.lan.addr, pd2.dst, key.af); PF_ACPY(&key.ext.addr, pd2.src, key.af); key.lan.port = uh.uh_dport; key.ext.port = uh.uh_sport; } STATE_LOOKUP(); if (STATE_TRANSLATE(*state)) { if (direction == PF_IN) { pf_change_icmp(pd2.src, &uh.uh_sport, daddr, &(*state)->lan.addr, (*state)->lan.port, &uh.uh_sum, pd2.ip_sum, icmpsum, pd->ip_sum, 1, pd2.af); } else { pf_change_icmp(pd2.dst, &uh.uh_dport, saddr, &(*state)->gwy.addr, (*state)->gwy.port, &uh.uh_sum, pd2.ip_sum, icmpsum, pd->ip_sum, 1, pd2.af); } switch (pd2.af) { #ifdef INET case AF_INET: m_copyback(m, off, ICMP_MINLEN, (caddr_t)pd->hdr.icmp); m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2); break; #endif /* INET */ #ifdef INET6 case AF_INET6: m_copyback(m, off, sizeof(struct icmp6_hdr), (caddr_t)pd->hdr.icmp6); m_copyback(m, ipoff2, sizeof(h2_6), (caddr_t)&h2_6); break; #endif /* INET6 */ } m_copyback(m, off2, sizeof(uh), (caddr_t)&uh); } return (PF_PASS); break; } #ifdef INET case IPPROTO_ICMP: { struct icmp iih; if (!pf_pull_hdr(m, off2, &iih, ICMP_MINLEN, NULL, reason, pd2.af)) { DPFPRINTF(PF_DEBUG_MISC, ("pf: ICMP error message too short i" "(icmp)\n")); return (PF_DROP); } key.af = pd2.af; key.proto = IPPROTO_ICMP; if (direction == PF_IN) { PF_ACPY(&key.ext.addr, pd2.dst, key.af); PF_ACPY(&key.gwy.addr, pd2.src, key.af); key.ext.port = 0; key.gwy.port = iih.icmp_id; } else { PF_ACPY(&key.lan.addr, pd2.dst, key.af); PF_ACPY(&key.ext.addr, pd2.src, key.af); key.lan.port = iih.icmp_id; key.ext.port = 0; } STATE_LOOKUP(); if (STATE_TRANSLATE(*state)) { if (direction == PF_IN) { pf_change_icmp(pd2.src, &iih.icmp_id, daddr, &(*state)->lan.addr, (*state)->lan.port, NULL, pd2.ip_sum, icmpsum, pd->ip_sum, 0, AF_INET); } else { pf_change_icmp(pd2.dst, &iih.icmp_id, saddr, &(*state)->gwy.addr, (*state)->gwy.port, NULL, pd2.ip_sum, icmpsum, pd->ip_sum, 0, AF_INET); } m_copyback(m, off, ICMP_MINLEN, (caddr_t)pd->hdr.icmp); m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2); m_copyback(m, off2, ICMP_MINLEN, (caddr_t)&iih); } return (PF_PASS); break; } #endif /* INET */ #ifdef INET6 case IPPROTO_ICMPV6: { struct icmp6_hdr iih; if (!pf_pull_hdr(m, off2, &iih, sizeof(struct icmp6_hdr), NULL, reason, pd2.af)) { DPFPRINTF(PF_DEBUG_MISC, ("pf: ICMP error message too short " "(icmp6)\n")); return (PF_DROP); } key.af = pd2.af; key.proto = IPPROTO_ICMPV6; if (direction == PF_IN) { PF_ACPY(&key.ext.addr, pd2.dst, key.af); PF_ACPY(&key.gwy.addr, pd2.src, key.af); key.ext.port = 0; key.gwy.port = iih.icmp6_id; } else { PF_ACPY(&key.lan.addr, pd2.dst, key.af); PF_ACPY(&key.ext.addr, pd2.src, key.af); key.lan.port = iih.icmp6_id; key.ext.port = 0; } STATE_LOOKUP(); if (STATE_TRANSLATE(*state)) { if (direction == PF_IN) { pf_change_icmp(pd2.src, &iih.icmp6_id, daddr, &(*state)->lan.addr, (*state)->lan.port, NULL, pd2.ip_sum, icmpsum, pd->ip_sum, 0, AF_INET6); } else { pf_change_icmp(pd2.dst, &iih.icmp6_id, saddr, &(*state)->gwy.addr, (*state)->gwy.port, NULL, pd2.ip_sum, icmpsum, pd->ip_sum, 0, AF_INET6); } m_copyback(m, off, sizeof(struct icmp6_hdr), (caddr_t)pd->hdr.icmp6); m_copyback(m, ipoff2, sizeof(h2_6), (caddr_t)&h2_6); m_copyback(m, off2, sizeof(struct icmp6_hdr), (caddr_t)&iih); } return (PF_PASS); break; } #endif /* INET6 */ default: { key.af = pd2.af; key.proto = pd2.proto; if (direction == PF_IN) { PF_ACPY(&key.ext.addr, pd2.dst, key.af); PF_ACPY(&key.gwy.addr, pd2.src, key.af); key.ext.port = 0; key.gwy.port = 0; } else { PF_ACPY(&key.lan.addr, pd2.dst, key.af); PF_ACPY(&key.ext.addr, pd2.src, key.af); key.lan.port = 0; key.ext.port = 0; } STATE_LOOKUP(); if (STATE_TRANSLATE(*state)) { if (direction == PF_IN) { pf_change_icmp(pd2.src, NULL, daddr, &(*state)->lan.addr, 0, NULL, pd2.ip_sum, icmpsum, pd->ip_sum, 0, pd2.af); } else { pf_change_icmp(pd2.dst, NULL, saddr, &(*state)->gwy.addr, 0, NULL, pd2.ip_sum, icmpsum, pd->ip_sum, 0, pd2.af); } switch (pd2.af) { #ifdef INET case AF_INET: m_copyback(m, off, ICMP_MINLEN, (caddr_t)pd->hdr.icmp); m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2); break; #endif /* INET */ #ifdef INET6 case AF_INET6: m_copyback(m, off, sizeof(struct icmp6_hdr), (caddr_t)pd->hdr.icmp6); m_copyback(m, ipoff2, sizeof(h2_6), (caddr_t)&h2_6); break; #endif /* INET6 */ } } return (PF_PASS); break; } } } } int pf_test_state_other(struct pf_state **state, int direction, struct pfi_kif *kif, struct pf_pdesc *pd) { struct pf_state_peer *src, *dst; struct pf_state_cmp key; key.af = pd->af; key.proto = pd->proto; if (direction == PF_IN) { PF_ACPY(&key.ext.addr, pd->src, key.af); PF_ACPY(&key.gwy.addr, pd->dst, key.af); key.ext.port = 0; key.gwy.port = 0; } else { PF_ACPY(&key.lan.addr, pd->src, key.af); PF_ACPY(&key.ext.addr, pd->dst, key.af); key.lan.port = 0; key.ext.port = 0; } STATE_LOOKUP(); if (direction == (*state)->direction) { src = &(*state)->src; dst = &(*state)->dst; } else { src = &(*state)->dst; dst = &(*state)->src; } /* update states */ if (src->state < PFOTHERS_SINGLE) src->state = PFOTHERS_SINGLE; if (dst->state == PFOTHERS_SINGLE) dst->state = PFOTHERS_MULTIPLE; /* update expire time */ (*state)->expire = time_second; if (src->state == PFOTHERS_MULTIPLE && dst->state == PFOTHERS_MULTIPLE) (*state)->timeout = PFTM_OTHER_MULTIPLE; else (*state)->timeout = PFTM_OTHER_SINGLE; /* translate source/destination address, if necessary */ if (STATE_TRANSLATE(*state)) { if (direction == PF_OUT) switch (pd->af) { #ifdef INET case AF_INET: pf_change_a(&pd->src->v4.s_addr, pd->ip_sum, (*state)->gwy.addr.v4.s_addr, 0); break; #endif /* INET */ #ifdef INET6 case AF_INET6: PF_ACPY(pd->src, &(*state)->gwy.addr, pd->af); break; #endif /* INET6 */ } else switch (pd->af) { #ifdef INET case AF_INET: pf_change_a(&pd->dst->v4.s_addr, pd->ip_sum, (*state)->lan.addr.v4.s_addr, 0); break; #endif /* INET */ #ifdef INET6 case AF_INET6: PF_ACPY(pd->dst, &(*state)->lan.addr, pd->af); break; #endif /* INET6 */ } } return (PF_PASS); } /* * ipoff and off are measured from the start of the mbuf chain. * h must be at "ipoff" on the mbuf chain. */ void * pf_pull_hdr(struct mbuf *m, int off, void *p, int len, u_short *actionp, u_short *reasonp, sa_family_t af) { switch (af) { #ifdef INET case AF_INET: { struct ip *h = mtod(m, struct ip *); u_int16_t fragoff = (ntohs(h->ip_off) & IP_OFFMASK) << 3; if (fragoff) { if (fragoff >= len) ACTION_SET(actionp, PF_PASS); else { ACTION_SET(actionp, PF_DROP); REASON_SET(reasonp, PFRES_FRAG); } return (NULL); } if (m->m_pkthdr.len < off + len || ntohs(h->ip_len) < off + len) { ACTION_SET(actionp, PF_DROP); REASON_SET(reasonp, PFRES_SHORT); return (NULL); } break; } #endif /* INET */ #ifdef INET6 case AF_INET6: { struct ip6_hdr *h = mtod(m, struct ip6_hdr *); if (m->m_pkthdr.len < off + len || (ntohs(h->ip6_plen) + sizeof(struct ip6_hdr)) < (unsigned)(off + len)) { ACTION_SET(actionp, PF_DROP); REASON_SET(reasonp, PFRES_SHORT); return (NULL); } break; } #endif /* INET6 */ } m_copydata(m, off, len, p); return (p); } int pf_routable(struct pf_addr *addr, sa_family_t af, struct pfi_kif *kif) { struct sockaddr_in *dst; int ret = 1; int check_mpath; #ifndef __FreeBSD__ extern int ipmultipath; #endif #ifdef INET6 #ifndef __FreeBSD__ extern int ip6_multipath; #endif struct sockaddr_in6 *dst6; struct route_in6 ro; #else struct route ro; #endif struct radix_node *rn; struct rtentry *rt; struct ifnet *ifp; check_mpath = 0; bzero(&ro, sizeof(ro)); switch (af) { case AF_INET: dst = satosin(&ro.ro_dst); dst->sin_family = AF_INET; dst->sin_len = sizeof(*dst); dst->sin_addr = addr->v4; #ifndef __FreeBSD__ /* MULTIPATH_ROUTING */ if (ipmultipath) check_mpath = 1; #endif break; #ifdef INET6 case AF_INET6: dst6 = (struct sockaddr_in6 *)&ro.ro_dst; dst6->sin6_family = AF_INET6; dst6->sin6_len = sizeof(*dst6); dst6->sin6_addr = addr->v6; #ifndef __FreeBSD__ /* MULTIPATH_ROUTING */ if (ip6_multipath) check_mpath = 1; #endif break; #endif /* INET6 */ default: return (0); } /* Skip checks for ipsec interfaces */ if (kif != NULL && kif->pfik_ifp->if_type == IFT_ENC) goto out; #ifdef __FreeBSD__ /* XXX MRT not always INET */ /* stick with table 0 though */ if (af == AF_INET) in_rtalloc_ign((struct route *)&ro, 0, 0); else rtalloc_ign((struct route *)&ro, 0); #else /* ! __FreeBSD__ */ rtalloc_noclone((struct route *)&ro, NO_CLONING); #endif if (ro.ro_rt != NULL) { /* No interface given, this is a no-route check */ if (kif == NULL) goto out; if (kif->pfik_ifp == NULL) { ret = 0; goto out; } /* Perform uRPF check if passed input interface */ ret = 0; rn = (struct radix_node *)ro.ro_rt; do { rt = (struct rtentry *)rn; #ifndef __FreeBSD__ /* CARPDEV */ if (rt->rt_ifp->if_type == IFT_CARP) ifp = rt->rt_ifp->if_carpdev; else #endif ifp = rt->rt_ifp; if (kif->pfik_ifp == ifp) ret = 1; #ifdef __FreeBSD__ /* MULTIPATH_ROUTING */ rn = NULL; #else rn = rn_mpath_next(rn); #endif } while (check_mpath == 1 && rn != NULL && ret == 0); } else ret = 0; out: if (ro.ro_rt != NULL) RTFREE(ro.ro_rt); return (ret); } int pf_rtlabel_match(struct pf_addr *addr, sa_family_t af, struct pf_addr_wrap *aw) { struct sockaddr_in *dst; #ifdef INET6 struct sockaddr_in6 *dst6; struct route_in6 ro; #else struct route ro; #endif int ret = 0; bzero(&ro, sizeof(ro)); switch (af) { case AF_INET: dst = satosin(&ro.ro_dst); dst->sin_family = AF_INET; dst->sin_len = sizeof(*dst); dst->sin_addr = addr->v4; break; #ifdef INET6 case AF_INET6: dst6 = (struct sockaddr_in6 *)&ro.ro_dst; dst6->sin6_family = AF_INET6; dst6->sin6_len = sizeof(*dst6); dst6->sin6_addr = addr->v6; break; #endif /* INET6 */ default: return (0); } #ifdef __FreeBSD__ # ifdef RTF_PRCLONING rtalloc_ign((struct route *)&ro, (RTF_CLONING|RTF_PRCLONING)); # else /* !RTF_PRCLONING */ if (af == AF_INET) in_rtalloc_ign((struct route *)&ro, 0, 0); else rtalloc_ign((struct route *)&ro, 0); # endif #else /* ! __FreeBSD__ */ rtalloc_noclone((struct route *)&ro, NO_CLONING); #endif if (ro.ro_rt != NULL) { #ifdef __FreeBSD__ /* XXX_IMPORT: later */ #else if (ro.ro_rt->rt_labelid == aw->v.rtlabel) ret = 1; #endif RTFREE(ro.ro_rt); } return (ret); } #ifdef INET void pf_route(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp, struct pf_state *s, struct pf_pdesc *pd) { struct mbuf *m0, *m1; struct route iproute; struct route *ro = NULL; struct sockaddr_in *dst; struct ip *ip; struct ifnet *ifp = NULL; struct pf_addr naddr; struct pf_src_node *sn = NULL; int error = 0; #ifdef __FreeBSD__ int sw_csum; #endif #ifdef IPSEC struct m_tag *mtag; #endif /* IPSEC */ if (m == NULL || *m == NULL || r == NULL || (dir != PF_IN && dir != PF_OUT) || oifp == NULL) panic("pf_route: invalid parameters"); if (pd->pf_mtag->routed++ > 3) { m0 = *m; *m = NULL; goto bad; } if (r->rt == PF_DUPTO) { #ifdef __FreeBSD__ if ((m0 = m_dup(*m, M_DONTWAIT)) == NULL) #else if ((m0 = m_copym2(*m, 0, M_COPYALL, M_NOWAIT)) == NULL) #endif return; } else { if ((r->rt == PF_REPLYTO) == (r->direction == dir)) return; m0 = *m; } if (m0->m_len < sizeof(struct ip)) { DPFPRINTF(PF_DEBUG_URGENT, ("pf_route: m0->m_len < sizeof(struct ip)\n")); goto bad; } ip = mtod(m0, struct ip *); ro = &iproute; bzero((caddr_t)ro, sizeof(*ro)); dst = satosin(&ro->ro_dst); dst->sin_family = AF_INET; dst->sin_len = sizeof(*dst); dst->sin_addr = ip->ip_dst; if (r->rt == PF_FASTROUTE) { in_rtalloc(ro, 0); if (ro->ro_rt == 0) { KMOD_IPSTAT_INC(ips_noroute); goto bad; } ifp = ro->ro_rt->rt_ifp; ro->ro_rt->rt_use++; if (ro->ro_rt->rt_flags & RTF_GATEWAY) dst = satosin(ro->ro_rt->rt_gateway); } else { if (TAILQ_EMPTY(&r->rpool.list)) { DPFPRINTF(PF_DEBUG_URGENT, ("pf_route: TAILQ_EMPTY(&r->rpool.list)\n")); goto bad; } if (s == NULL) { pf_map_addr(AF_INET, r, (struct pf_addr *)&ip->ip_src, &naddr, NULL, &sn); if (!PF_AZERO(&naddr, AF_INET)) dst->sin_addr.s_addr = naddr.v4.s_addr; ifp = r->rpool.cur->kif ? r->rpool.cur->kif->pfik_ifp : NULL; } else { if (!PF_AZERO(&s->rt_addr, AF_INET)) dst->sin_addr.s_addr = s->rt_addr.v4.s_addr; ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL; } } if (ifp == NULL) goto bad; if (oifp != ifp) { #ifdef __FreeBSD__ PF_UNLOCK(); if (pf_test(PF_OUT, ifp, &m0, NULL, NULL) != PF_PASS) { PF_LOCK(); goto bad; } else if (m0 == NULL) { PF_LOCK(); goto done; } PF_LOCK(); #else if (pf_test(PF_OUT, ifp, &m0, NULL) != PF_PASS) goto bad; else if (m0 == NULL) goto done; #endif if (m0->m_len < sizeof(struct ip)) { DPFPRINTF(PF_DEBUG_URGENT, ("pf_route: m0->m_len < sizeof(struct ip)\n")); goto bad; } ip = mtod(m0, struct ip *); } #ifdef __FreeBSD__ /* Copied from FreeBSD 5.1-CURRENT ip_output. */ m0->m_pkthdr.csum_flags |= CSUM_IP; sw_csum = m0->m_pkthdr.csum_flags & ~ifp->if_hwassist; if (sw_csum & CSUM_DELAY_DATA) { /* * XXX: in_delayed_cksum assumes HBO for ip->ip_len (at least) */ NTOHS(ip->ip_len); NTOHS(ip->ip_off); /* XXX: needed? */ in_delayed_cksum(m0); HTONS(ip->ip_len); HTONS(ip->ip_off); sw_csum &= ~CSUM_DELAY_DATA; } m0->m_pkthdr.csum_flags &= ifp->if_hwassist; if (ntohs(ip->ip_len) <= ifp->if_mtu || (m0->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0 || (ifp->if_hwassist & CSUM_FRAGMENT && ((ip->ip_off & htons(IP_DF)) == 0))) { /* * ip->ip_len = htons(ip->ip_len); * ip->ip_off = htons(ip->ip_off); */ ip->ip_sum = 0; if (sw_csum & CSUM_DELAY_IP) { /* From KAME */ if (ip->ip_v == IPVERSION && (ip->ip_hl << 2) == sizeof(*ip)) { ip->ip_sum = in_cksum_hdr(ip); } else { ip->ip_sum = in_cksum(m0, ip->ip_hl << 2); } } PF_UNLOCK(); error = (*ifp->if_output)(ifp, m0, sintosa(dst), ro); PF_LOCK(); goto done; } #else /* Copied from ip_output. */ #ifdef IPSEC /* * If deferred crypto processing is needed, check that the * interface supports it. */ if ((mtag = m_tag_find(m0, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL)) != NULL && (ifp->if_capabilities & IFCAP_IPSEC) == 0) { /* Notify IPsec to do its own crypto. */ ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1)); goto bad; } #endif /* IPSEC */ /* Catch routing changes wrt. hardware checksumming for TCP or UDP. */ if (m0->m_pkthdr.csum_flags & M_TCPV4_CSUM_OUT) { if (!(ifp->if_capabilities & IFCAP_CSUM_TCPv4) || ifp->if_bridge != NULL) { in_delayed_cksum(m0); m0->m_pkthdr.csum_flags &= ~M_TCPV4_CSUM_OUT; /* Clear */ } } else if (m0->m_pkthdr.csum_flags & M_UDPV4_CSUM_OUT) { if (!(ifp->if_capabilities & IFCAP_CSUM_UDPv4) || ifp->if_bridge != NULL) { in_delayed_cksum(m0); m0->m_pkthdr.csum_flags &= ~M_UDPV4_CSUM_OUT; /* Clear */ } } if (ntohs(ip->ip_len) <= ifp->if_mtu) { if ((ifp->if_capabilities & IFCAP_CSUM_IPv4) && ifp->if_bridge == NULL) { m0->m_pkthdr.csum_flags |= M_IPV4_CSUM_OUT; KMOD_IPSTAT_INC(ips_outhwcsum); } else { ip->ip_sum = 0; ip->ip_sum = in_cksum(m0, ip->ip_hl << 2); } /* Update relevant hardware checksum stats for TCP/UDP */ if (m0->m_pkthdr.csum_flags & M_TCPV4_CSUM_OUT) KMOD_TCPSTAT_INC(tcps_outhwcsum); else if (m0->m_pkthdr.csum_flags & M_UDPV4_CSUM_OUT) KMOD_UDPSTAT_INC(udps_outhwcsum); error = (*ifp->if_output)(ifp, m0, sintosa(dst), NULL); goto done; } #endif /* * Too large for interface; fragment if possible. * Must be able to put at least 8 bytes per fragment. */ if (ip->ip_off & htons(IP_DF) || (m0->m_pkthdr.csum_flags & CSUM_TSO)) { KMOD_IPSTAT_INC(ips_cantfrag); if (r->rt != PF_DUPTO) { #ifdef __FreeBSD__ /* icmp_error() expects host byte ordering */ NTOHS(ip->ip_len); NTOHS(ip->ip_off); PF_UNLOCK(); icmp_error(m0, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG, 0, ifp->if_mtu); PF_LOCK(); #else icmp_error(m0, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG, 0, ifp->if_mtu); #endif goto done; } else goto bad; } m1 = m0; #ifdef __FreeBSD__ /* * XXX: is cheaper + less error prone than own function */ NTOHS(ip->ip_len); NTOHS(ip->ip_off); error = ip_fragment(ip, &m0, ifp->if_mtu, ifp->if_hwassist, sw_csum); #else error = ip_fragment(m0, ifp, ifp->if_mtu); #endif if (error) { #ifndef __FreeBSD__ /* ip_fragment does not do m_freem() on FreeBSD */ m0 = NULL; #endif goto bad; } for (m0 = m1; m0; m0 = m1) { m1 = m0->m_nextpkt; m0->m_nextpkt = 0; #ifdef __FreeBSD__ if (error == 0) { PF_UNLOCK(); error = (*ifp->if_output)(ifp, m0, sintosa(dst), NULL); PF_LOCK(); } else #else if (error == 0) error = (*ifp->if_output)(ifp, m0, sintosa(dst), NULL); else #endif m_freem(m0); } if (error == 0) KMOD_IPSTAT_INC(ips_fragmented); done: if (r->rt != PF_DUPTO) *m = NULL; if (ro == &iproute && ro->ro_rt) RTFREE(ro->ro_rt); return; bad: m_freem(m0); goto done; } #endif /* INET */ #ifdef INET6 void pf_route6(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp, struct pf_state *s, struct pf_pdesc *pd) { struct mbuf *m0; struct route_in6 ip6route; struct route_in6 *ro; struct sockaddr_in6 *dst; struct ip6_hdr *ip6; struct ifnet *ifp = NULL; struct pf_addr naddr; struct pf_src_node *sn = NULL; int error = 0; if (m == NULL || *m == NULL || r == NULL || (dir != PF_IN && dir != PF_OUT) || oifp == NULL) panic("pf_route6: invalid parameters"); if (pd->pf_mtag->routed++ > 3) { m0 = *m; *m = NULL; goto bad; } if (r->rt == PF_DUPTO) { #ifdef __FreeBSD__ if ((m0 = m_dup(*m, M_DONTWAIT)) == NULL) #else if ((m0 = m_copym2(*m, 0, M_COPYALL, M_NOWAIT)) == NULL) #endif return; } else { if ((r->rt == PF_REPLYTO) == (r->direction == dir)) return; m0 = *m; } if (m0->m_len < sizeof(struct ip6_hdr)) { DPFPRINTF(PF_DEBUG_URGENT, ("pf_route6: m0->m_len < sizeof(struct ip6_hdr)\n")); goto bad; } ip6 = mtod(m0, struct ip6_hdr *); ro = &ip6route; bzero((caddr_t)ro, sizeof(*ro)); dst = (struct sockaddr_in6 *)&ro->ro_dst; dst->sin6_family = AF_INET6; dst->sin6_len = sizeof(*dst); dst->sin6_addr = ip6->ip6_dst; /* Cheat. XXX why only in the v6 case??? */ if (r->rt == PF_FASTROUTE) { #ifdef __FreeBSD__ m0->m_flags |= M_SKIP_FIREWALL; PF_UNLOCK(); ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL); PF_LOCK(); #else mtag = m_tag_get(PACKET_TAG_PF_GENERATED, 0, M_NOWAIT); if (mtag == NULL) goto bad; m_tag_prepend(m0, mtag); pd->pf_mtag->flags |= PF_TAG_GENERATED; ip6_output(m0, NULL, NULL, 0, NULL, NULL); #endif return; } if (TAILQ_EMPTY(&r->rpool.list)) { DPFPRINTF(PF_DEBUG_URGENT, ("pf_route6: TAILQ_EMPTY(&r->rpool.list)\n")); goto bad; } if (s == NULL) { pf_map_addr(AF_INET6, r, (struct pf_addr *)&ip6->ip6_src, &naddr, NULL, &sn); if (!PF_AZERO(&naddr, AF_INET6)) PF_ACPY((struct pf_addr *)&dst->sin6_addr, &naddr, AF_INET6); ifp = r->rpool.cur->kif ? r->rpool.cur->kif->pfik_ifp : NULL; } else { if (!PF_AZERO(&s->rt_addr, AF_INET6)) PF_ACPY((struct pf_addr *)&dst->sin6_addr, &s->rt_addr, AF_INET6); ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL; } if (ifp == NULL) goto bad; if (oifp != ifp) { #ifdef __FreeBSD__ PF_UNLOCK(); if (pf_test6(PF_OUT, ifp, &m0, NULL, NULL) != PF_PASS) { PF_LOCK(); goto bad; } else if (m0 == NULL) { PF_LOCK(); goto done; } PF_LOCK(); #else if (pf_test6(PF_OUT, ifp, &m0, NULL) != PF_PASS) goto bad; else if (m0 == NULL) goto done; #endif if (m0->m_len < sizeof(struct ip6_hdr)) { DPFPRINTF(PF_DEBUG_URGENT, ("pf_route6: m0->m_len < sizeof(struct ip6_hdr)\n")); goto bad; } ip6 = mtod(m0, struct ip6_hdr *); } /* * If the packet is too large for the outgoing interface, * send back an icmp6 error. */ if (IN6_IS_SCOPE_EMBED(&dst->sin6_addr)) dst->sin6_addr.s6_addr16[1] = htons(ifp->if_index); if ((u_long)m0->m_pkthdr.len <= ifp->if_mtu) { #ifdef __FreeBSD__ PF_UNLOCK(); #endif error = nd6_output(ifp, ifp, m0, dst, NULL); #ifdef __FreeBSD__ PF_LOCK(); #endif } else { in6_ifstat_inc(ifp, ifs6_in_toobig); #ifdef __FreeBSD__ if (r->rt != PF_DUPTO) { PF_UNLOCK(); icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu); PF_LOCK(); } else #else if (r->rt != PF_DUPTO) icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu); else #endif goto bad; } done: if (r->rt != PF_DUPTO) *m = NULL; return; bad: m_freem(m0); goto done; } #endif /* INET6 */ #ifdef __FreeBSD__ /* * FreeBSD supports cksum offloads for the following drivers. * em(4), fxp(4), ixgb(4), lge(4), ndis(4), nge(4), re(4), * ti(4), txp(4), xl(4) * * CSUM_DATA_VALID | CSUM_PSEUDO_HDR : * network driver performed cksum including pseudo header, need to verify * csum_data * CSUM_DATA_VALID : * network driver performed cksum, needs to additional pseudo header * cksum computation with partial csum_data(i.e. lack of H/W support for * pseudo header, for instance hme(4), sk(4) and possibly gem(4)) * * After validating the cksum of packet, set both flag CSUM_DATA_VALID and * CSUM_PSEUDO_HDR in order to avoid recomputation of the cksum in upper * TCP/UDP layer. * Also, set csum_data to 0xffff to force cksum validation. */ int pf_check_proto_cksum(struct mbuf *m, int off, int len, u_int8_t p, sa_family_t af) { u_int16_t sum = 0; int hw_assist = 0; struct ip *ip; if (off < sizeof(struct ip) || len < sizeof(struct udphdr)) return (1); if (m->m_pkthdr.len < off + len) return (1); switch (p) { case IPPROTO_TCP: if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) { if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) { sum = m->m_pkthdr.csum_data; } else { ip = mtod(m, struct ip *); sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr, htonl((u_short)len + m->m_pkthdr.csum_data + IPPROTO_TCP)); } sum ^= 0xffff; ++hw_assist; } break; case IPPROTO_UDP: if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) { if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) { sum = m->m_pkthdr.csum_data; } else { ip = mtod(m, struct ip *); sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr, htonl((u_short)len + m->m_pkthdr.csum_data + IPPROTO_UDP)); } sum ^= 0xffff; ++hw_assist; } break; case IPPROTO_ICMP: #ifdef INET6 case IPPROTO_ICMPV6: #endif /* INET6 */ break; default: return (1); } if (!hw_assist) { switch (af) { case AF_INET: if (p == IPPROTO_ICMP) { if (m->m_len < off) return (1); m->m_data += off; m->m_len -= off; sum = in_cksum(m, len); m->m_data -= off; m->m_len += off; } else { if (m->m_len < sizeof(struct ip)) return (1); sum = in4_cksum(m, p, off, len); } break; #ifdef INET6 case AF_INET6: if (m->m_len < sizeof(struct ip6_hdr)) return (1); sum = in6_cksum(m, p, off, len); break; #endif /* INET6 */ default: return (1); } } if (sum) { switch (p) { case IPPROTO_TCP: { KMOD_TCPSTAT_INC(tcps_rcvbadsum); break; } case IPPROTO_UDP: { KMOD_UDPSTAT_INC(udps_badsum); break; } case IPPROTO_ICMP: { KMOD_ICMPSTAT_INC(icps_checksum); break; } #ifdef INET6 case IPPROTO_ICMPV6: { KMOD_ICMP6STAT_INC(icp6s_checksum); break; } #endif /* INET6 */ } return (1); } else { if (p == IPPROTO_TCP || p == IPPROTO_UDP) { m->m_pkthdr.csum_flags |= (CSUM_DATA_VALID | CSUM_PSEUDO_HDR); m->m_pkthdr.csum_data = 0xffff; } } return (0); } #else /* !__FreeBSD__ */ /* * check protocol (tcp/udp/icmp/icmp6) checksum and set mbuf flag * off is the offset where the protocol header starts * len is the total length of protocol header plus payload * returns 0 when the checksum is valid, otherwise returns 1. */ int pf_check_proto_cksum(struct mbuf *m, int off, int len, u_int8_t p, sa_family_t af) { u_int16_t flag_ok, flag_bad; u_int16_t sum; switch (p) { case IPPROTO_TCP: flag_ok = M_TCP_CSUM_IN_OK; flag_bad = M_TCP_CSUM_IN_BAD; break; case IPPROTO_UDP: flag_ok = M_UDP_CSUM_IN_OK; flag_bad = M_UDP_CSUM_IN_BAD; break; case IPPROTO_ICMP: #ifdef INET6 case IPPROTO_ICMPV6: #endif /* INET6 */ flag_ok = flag_bad = 0; break; default: return (1); } if (m->m_pkthdr.csum_flags & flag_ok) return (0); if (m->m_pkthdr.csum_flags & flag_bad) return (1); if (off < sizeof(struct ip) || len < sizeof(struct udphdr)) return (1); if (m->m_pkthdr.len < off + len) return (1); switch (af) { #ifdef INET case AF_INET: if (p == IPPROTO_ICMP) { if (m->m_len < off) return (1); m->m_data += off; m->m_len -= off; sum = in_cksum(m, len); m->m_data -= off; m->m_len += off; } else { if (m->m_len < sizeof(struct ip)) return (1); sum = in4_cksum(m, p, off, len); } break; #endif /* INET */ #ifdef INET6 case AF_INET6: if (m->m_len < sizeof(struct ip6_hdr)) return (1); sum = in6_cksum(m, p, off, len); break; #endif /* INET6 */ default: return (1); } if (sum) { m->m_pkthdr.csum_flags |= flag_bad; switch (p) { case IPPROTO_TCP: KMOD_TCPSTAT_INC(tcps_rcvbadsum); break; case IPPROTO_UDP: KMOD_UDPSTAT_INC(udps_badsum); break; case IPPROTO_ICMP: KMOD_ICMPSTAT_INC(icps_checksum); break; #ifdef INET6 case IPPROTO_ICMPV6: KMOD_ICMP6STAT_INC(icp6s_checksum); break; #endif /* INET6 */ } return (1); } m->m_pkthdr.csum_flags |= flag_ok; return (0); } #endif /* __FreeBSD__ */ #ifdef INET int #ifdef __FreeBSD__ pf_test(int dir, struct ifnet *ifp, struct mbuf **m0, struct ether_header *eh, struct inpcb *inp) #else pf_test(int dir, struct ifnet *ifp, struct mbuf **m0, struct ether_header *eh) #endif { struct pfi_kif *kif; u_short action, reason = 0, log = 0; struct mbuf *m = *m0; struct ip *h = NULL; /* make the compiler happy */ struct pf_rule *a = NULL, *r = &pf_default_rule, *tr, *nr; struct pf_state *s = NULL; struct pf_ruleset *ruleset = NULL; struct pf_pdesc pd; int off, dirndx, pqid = 0; #ifdef __FreeBSD__ PF_LOCK(); #endif if (!pf_status.running) #ifdef __FreeBSD__ { PF_UNLOCK(); #endif return (PF_PASS); #ifdef __FreeBSD__ } #endif memset(&pd, 0, sizeof(pd)); if ((pd.pf_mtag = pf_get_mtag(m)) == NULL) { #ifdef __FreeBSD__ PF_UNLOCK(); #endif DPFPRINTF(PF_DEBUG_URGENT, ("pf_test: pf_get_mtag returned NULL\n")); return (PF_DROP); } #ifdef __FreeBSD__ if (m->m_flags & M_SKIP_FIREWALL) { PF_UNLOCK(); return (PF_PASS); } #else if (pd.pf_mtag->flags & PF_TAG_GENERATED) return (PF_PASS); #endif #ifdef __FreeBSD__ /* XXX_IMPORT: later */ #else if (ifp->if_type == IFT_CARP && ifp->if_carpdev) ifp = ifp->if_carpdev; #endif kif = (struct pfi_kif *)ifp->if_pf_kif; if (kif == NULL) { #ifdef __FreeBSD__ PF_UNLOCK(); #endif DPFPRINTF(PF_DEBUG_URGENT, ("pf_test: kif == NULL, if_xname %s\n", ifp->if_xname)); return (PF_DROP); } if (kif->pfik_flags & PFI_IFLAG_SKIP) { #ifdef __FreeBSD__ PF_UNLOCK(); #endif return (PF_PASS); } #ifdef __FreeBSD__ M_ASSERTPKTHDR(m); #else #ifdef DIAGNOSTIC if ((m->m_flags & M_PKTHDR) == 0) panic("non-M_PKTHDR is passed to pf_test"); #endif /* DIAGNOSTIC */ #endif /* __FreeBSD__ */ if (m->m_pkthdr.len < (int)sizeof(*h)) { action = PF_DROP; REASON_SET(&reason, PFRES_SHORT); log = 1; goto done; } /* We do IP header normalization and packet reassembly here */ if (pf_normalize_ip(m0, dir, kif, &reason, &pd) != PF_PASS) { action = PF_DROP; goto done; } m = *m0; h = mtod(m, struct ip *); off = h->ip_hl << 2; if (off < (int)sizeof(*h)) { action = PF_DROP; REASON_SET(&reason, PFRES_SHORT); log = 1; goto done; } pd.src = (struct pf_addr *)&h->ip_src; pd.dst = (struct pf_addr *)&h->ip_dst; PF_ACPY(&pd.baddr, dir == PF_OUT ? pd.src : pd.dst, AF_INET); pd.ip_sum = &h->ip_sum; pd.proto = h->ip_p; pd.af = AF_INET; pd.tos = h->ip_tos; pd.tot_len = ntohs(h->ip_len); pd.eh = eh; /* handle fragments that didn't get reassembled by normalization */ if (h->ip_off & htons(IP_MF | IP_OFFMASK)) { action = pf_test_fragment(&r, dir, kif, m, h, &pd, &a, &ruleset); goto done; } switch (h->ip_p) { case IPPROTO_TCP: { struct tcphdr th; pd.hdr.tcp = &th; if (!pf_pull_hdr(m, off, &th, sizeof(th), &action, &reason, AF_INET)) { log = action != PF_PASS; goto done; } if (dir == PF_IN && pf_check_proto_cksum(m, off, ntohs(h->ip_len) - off, IPPROTO_TCP, AF_INET)) { REASON_SET(&reason, PFRES_PROTCKSUM); action = PF_DROP; goto done; } pd.p_len = pd.tot_len - off - (th.th_off << 2); if ((th.th_flags & TH_ACK) && pd.p_len == 0) pqid = 1; action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd); if (action == PF_DROP) goto done; action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd, &reason); if (action == PF_PASS) { #if NPFSYNC pfsync_update_state(s); #endif /* NPFSYNC */ r = s->rule.ptr; a = s->anchor.ptr; log = s->log; } else if (s == NULL) #ifdef __FreeBSD__ action = pf_test_tcp(&r, &s, dir, kif, m, off, h, &pd, &a, &ruleset, NULL, inp); #else action = pf_test_tcp(&r, &s, dir, kif, m, off, h, &pd, &a, &ruleset, &ipintrq); #endif break; } case IPPROTO_UDP: { struct udphdr uh; pd.hdr.udp = &uh; if (!pf_pull_hdr(m, off, &uh, sizeof(uh), &action, &reason, AF_INET)) { log = action != PF_PASS; goto done; } if (dir == PF_IN && uh.uh_sum && pf_check_proto_cksum(m, off, ntohs(h->ip_len) - off, IPPROTO_UDP, AF_INET)) { action = PF_DROP; REASON_SET(&reason, PFRES_PROTCKSUM); goto done; } if (uh.uh_dport == 0 || ntohs(uh.uh_ulen) > m->m_pkthdr.len - off || ntohs(uh.uh_ulen) < sizeof(struct udphdr)) { action = PF_DROP; REASON_SET(&reason, PFRES_SHORT); goto done; } action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd); if (action == PF_PASS) { #if NPFSYNC pfsync_update_state(s); #endif /* NPFSYNC */ r = s->rule.ptr; a = s->anchor.ptr; log = s->log; } else if (s == NULL) #ifdef __FreeBSD__ action = pf_test_udp(&r, &s, dir, kif, m, off, h, &pd, &a, &ruleset, NULL, inp); #else action = pf_test_udp(&r, &s, dir, kif, m, off, h, &pd, &a, &ruleset, &ipintrq); #endif break; } case IPPROTO_ICMP: { struct icmp ih; pd.hdr.icmp = &ih; if (!pf_pull_hdr(m, off, &ih, ICMP_MINLEN, &action, &reason, AF_INET)) { log = action != PF_PASS; goto done; } if (dir == PF_IN && pf_check_proto_cksum(m, off, ntohs(h->ip_len) - off, IPPROTO_ICMP, AF_INET)) { action = PF_DROP; REASON_SET(&reason, PFRES_PROTCKSUM); goto done; } action = pf_test_state_icmp(&s, dir, kif, m, off, h, &pd, &reason); if (action == PF_PASS) { #if NPFSYNC pfsync_update_state(s); #endif /* NPFSYNC */ r = s->rule.ptr; a = s->anchor.ptr; log = s->log; } else if (s == NULL) #ifdef __FreeBSD__ action = pf_test_icmp(&r, &s, dir, kif, m, off, h, &pd, &a, &ruleset, NULL); #else action = pf_test_icmp(&r, &s, dir, kif, m, off, h, &pd, &a, &ruleset, &ipintrq); #endif break; } default: action = pf_test_state_other(&s, dir, kif, &pd); if (action == PF_PASS) { #if NPFSYNC pfsync_update_state(s); #endif /* NPFSYNC */ r = s->rule.ptr; a = s->anchor.ptr; log = s->log; } else if (s == NULL) #ifdef __FreeBSD__ action = pf_test_other(&r, &s, dir, kif, m, off, h, &pd, &a, &ruleset, NULL); #else action = pf_test_other(&r, &s, dir, kif, m, off, h, &pd, &a, &ruleset, &ipintrq); #endif break; } done: if (action == PF_PASS && h->ip_hl > 5 && !((s && s->state_flags & PFSTATE_ALLOWOPTS) || r->allow_opts)) { action = PF_DROP; REASON_SET(&reason, PFRES_IPOPTIONS); log = 1; DPFPRINTF(PF_DEBUG_MISC, ("pf: dropping packet with ip options\n")); } if ((s && s->tag) || r->rtableid) pf_tag_packet(m, pd.pf_mtag, s ? s->tag : 0, r->rtableid); #ifdef ALTQ if (action == PF_PASS && r->qid) { if (pqid || (pd.tos & IPTOS_LOWDELAY)) pd.pf_mtag->qid = r->pqid; else pd.pf_mtag->qid = r->qid; /* add hints for ecn */ pd.pf_mtag->af = AF_INET; pd.pf_mtag->hdr = h; } #endif /* ALTQ */ /* * connections redirected to loopback should not match sockets * bound specifically to loopback due to security implications, * see tcp_input() and in_pcblookup_listen(). */ if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP || pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL && (s->nat_rule.ptr->action == PF_RDR || s->nat_rule.ptr->action == PF_BINAT) && (ntohl(pd.dst->v4.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) pd.pf_mtag->flags |= PF_TAG_TRANSLATE_LOCALHOST; if (log) { struct pf_rule *lr; if (s != NULL && s->nat_rule.ptr != NULL && s->nat_rule.ptr->log & PF_LOG_ALL) lr = s->nat_rule.ptr; else lr = r; PFLOG_PACKET(kif, h, m, AF_INET, dir, reason, lr, a, ruleset, &pd); } kif->pfik_bytes[0][dir == PF_OUT][action != PF_PASS] += pd.tot_len; kif->pfik_packets[0][dir == PF_OUT][action != PF_PASS]++; if (action == PF_PASS || r->action == PF_DROP) { dirndx = (dir == PF_OUT); r->packets[dirndx]++; r->bytes[dirndx] += pd.tot_len; if (a != NULL) { a->packets[dirndx]++; a->bytes[dirndx] += pd.tot_len; } if (s != NULL) { if (s->nat_rule.ptr != NULL) { s->nat_rule.ptr->packets[dirndx]++; s->nat_rule.ptr->bytes[dirndx] += pd.tot_len; } if (s->src_node != NULL) { s->src_node->packets[dirndx]++; s->src_node->bytes[dirndx] += pd.tot_len; } if (s->nat_src_node != NULL) { s->nat_src_node->packets[dirndx]++; s->nat_src_node->bytes[dirndx] += pd.tot_len; } dirndx = (dir == s->direction) ? 0 : 1; s->packets[dirndx]++; s->bytes[dirndx] += pd.tot_len; } tr = r; nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule; if (nr != NULL) { struct pf_addr *x; /* * XXX: we need to make sure that the addresses * passed to pfr_update_stats() are the same than * the addresses used during matching (pfr_match) */ if (r == &pf_default_rule) { tr = nr; x = (s == NULL || s->direction == dir) ? &pd.baddr : &pd.naddr; } else x = (s == NULL || s->direction == dir) ? &pd.naddr : &pd.baddr; if (x == &pd.baddr || s == NULL) { /* we need to change the address */ if (dir == PF_OUT) pd.src = x; else pd.dst = x; } } if (tr->src.addr.type == PF_ADDR_TABLE) pfr_update_stats(tr->src.addr.p.tbl, (s == NULL || s->direction == dir) ? pd.src : pd.dst, pd.af, pd.tot_len, dir == PF_OUT, r->action == PF_PASS, tr->src.neg); if (tr->dst.addr.type == PF_ADDR_TABLE) pfr_update_stats(tr->dst.addr.p.tbl, (s == NULL || s->direction == dir) ? pd.dst : pd.src, pd.af, pd.tot_len, dir == PF_OUT, r->action == PF_PASS, tr->dst.neg); } if (action == PF_SYNPROXY_DROP) { m_freem(*m0); *m0 = NULL; action = PF_PASS; } else if (r->rt) /* pf_route can free the mbuf causing *m0 to become NULL */ pf_route(m0, r, dir, ifp, s, &pd); #ifdef __FreeBSD__ PF_UNLOCK(); #endif return (action); } #endif /* INET */ #ifdef INET6 int #ifdef __FreeBSD__ pf_test6(int dir, struct ifnet *ifp, struct mbuf **m0, struct ether_header *eh, struct inpcb *inp) #else pf_test6(int dir, struct ifnet *ifp, struct mbuf **m0, struct ether_header *eh) #endif { struct pfi_kif *kif; u_short action, reason = 0, log = 0; struct mbuf *m = *m0, *n = NULL; struct ip6_hdr *h; struct pf_rule *a = NULL, *r = &pf_default_rule, *tr, *nr; struct pf_state *s = NULL; struct pf_ruleset *ruleset = NULL; struct pf_pdesc pd; int off, terminal = 0, dirndx, rh_cnt = 0; #ifdef __FreeBSD__ PF_LOCK(); #endif if (!pf_status.running) #ifdef __FreeBSD__ { PF_UNLOCK(); #endif return (PF_PASS); #ifdef __FreeBSD__ } #endif memset(&pd, 0, sizeof(pd)); if ((pd.pf_mtag = pf_get_mtag(m)) == NULL) { #ifdef __FreeBSD__ PF_UNLOCK(); #endif DPFPRINTF(PF_DEBUG_URGENT, ("pf_test6: pf_get_mtag returned NULL\n")); return (PF_DROP); } if (pd.pf_mtag->flags & PF_TAG_GENERATED) return (PF_PASS); #ifdef __FreeBSD__ /* XXX_IMPORT: later */ #else if (ifp->if_type == IFT_CARP && ifp->if_carpdev) ifp = ifp->if_carpdev; #endif kif = (struct pfi_kif *)ifp->if_pf_kif; if (kif == NULL) { #ifdef __FreeBSD__ PF_UNLOCK(); #endif DPFPRINTF(PF_DEBUG_URGENT, ("pf_test6: kif == NULL, if_xname %s\n", ifp->if_xname)); return (PF_DROP); } if (kif->pfik_flags & PFI_IFLAG_SKIP) { #ifdef __FreeBSD__ PF_UNLOCK(); #endif return (PF_PASS); } #ifdef __FreeBSD__ M_ASSERTPKTHDR(m); #else #ifdef DIAGNOSTIC if ((m->m_flags & M_PKTHDR) == 0) panic("non-M_PKTHDR is passed to pf_test6"); #endif /* DIAGNOSTIC */ #endif #ifdef __FreeBSD__ h = NULL; /* make the compiler happy */ #endif if (m->m_pkthdr.len < (int)sizeof(*h)) { action = PF_DROP; REASON_SET(&reason, PFRES_SHORT); log = 1; goto done; } /* We do IP header normalization and packet reassembly here */ if (pf_normalize_ip6(m0, dir, kif, &reason, &pd) != PF_PASS) { action = PF_DROP; goto done; } m = *m0; h = mtod(m, struct ip6_hdr *); #if 1 /* * we do not support jumbogram yet. if we keep going, zero ip6_plen * will do something bad, so drop the packet for now. */ if (htons(h->ip6_plen) == 0) { action = PF_DROP; REASON_SET(&reason, PFRES_NORM); /*XXX*/ goto done; } #endif pd.src = (struct pf_addr *)&h->ip6_src; pd.dst = (struct pf_addr *)&h->ip6_dst; PF_ACPY(&pd.baddr, dir == PF_OUT ? pd.src : pd.dst, AF_INET6); pd.ip_sum = NULL; pd.af = AF_INET6; pd.tos = 0; pd.tot_len = ntohs(h->ip6_plen) + sizeof(struct ip6_hdr); pd.eh = eh; off = ((caddr_t)h - m->m_data) + sizeof(struct ip6_hdr); pd.proto = h->ip6_nxt; do { switch (pd.proto) { case IPPROTO_FRAGMENT: action = pf_test_fragment(&r, dir, kif, m, h, &pd, &a, &ruleset); if (action == PF_DROP) REASON_SET(&reason, PFRES_FRAG); goto done; case IPPROTO_ROUTING: { struct ip6_rthdr rthdr; if (rh_cnt++) { DPFPRINTF(PF_DEBUG_MISC, ("pf: IPv6 more than one rthdr\n")); action = PF_DROP; REASON_SET(&reason, PFRES_IPOPTIONS); log = 1; goto done; } if (!pf_pull_hdr(m, off, &rthdr, sizeof(rthdr), NULL, &reason, pd.af)) { DPFPRINTF(PF_DEBUG_MISC, ("pf: IPv6 short rthdr\n")); action = PF_DROP; REASON_SET(&reason, PFRES_SHORT); log = 1; goto done; } if (rthdr.ip6r_type == IPV6_RTHDR_TYPE_0) { DPFPRINTF(PF_DEBUG_MISC, ("pf: IPv6 rthdr0\n")); action = PF_DROP; REASON_SET(&reason, PFRES_IPOPTIONS); log = 1; goto done; } /* fallthrough */ } case IPPROTO_AH: case IPPROTO_HOPOPTS: case IPPROTO_DSTOPTS: { /* get next header and header length */ struct ip6_ext opt6; if (!pf_pull_hdr(m, off, &opt6, sizeof(opt6), NULL, &reason, pd.af)) { DPFPRINTF(PF_DEBUG_MISC, ("pf: IPv6 short opt\n")); action = PF_DROP; log = 1; goto done; } if (pd.proto == IPPROTO_AH) off += (opt6.ip6e_len + 2) * 4; else off += (opt6.ip6e_len + 1) * 8; pd.proto = opt6.ip6e_nxt; /* goto the next header */ break; } default: terminal++; break; } } while (!terminal); /* if there's no routing header, use unmodified mbuf for checksumming */ if (!n) n = m; switch (pd.proto) { case IPPROTO_TCP: { struct tcphdr th; pd.hdr.tcp = &th; if (!pf_pull_hdr(m, off, &th, sizeof(th), &action, &reason, AF_INET6)) { log = action != PF_PASS; goto done; } if (dir == PF_IN && pf_check_proto_cksum(n, off, ntohs(h->ip6_plen) - (off - sizeof(struct ip6_hdr)), IPPROTO_TCP, AF_INET6)) { action = PF_DROP; REASON_SET(&reason, PFRES_PROTCKSUM); goto done; } pd.p_len = pd.tot_len - off - (th.th_off << 2); action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd); if (action == PF_DROP) goto done; action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd, &reason); if (action == PF_PASS) { #if NPFSYNC pfsync_update_state(s); #endif /* NPFSYNC */ r = s->rule.ptr; a = s->anchor.ptr; log = s->log; } else if (s == NULL) #ifdef __FreeBSD__ action = pf_test_tcp(&r, &s, dir, kif, m, off, h, &pd, &a, &ruleset, NULL, inp); #else action = pf_test_tcp(&r, &s, dir, kif, m, off, h, &pd, &a, &ruleset, &ip6intrq); #endif break; } case IPPROTO_UDP: { struct udphdr uh; pd.hdr.udp = &uh; if (!pf_pull_hdr(m, off, &uh, sizeof(uh), &action, &reason, AF_INET6)) { log = action != PF_PASS; goto done; } if (dir == PF_IN && uh.uh_sum && pf_check_proto_cksum(n, off, ntohs(h->ip6_plen) - (off - sizeof(struct ip6_hdr)), IPPROTO_UDP, AF_INET6)) { action = PF_DROP; REASON_SET(&reason, PFRES_PROTCKSUM); goto done; } if (uh.uh_dport == 0 || ntohs(uh.uh_ulen) > m->m_pkthdr.len - off || ntohs(uh.uh_ulen) < sizeof(struct udphdr)) { action = PF_DROP; REASON_SET(&reason, PFRES_SHORT); goto done; } action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd); if (action == PF_PASS) { #if NPFSYNC pfsync_update_state(s); #endif /* NPFSYNC */ r = s->rule.ptr; a = s->anchor.ptr; log = s->log; } else if (s == NULL) #ifdef __FreeBSD__ action = pf_test_udp(&r, &s, dir, kif, m, off, h, &pd, &a, &ruleset, NULL, inp); #else action = pf_test_udp(&r, &s, dir, kif, m, off, h, &pd, &a, &ruleset, &ip6intrq); #endif break; } case IPPROTO_ICMPV6: { struct icmp6_hdr ih; pd.hdr.icmp6 = &ih; if (!pf_pull_hdr(m, off, &ih, sizeof(ih), &action, &reason, AF_INET6)) { log = action != PF_PASS; goto done; } if (dir == PF_IN && pf_check_proto_cksum(n, off, ntohs(h->ip6_plen) - (off - sizeof(struct ip6_hdr)), IPPROTO_ICMPV6, AF_INET6)) { action = PF_DROP; REASON_SET(&reason, PFRES_PROTCKSUM); goto done; } action = pf_test_state_icmp(&s, dir, kif, m, off, h, &pd, &reason); if (action == PF_PASS) { #if NPFSYNC pfsync_update_state(s); #endif /* NPFSYNC */ r = s->rule.ptr; a = s->anchor.ptr; log = s->log; } else if (s == NULL) #ifdef __FreeBSD__ action = pf_test_icmp(&r, &s, dir, kif, m, off, h, &pd, &a, &ruleset, NULL); #else action = pf_test_icmp(&r, &s, dir, kif, m, off, h, &pd, &a, &ruleset, &ip6intrq); #endif break; } default: action = pf_test_state_other(&s, dir, kif, &pd); if (action == PF_PASS) { #if NPFSYNC pfsync_update_state(s); #endif /* NPFSYNC */ r = s->rule.ptr; a = s->anchor.ptr; log = s->log; } else if (s == NULL) #ifdef __FreeBSD__ action = pf_test_other(&r, &s, dir, kif, m, off, h, &pd, &a, &ruleset, NULL); #else action = pf_test_other(&r, &s, dir, kif, m, off, h, &pd, &a, &ruleset, &ip6intrq); #endif break; } done: /* handle dangerous IPv6 extension headers. */ if (action == PF_PASS && rh_cnt && !((s && s->state_flags & PFSTATE_ALLOWOPTS) || r->allow_opts)) { action = PF_DROP; REASON_SET(&reason, PFRES_IPOPTIONS); log = 1; DPFPRINTF(PF_DEBUG_MISC, ("pf: dropping packet with dangerous v6 headers\n")); } if ((s && s->tag) || r->rtableid) pf_tag_packet(m, pd.pf_mtag, s ? s->tag : 0, r->rtableid); #ifdef ALTQ if (action == PF_PASS && r->qid) { if (pd.tos & IPTOS_LOWDELAY) pd.pf_mtag->qid = r->pqid; else pd.pf_mtag->qid = r->qid; /* add hints for ecn */ pd.pf_mtag->af = AF_INET6; pd.pf_mtag->hdr = h; } #endif /* ALTQ */ if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP || pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL && (s->nat_rule.ptr->action == PF_RDR || s->nat_rule.ptr->action == PF_BINAT) && IN6_IS_ADDR_LOOPBACK(&pd.dst->v6)) pd.pf_mtag->flags |= PF_TAG_TRANSLATE_LOCALHOST; if (log) { struct pf_rule *lr; if (s != NULL && s->nat_rule.ptr != NULL && s->nat_rule.ptr->log & PF_LOG_ALL) lr = s->nat_rule.ptr; else lr = r; PFLOG_PACKET(kif, h, m, AF_INET6, dir, reason, lr, a, ruleset, &pd); } kif->pfik_bytes[1][dir == PF_OUT][action != PF_PASS] += pd.tot_len; kif->pfik_packets[1][dir == PF_OUT][action != PF_PASS]++; if (action == PF_PASS || r->action == PF_DROP) { dirndx = (dir == PF_OUT); r->packets[dirndx]++; r->bytes[dirndx] += pd.tot_len; if (a != NULL) { a->packets[dirndx]++; a->bytes[dirndx] += pd.tot_len; } if (s != NULL) { if (s->nat_rule.ptr != NULL) { s->nat_rule.ptr->packets[dirndx]++; s->nat_rule.ptr->bytes[dirndx] += pd.tot_len; } if (s->src_node != NULL) { s->src_node->packets[dirndx]++; s->src_node->bytes[dirndx] += pd.tot_len; } if (s->nat_src_node != NULL) { s->nat_src_node->packets[dirndx]++; s->nat_src_node->bytes[dirndx] += pd.tot_len; } dirndx = (dir == s->direction) ? 0 : 1; s->packets[dirndx]++; s->bytes[dirndx] += pd.tot_len; } tr = r; nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule; if (nr != NULL) { struct pf_addr *x; /* * XXX: we need to make sure that the addresses * passed to pfr_update_stats() are the same than * the addresses used during matching (pfr_match) */ if (r == &pf_default_rule) { tr = nr; x = (s == NULL || s->direction == dir) ? &pd.baddr : &pd.naddr; } else { x = (s == NULL || s->direction == dir) ? &pd.naddr : &pd.baddr; } if (x == &pd.baddr || s == NULL) { if (dir == PF_OUT) pd.src = x; else pd.dst = x; } } if (tr->src.addr.type == PF_ADDR_TABLE) pfr_update_stats(tr->src.addr.p.tbl, (s == NULL || s->direction == dir) ? pd.src : pd.dst, pd.af, pd.tot_len, dir == PF_OUT, r->action == PF_PASS, tr->src.neg); if (tr->dst.addr.type == PF_ADDR_TABLE) pfr_update_stats(tr->dst.addr.p.tbl, (s == NULL || s->direction == dir) ? pd.dst : pd.src, pd.af, pd.tot_len, dir == PF_OUT, r->action == PF_PASS, tr->dst.neg); } if (action == PF_SYNPROXY_DROP) { m_freem(*m0); *m0 = NULL; action = PF_PASS; } else if (r->rt) /* pf_route6 can free the mbuf causing *m0 to become NULL */ pf_route6(m0, r, dir, ifp, s, &pd); #ifdef __FreeBSD__ PF_UNLOCK(); #endif return (action); } #endif /* INET6 */ int pf_check_congestion(struct ifqueue *ifq) { #ifdef __FreeBSD__ /* XXX_IMPORT: later */ return (0); #else if (ifq->ifq_congestion) return (1); else return (0); #endif } Index: head/sys/netinet/in_pcb.c =================================================================== --- head/sys/netinet/in_pcb.c (revision 222487) +++ head/sys/netinet/in_pcb.c (revision 222488) @@ -1,2126 +1,2197 @@ /*- * Copyright (c) 1982, 1986, 1991, 1993, 1995 * The Regents of the University of California. * Copyright (c) 2007-2009 Robert N. M. Watson * Copyright (c) 2010-2011 Juniper Networks, Inc. * All rights reserved. * * Portions of this software were developed by Robert N. M. Watson under * contract to Juniper Networks, Inc. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)in_pcb.c 8.4 (Berkeley) 5/24/95 */ #include __FBSDID("$FreeBSD$"); #include "opt_ddb.h" #include "opt_ipsec.h" #include "opt_inet.h" #include "opt_inet6.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef DDB #include #endif #include #include #include #include #include #if defined(INET) || defined(INET6) #include #include #include #include #include #include #endif #ifdef INET #include #endif #ifdef INET6 #include #include #include #include #endif /* INET6 */ #ifdef IPSEC #include #include #endif /* IPSEC */ #include static struct callout ipport_tick_callout; /* * These configure the range of local port addresses assigned to * "unspecified" outgoing connections/packets/whatever. */ VNET_DEFINE(int, ipport_lowfirstauto) = IPPORT_RESERVED - 1; /* 1023 */ VNET_DEFINE(int, ipport_lowlastauto) = IPPORT_RESERVEDSTART; /* 600 */ VNET_DEFINE(int, ipport_firstauto) = IPPORT_EPHEMERALFIRST; /* 10000 */ VNET_DEFINE(int, ipport_lastauto) = IPPORT_EPHEMERALLAST; /* 65535 */ VNET_DEFINE(int, ipport_hifirstauto) = IPPORT_HIFIRSTAUTO; /* 49152 */ VNET_DEFINE(int, ipport_hilastauto) = IPPORT_HILASTAUTO; /* 65535 */ /* * Reserved ports accessible only to root. There are significant * security considerations that must be accounted for when changing these, * but the security benefits can be great. Please be careful. */ VNET_DEFINE(int, ipport_reservedhigh) = IPPORT_RESERVED - 1; /* 1023 */ VNET_DEFINE(int, ipport_reservedlow); /* Variables dealing with random ephemeral port allocation. */ VNET_DEFINE(int, ipport_randomized) = 1; /* user controlled via sysctl */ VNET_DEFINE(int, ipport_randomcps) = 10; /* user controlled via sysctl */ VNET_DEFINE(int, ipport_randomtime) = 45; /* user controlled via sysctl */ VNET_DEFINE(int, ipport_stoprandom); /* toggled by ipport_tick */ VNET_DEFINE(int, ipport_tcpallocs); static VNET_DEFINE(int, ipport_tcplastcount); #define V_ipport_tcplastcount VNET(ipport_tcplastcount) +static struct inpcb *in_pcblookup_hash_locked(struct inpcbinfo *pcbinfo, + struct in_addr faddr, u_int fport_arg, + struct in_addr laddr, u_int lport_arg, + int lookupflags, struct ifnet *ifp); static void in_pcbremlists(struct inpcb *inp); #ifdef INET #define RANGECHK(var, min, max) \ if ((var) < (min)) { (var) = (min); } \ else if ((var) > (max)) { (var) = (max); } static int sysctl_net_ipport_check(SYSCTL_HANDLER_ARGS) { int error; #ifdef VIMAGE error = vnet_sysctl_handle_int(oidp, arg1, arg2, req); #else error = sysctl_handle_int(oidp, arg1, arg2, req); #endif if (error == 0) { RANGECHK(V_ipport_lowfirstauto, 1, IPPORT_RESERVED - 1); RANGECHK(V_ipport_lowlastauto, 1, IPPORT_RESERVED - 1); RANGECHK(V_ipport_firstauto, IPPORT_RESERVED, IPPORT_MAX); RANGECHK(V_ipport_lastauto, IPPORT_RESERVED, IPPORT_MAX); RANGECHK(V_ipport_hifirstauto, IPPORT_RESERVED, IPPORT_MAX); RANGECHK(V_ipport_hilastauto, IPPORT_RESERVED, IPPORT_MAX); } return (error); } #undef RANGECHK SYSCTL_NODE(_net_inet_ip, IPPROTO_IP, portrange, CTLFLAG_RW, 0, "IP Ports"); SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, lowfirst, CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_lowfirstauto), 0, &sysctl_net_ipport_check, "I", ""); SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, lowlast, CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_lowlastauto), 0, &sysctl_net_ipport_check, "I", ""); SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, first, CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_firstauto), 0, &sysctl_net_ipport_check, "I", ""); SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, last, CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_lastauto), 0, &sysctl_net_ipport_check, "I", ""); SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, hifirst, CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_hifirstauto), 0, &sysctl_net_ipport_check, "I", ""); SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, hilast, CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_hilastauto), 0, &sysctl_net_ipport_check, "I", ""); SYSCTL_VNET_INT(_net_inet_ip_portrange, OID_AUTO, reservedhigh, CTLFLAG_RW|CTLFLAG_SECURE, &VNET_NAME(ipport_reservedhigh), 0, ""); SYSCTL_VNET_INT(_net_inet_ip_portrange, OID_AUTO, reservedlow, CTLFLAG_RW|CTLFLAG_SECURE, &VNET_NAME(ipport_reservedlow), 0, ""); SYSCTL_VNET_INT(_net_inet_ip_portrange, OID_AUTO, randomized, CTLFLAG_RW, &VNET_NAME(ipport_randomized), 0, "Enable random port allocation"); SYSCTL_VNET_INT(_net_inet_ip_portrange, OID_AUTO, randomcps, CTLFLAG_RW, &VNET_NAME(ipport_randomcps), 0, "Maximum number of random port " "allocations before switching to a sequental one"); SYSCTL_VNET_INT(_net_inet_ip_portrange, OID_AUTO, randomtime, CTLFLAG_RW, &VNET_NAME(ipport_randomtime), 0, "Minimum time to keep sequental port " "allocation before switching to a random one"); #endif /* * in_pcb.c: manage the Protocol Control Blocks. * * NOTE: It is assumed that most of these functions will be called with * the pcbinfo lock held, and often, the inpcb lock held, as these utility * functions often modify hash chains or addresses in pcbs. */ /* * Initialize an inpcbinfo -- we should be able to reduce the number of * arguments in time. */ void in_pcbinfo_init(struct inpcbinfo *pcbinfo, const char *name, struct inpcbhead *listhead, int hash_nelements, int porthash_nelements, char *inpcbzone_name, uma_init inpcbzone_init, uma_fini inpcbzone_fini, uint32_t inpcbzone_flags) { INP_INFO_LOCK_INIT(pcbinfo, name); + INP_HASH_LOCK_INIT(pcbinfo, "pcbinfohash"); /* XXXRW: argument? */ #ifdef VIMAGE pcbinfo->ipi_vnet = curvnet; #endif pcbinfo->ipi_listhead = listhead; LIST_INIT(pcbinfo->ipi_listhead); + pcbinfo->ipi_count = 0; pcbinfo->ipi_hashbase = hashinit(hash_nelements, M_PCB, &pcbinfo->ipi_hashmask); pcbinfo->ipi_porthashbase = hashinit(porthash_nelements, M_PCB, &pcbinfo->ipi_porthashmask); pcbinfo->ipi_zone = uma_zcreate(inpcbzone_name, sizeof(struct inpcb), NULL, NULL, inpcbzone_init, inpcbzone_fini, UMA_ALIGN_PTR, inpcbzone_flags); uma_zone_set_max(pcbinfo->ipi_zone, maxsockets); } /* * Destroy an inpcbinfo. */ void in_pcbinfo_destroy(struct inpcbinfo *pcbinfo) { + KASSERT(pcbinfo->ipi_count == 0, + ("%s: ipi_count = %u", __func__, pcbinfo->ipi_count)); + hashdestroy(pcbinfo->ipi_hashbase, M_PCB, pcbinfo->ipi_hashmask); hashdestroy(pcbinfo->ipi_porthashbase, M_PCB, pcbinfo->ipi_porthashmask); uma_zdestroy(pcbinfo->ipi_zone); + INP_HASH_LOCK_DESTROY(pcbinfo); INP_INFO_LOCK_DESTROY(pcbinfo); } /* * Allocate a PCB and associate it with the socket. * On success return with the PCB locked. */ int in_pcballoc(struct socket *so, struct inpcbinfo *pcbinfo) { struct inpcb *inp; int error; INP_INFO_WLOCK_ASSERT(pcbinfo); error = 0; inp = uma_zalloc(pcbinfo->ipi_zone, M_NOWAIT); if (inp == NULL) return (ENOBUFS); bzero(inp, inp_zero_size); inp->inp_pcbinfo = pcbinfo; inp->inp_socket = so; inp->inp_cred = crhold(so->so_cred); inp->inp_inc.inc_fibnum = so->so_fibnum; #ifdef MAC error = mac_inpcb_init(inp, M_NOWAIT); if (error != 0) goto out; mac_inpcb_create(so, inp); #endif #ifdef IPSEC error = ipsec_init_policy(so, &inp->inp_sp); if (error != 0) { #ifdef MAC mac_inpcb_destroy(inp); #endif goto out; } #endif /*IPSEC*/ #ifdef INET6 if (INP_SOCKAF(so) == AF_INET6) { inp->inp_vflag |= INP_IPV6PROTO; if (V_ip6_v6only) inp->inp_flags |= IN6P_IPV6_V6ONLY; } #endif LIST_INSERT_HEAD(pcbinfo->ipi_listhead, inp, inp_list); pcbinfo->ipi_count++; so->so_pcb = (caddr_t)inp; #ifdef INET6 if (V_ip6_auto_flowlabel) inp->inp_flags |= IN6P_AUTOFLOWLABEL; #endif INP_WLOCK(inp); inp->inp_gencnt = ++pcbinfo->ipi_gencnt; refcount_init(&inp->inp_refcount, 1); /* Reference from inpcbinfo */ #if defined(IPSEC) || defined(MAC) out: if (error != 0) { crfree(inp->inp_cred); uma_zfree(pcbinfo->ipi_zone, inp); } #endif return (error); } #ifdef INET int in_pcbbind(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred) { int anonport, error; - INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo); INP_WLOCK_ASSERT(inp); + INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo); if (inp->inp_lport != 0 || inp->inp_laddr.s_addr != INADDR_ANY) return (EINVAL); anonport = inp->inp_lport == 0 && (nam == NULL || ((struct sockaddr_in *)nam)->sin_port == 0); error = in_pcbbind_setup(inp, nam, &inp->inp_laddr.s_addr, &inp->inp_lport, cred); if (error) return (error); if (in_pcbinshash(inp) != 0) { inp->inp_laddr.s_addr = INADDR_ANY; inp->inp_lport = 0; return (EAGAIN); } if (anonport) inp->inp_flags |= INP_ANONPORT; return (0); } #endif #if defined(INET) || defined(INET6) int in_pcb_lport(struct inpcb *inp, struct in_addr *laddrp, u_short *lportp, struct ucred *cred, int lookupflags) { struct inpcbinfo *pcbinfo; struct inpcb *tmpinp; unsigned short *lastport; int count, dorandom, error; u_short aux, first, last, lport; #ifdef INET struct in_addr laddr; #endif pcbinfo = inp->inp_pcbinfo; /* * Because no actual state changes occur here, a global write lock on * the pcbinfo isn't required. */ - INP_INFO_LOCK_ASSERT(pcbinfo); INP_LOCK_ASSERT(inp); + INP_HASH_LOCK_ASSERT(pcbinfo); if (inp->inp_flags & INP_HIGHPORT) { first = V_ipport_hifirstauto; /* sysctl */ last = V_ipport_hilastauto; lastport = &pcbinfo->ipi_lasthi; } else if (inp->inp_flags & INP_LOWPORT) { error = priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT, 0); if (error) return (error); first = V_ipport_lowfirstauto; /* 1023 */ last = V_ipport_lowlastauto; /* 600 */ lastport = &pcbinfo->ipi_lastlow; } else { first = V_ipport_firstauto; /* sysctl */ last = V_ipport_lastauto; lastport = &pcbinfo->ipi_lastport; } /* * For UDP, use random port allocation as long as the user * allows it. For TCP (and as of yet unknown) connections, * use random port allocation only if the user allows it AND * ipport_tick() allows it. */ if (V_ipport_randomized && (!V_ipport_stoprandom || pcbinfo == &V_udbinfo)) dorandom = 1; else dorandom = 0; /* * It makes no sense to do random port allocation if * we have the only port available. */ if (first == last) dorandom = 0; /* Make sure to not include UDP packets in the count. */ if (pcbinfo != &V_udbinfo) V_ipport_tcpallocs++; /* * Instead of having two loops further down counting up or down * make sure that first is always <= last and go with only one * code path implementing all logic. */ if (first > last) { aux = first; first = last; last = aux; } #ifdef INET /* Make the compiler happy. */ laddr.s_addr = 0; if ((inp->inp_vflag & (INP_IPV4|INP_IPV6)) == INP_IPV4) { KASSERT(laddrp != NULL, ("%s: laddrp NULL for v4 inp %p", __func__, inp)); laddr = *laddrp; } #endif tmpinp = NULL; /* Make compiler happy. */ lport = *lportp; if (dorandom) *lastport = first + (arc4random() % (last - first)); count = last - first; do { if (count-- < 0) /* completely used? */ return (EADDRNOTAVAIL); ++*lastport; if (*lastport < first || *lastport > last) *lastport = first; lport = htons(*lastport); #ifdef INET6 if ((inp->inp_vflag & INP_IPV6) != 0) tmpinp = in6_pcblookup_local(pcbinfo, &inp->in6p_laddr, lport, lookupflags, cred); #endif #if defined(INET) && defined(INET6) else #endif #ifdef INET tmpinp = in_pcblookup_local(pcbinfo, laddr, lport, lookupflags, cred); #endif } while (tmpinp != NULL); #ifdef INET if ((inp->inp_vflag & (INP_IPV4|INP_IPV6)) == INP_IPV4) laddrp->s_addr = laddr.s_addr; #endif *lportp = lport; return (0); } #endif /* INET || INET6 */ #ifdef INET /* * Set up a bind operation on a PCB, performing port allocation * as required, but do not actually modify the PCB. Callers can * either complete the bind by setting inp_laddr/inp_lport and * calling in_pcbinshash(), or they can just use the resulting * port and address to authorise the sending of a once-off packet. * * On error, the values of *laddrp and *lportp are not changed. */ int in_pcbbind_setup(struct inpcb *inp, struct sockaddr *nam, in_addr_t *laddrp, u_short *lportp, struct ucred *cred) { struct socket *so = inp->inp_socket; struct sockaddr_in *sin; struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; struct in_addr laddr; u_short lport = 0; int lookupflags = 0, reuseport = (so->so_options & SO_REUSEPORT); int error; /* - * Because no actual state changes occur here, a global write lock on - * the pcbinfo isn't required. + * No state changes, so read locks are sufficient here. */ - INP_INFO_LOCK_ASSERT(pcbinfo); INP_LOCK_ASSERT(inp); + INP_HASH_LOCK_ASSERT(pcbinfo); if (TAILQ_EMPTY(&V_in_ifaddrhead)) /* XXX broken! */ return (EADDRNOTAVAIL); laddr.s_addr = *laddrp; if (nam != NULL && laddr.s_addr != INADDR_ANY) return (EINVAL); if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0) lookupflags = INPLOOKUP_WILDCARD; if (nam == NULL) { if ((error = prison_local_ip4(cred, &laddr)) != 0) return (error); } else { sin = (struct sockaddr_in *)nam; if (nam->sa_len != sizeof (*sin)) return (EINVAL); #ifdef notdef /* * We should check the family, but old programs * incorrectly fail to initialize it. */ if (sin->sin_family != AF_INET) return (EAFNOSUPPORT); #endif error = prison_local_ip4(cred, &sin->sin_addr); if (error) return (error); if (sin->sin_port != *lportp) { /* Don't allow the port to change. */ if (*lportp != 0) return (EINVAL); lport = sin->sin_port; } /* NB: lport is left as 0 if the port isn't being changed. */ if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) { /* * Treat SO_REUSEADDR as SO_REUSEPORT for multicast; * allow complete duplication of binding if * SO_REUSEPORT is set, or if SO_REUSEADDR is set * and a multicast address is bound on both * new and duplicated sockets. */ if (so->so_options & SO_REUSEADDR) reuseport = SO_REUSEADDR|SO_REUSEPORT; } else if (sin->sin_addr.s_addr != INADDR_ANY) { sin->sin_port = 0; /* yech... */ bzero(&sin->sin_zero, sizeof(sin->sin_zero)); /* * Is the address a local IP address? * If INP_BINDANY is set, then the socket may be bound * to any endpoint address, local or not. */ if ((inp->inp_flags & INP_BINDANY) == 0 && ifa_ifwithaddr_check((struct sockaddr *)sin) == 0) return (EADDRNOTAVAIL); } laddr = sin->sin_addr; if (lport) { struct inpcb *t; struct tcptw *tw; /* GROSS */ if (ntohs(lport) <= V_ipport_reservedhigh && ntohs(lport) >= V_ipport_reservedlow && priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT, 0)) return (EACCES); if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)) && priv_check_cred(inp->inp_cred, PRIV_NETINET_REUSEPORT, 0) != 0) { t = in_pcblookup_local(pcbinfo, sin->sin_addr, lport, INPLOOKUP_WILDCARD, cred); /* * XXX * This entire block sorely needs a rewrite. */ if (t && ((t->inp_flags & INP_TIMEWAIT) == 0) && (so->so_type != SOCK_STREAM || ntohl(t->inp_faddr.s_addr) == INADDR_ANY) && (ntohl(sin->sin_addr.s_addr) != INADDR_ANY || ntohl(t->inp_laddr.s_addr) != INADDR_ANY || (t->inp_socket->so_options & SO_REUSEPORT) == 0) && (inp->inp_cred->cr_uid != t->inp_cred->cr_uid)) return (EADDRINUSE); } t = in_pcblookup_local(pcbinfo, sin->sin_addr, lport, lookupflags, cred); if (t && (t->inp_flags & INP_TIMEWAIT)) { /* * XXXRW: If an incpb has had its timewait * state recycled, we treat the address as * being in use (for now). This is better * than a panic, but not desirable. */ tw = intotw(inp); if (tw == NULL || (reuseport & tw->tw_so_options) == 0) return (EADDRINUSE); } else if (t && (reuseport & t->inp_socket->so_options) == 0) { #ifdef INET6 if (ntohl(sin->sin_addr.s_addr) != INADDR_ANY || ntohl(t->inp_laddr.s_addr) != INADDR_ANY || INP_SOCKAF(so) == INP_SOCKAF(t->inp_socket)) #endif return (EADDRINUSE); } } } if (*lportp != 0) lport = *lportp; if (lport == 0) { error = in_pcb_lport(inp, &laddr, &lport, cred, lookupflags); if (error != 0) return (error); } *laddrp = laddr.s_addr; *lportp = lport; return (0); } /* * Connect from a socket to a specified address. * Both address and port must be specified in argument sin. * If don't have a local address for this socket yet, * then pick one. */ int in_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred) { u_short lport, fport; in_addr_t laddr, faddr; int anonport, error; - INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo); INP_WLOCK_ASSERT(inp); + INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo); lport = inp->inp_lport; laddr = inp->inp_laddr.s_addr; anonport = (lport == 0); error = in_pcbconnect_setup(inp, nam, &laddr, &lport, &faddr, &fport, NULL, cred); if (error) return (error); /* Do the initial binding of the local address if required. */ if (inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0) { inp->inp_lport = lport; inp->inp_laddr.s_addr = laddr; if (in_pcbinshash(inp) != 0) { inp->inp_laddr.s_addr = INADDR_ANY; inp->inp_lport = 0; return (EAGAIN); } } /* Commit the remaining changes. */ inp->inp_lport = lport; inp->inp_laddr.s_addr = laddr; inp->inp_faddr.s_addr = faddr; inp->inp_fport = fport; in_pcbrehash(inp); if (anonport) inp->inp_flags |= INP_ANONPORT; return (0); } /* * Do proper source address selection on an unbound socket in case * of connect. Take jails into account as well. */ static int in_pcbladdr(struct inpcb *inp, struct in_addr *faddr, struct in_addr *laddr, struct ucred *cred) { struct ifaddr *ifa; struct sockaddr *sa; struct sockaddr_in *sin; struct route sro; int error; KASSERT(laddr != NULL, ("%s: laddr NULL", __func__)); /* * Bypass source address selection and use the primary jail IP * if requested. */ if (cred != NULL && !prison_saddrsel_ip4(cred, laddr)) return (0); error = 0; bzero(&sro, sizeof(sro)); sin = (struct sockaddr_in *)&sro.ro_dst; sin->sin_family = AF_INET; sin->sin_len = sizeof(struct sockaddr_in); sin->sin_addr.s_addr = faddr->s_addr; /* * If route is known our src addr is taken from the i/f, * else punt. * * Find out route to destination. */ if ((inp->inp_socket->so_options & SO_DONTROUTE) == 0) in_rtalloc_ign(&sro, 0, inp->inp_inc.inc_fibnum); /* * If we found a route, use the address corresponding to * the outgoing interface. * * Otherwise assume faddr is reachable on a directly connected * network and try to find a corresponding interface to take * the source address from. */ if (sro.ro_rt == NULL || sro.ro_rt->rt_ifp == NULL) { struct in_ifaddr *ia; struct ifnet *ifp; ia = ifatoia(ifa_ifwithdstaddr((struct sockaddr *)sin)); if (ia == NULL) ia = ifatoia(ifa_ifwithnet((struct sockaddr *)sin, 0)); if (ia == NULL) { error = ENETUNREACH; goto done; } if (cred == NULL || !prison_flag(cred, PR_IP4)) { laddr->s_addr = ia->ia_addr.sin_addr.s_addr; ifa_free(&ia->ia_ifa); goto done; } ifp = ia->ia_ifp; ifa_free(&ia->ia_ifa); ia = NULL; IF_ADDR_LOCK(ifp); TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { sa = ifa->ifa_addr; if (sa->sa_family != AF_INET) continue; sin = (struct sockaddr_in *)sa; if (prison_check_ip4(cred, &sin->sin_addr) == 0) { ia = (struct in_ifaddr *)ifa; break; } } if (ia != NULL) { laddr->s_addr = ia->ia_addr.sin_addr.s_addr; IF_ADDR_UNLOCK(ifp); goto done; } IF_ADDR_UNLOCK(ifp); /* 3. As a last resort return the 'default' jail address. */ error = prison_get_ip4(cred, laddr); goto done; } /* * If the outgoing interface on the route found is not * a loopback interface, use the address from that interface. * In case of jails do those three steps: * 1. check if the interface address belongs to the jail. If so use it. * 2. check if we have any address on the outgoing interface * belonging to this jail. If so use it. * 3. as a last resort return the 'default' jail address. */ if ((sro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0) { struct in_ifaddr *ia; struct ifnet *ifp; /* If not jailed, use the default returned. */ if (cred == NULL || !prison_flag(cred, PR_IP4)) { ia = (struct in_ifaddr *)sro.ro_rt->rt_ifa; laddr->s_addr = ia->ia_addr.sin_addr.s_addr; goto done; } /* Jailed. */ /* 1. Check if the iface address belongs to the jail. */ sin = (struct sockaddr_in *)sro.ro_rt->rt_ifa->ifa_addr; if (prison_check_ip4(cred, &sin->sin_addr) == 0) { ia = (struct in_ifaddr *)sro.ro_rt->rt_ifa; laddr->s_addr = ia->ia_addr.sin_addr.s_addr; goto done; } /* * 2. Check if we have any address on the outgoing interface * belonging to this jail. */ ia = NULL; ifp = sro.ro_rt->rt_ifp; IF_ADDR_LOCK(ifp); TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { sa = ifa->ifa_addr; if (sa->sa_family != AF_INET) continue; sin = (struct sockaddr_in *)sa; if (prison_check_ip4(cred, &sin->sin_addr) == 0) { ia = (struct in_ifaddr *)ifa; break; } } if (ia != NULL) { laddr->s_addr = ia->ia_addr.sin_addr.s_addr; IF_ADDR_UNLOCK(ifp); goto done; } IF_ADDR_UNLOCK(ifp); /* 3. As a last resort return the 'default' jail address. */ error = prison_get_ip4(cred, laddr); goto done; } /* * The outgoing interface is marked with 'loopback net', so a route * to ourselves is here. * Try to find the interface of the destination address and then * take the address from there. That interface is not necessarily * a loopback interface. * In case of jails, check that it is an address of the jail * and if we cannot find, fall back to the 'default' jail address. */ if ((sro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) { struct sockaddr_in sain; struct in_ifaddr *ia; bzero(&sain, sizeof(struct sockaddr_in)); sain.sin_family = AF_INET; sain.sin_len = sizeof(struct sockaddr_in); sain.sin_addr.s_addr = faddr->s_addr; ia = ifatoia(ifa_ifwithdstaddr(sintosa(&sain))); if (ia == NULL) ia = ifatoia(ifa_ifwithnet(sintosa(&sain), 0)); if (ia == NULL) ia = ifatoia(ifa_ifwithaddr(sintosa(&sain))); if (cred == NULL || !prison_flag(cred, PR_IP4)) { if (ia == NULL) { error = ENETUNREACH; goto done; } laddr->s_addr = ia->ia_addr.sin_addr.s_addr; ifa_free(&ia->ia_ifa); goto done; } /* Jailed. */ if (ia != NULL) { struct ifnet *ifp; ifp = ia->ia_ifp; ifa_free(&ia->ia_ifa); ia = NULL; IF_ADDR_LOCK(ifp); TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { sa = ifa->ifa_addr; if (sa->sa_family != AF_INET) continue; sin = (struct sockaddr_in *)sa; if (prison_check_ip4(cred, &sin->sin_addr) == 0) { ia = (struct in_ifaddr *)ifa; break; } } if (ia != NULL) { laddr->s_addr = ia->ia_addr.sin_addr.s_addr; IF_ADDR_UNLOCK(ifp); goto done; } IF_ADDR_UNLOCK(ifp); } /* 3. As a last resort return the 'default' jail address. */ error = prison_get_ip4(cred, laddr); goto done; } done: if (sro.ro_rt != NULL) RTFREE(sro.ro_rt); return (error); } /* * Set up for a connect from a socket to the specified address. * On entry, *laddrp and *lportp should contain the current local * address and port for the PCB; these are updated to the values * that should be placed in inp_laddr and inp_lport to complete * the connect. * * On success, *faddrp and *fportp will be set to the remote address * and port. These are not updated in the error case. * * If the operation fails because the connection already exists, * *oinpp will be set to the PCB of that connection so that the * caller can decide to override it. In all other cases, *oinpp * is set to NULL. */ int in_pcbconnect_setup(struct inpcb *inp, struct sockaddr *nam, in_addr_t *laddrp, u_short *lportp, in_addr_t *faddrp, u_short *fportp, struct inpcb **oinpp, struct ucred *cred) { struct sockaddr_in *sin = (struct sockaddr_in *)nam; struct in_ifaddr *ia; struct inpcb *oinp; struct in_addr laddr, faddr; u_short lport, fport; int error; /* * Because a global state change doesn't actually occur here, a read * lock is sufficient. */ - INP_INFO_LOCK_ASSERT(inp->inp_pcbinfo); INP_LOCK_ASSERT(inp); + INP_HASH_LOCK_ASSERT(inp->inp_pcbinfo); if (oinpp != NULL) *oinpp = NULL; if (nam->sa_len != sizeof (*sin)) return (EINVAL); if (sin->sin_family != AF_INET) return (EAFNOSUPPORT); if (sin->sin_port == 0) return (EADDRNOTAVAIL); laddr.s_addr = *laddrp; lport = *lportp; faddr = sin->sin_addr; fport = sin->sin_port; if (!TAILQ_EMPTY(&V_in_ifaddrhead)) { /* * If the destination address is INADDR_ANY, * use the primary local address. * If the supplied address is INADDR_BROADCAST, * and the primary interface supports broadcast, * choose the broadcast address for that interface. */ if (faddr.s_addr == INADDR_ANY) { IN_IFADDR_RLOCK(); faddr = IA_SIN(TAILQ_FIRST(&V_in_ifaddrhead))->sin_addr; IN_IFADDR_RUNLOCK(); if (cred != NULL && (error = prison_get_ip4(cred, &faddr)) != 0) return (error); } else if (faddr.s_addr == (u_long)INADDR_BROADCAST) { IN_IFADDR_RLOCK(); if (TAILQ_FIRST(&V_in_ifaddrhead)->ia_ifp->if_flags & IFF_BROADCAST) faddr = satosin(&TAILQ_FIRST( &V_in_ifaddrhead)->ia_broadaddr)->sin_addr; IN_IFADDR_RUNLOCK(); } } if (laddr.s_addr == INADDR_ANY) { error = in_pcbladdr(inp, &faddr, &laddr, cred); /* * If the destination address is multicast and an outgoing * interface has been set as a multicast option, prefer the * address of that interface as our source address. */ if (IN_MULTICAST(ntohl(faddr.s_addr)) && inp->inp_moptions != NULL) { struct ip_moptions *imo; struct ifnet *ifp; imo = inp->inp_moptions; if (imo->imo_multicast_ifp != NULL) { ifp = imo->imo_multicast_ifp; IN_IFADDR_RLOCK(); TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { if ((ia->ia_ifp == ifp) && (cred == NULL || prison_check_ip4(cred, &ia->ia_addr.sin_addr) == 0)) break; } if (ia == NULL) error = EADDRNOTAVAIL; else { laddr = ia->ia_addr.sin_addr; error = 0; } IN_IFADDR_RUNLOCK(); } } if (error) return (error); } - oinp = in_pcblookup_hash(inp->inp_pcbinfo, faddr, fport, laddr, lport, - 0, NULL); + oinp = in_pcblookup_hash_locked(inp->inp_pcbinfo, faddr, fport, + laddr, lport, 0, NULL); if (oinp != NULL) { if (oinpp != NULL) *oinpp = oinp; return (EADDRINUSE); } if (lport == 0) { error = in_pcbbind_setup(inp, NULL, &laddr.s_addr, &lport, cred); if (error) return (error); } *laddrp = laddr.s_addr; *lportp = lport; *faddrp = faddr.s_addr; *fportp = fport; return (0); } void in_pcbdisconnect(struct inpcb *inp) { - INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo); INP_WLOCK_ASSERT(inp); + INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo); inp->inp_faddr.s_addr = INADDR_ANY; inp->inp_fport = 0; in_pcbrehash(inp); } #endif /* * in_pcbdetach() is responsibe for disassociating a socket from an inpcb. * For most protocols, this will be invoked immediately prior to calling * in_pcbfree(). However, with TCP the inpcb may significantly outlive the * socket, in which case in_pcbfree() is deferred. */ void in_pcbdetach(struct inpcb *inp) { KASSERT(inp->inp_socket != NULL, ("%s: inp_socket == NULL", __func__)); inp->inp_socket->so_pcb = NULL; inp->inp_socket = NULL; } /* * in_pcbref() bumps the reference count on an inpcb in order to maintain * stability of an inpcb pointer despite the inpcb lock being released. This * is used in TCP when the inpcbinfo lock needs to be acquired or upgraded, * but where the inpcb lock is already held. * * in_pcbref() should be used only to provide brief memory stability, and * must always be followed by a call to INP_WLOCK() and in_pcbrele() to * garbage collect the inpcb if it has been in_pcbfree()'d from another * context. Until in_pcbrele() has returned that the inpcb is still valid, * lock and rele are the *only* safe operations that may be performed on the * inpcb. * * While the inpcb will not be freed, releasing the inpcb lock means that the * connection's state may change, so the caller should be careful to * revalidate any cached state on reacquiring the lock. Drop the reference * using in_pcbrele(). */ void in_pcbref(struct inpcb *inp) { KASSERT(inp->inp_refcount > 0, ("%s: refcount 0", __func__)); refcount_acquire(&inp->inp_refcount); } /* * Drop a refcount on an inpcb elevated using in_pcbref(); because a call to * in_pcbfree() may have been made between in_pcbref() and in_pcbrele(), we * return a flag indicating whether or not the inpcb remains valid. If it is * valid, we return with the inpcb lock held. * * Notice that, unlike in_pcbref(), the inpcb lock must be held to drop a * reference on an inpcb. Historically more work was done here (actually, in * in_pcbfree_internal()) but has been moved to in_pcbfree() to avoid the * need for the pcbinfo lock in in_pcbrele(). Deferring the free is entirely * about memory stability (and continued use of the write lock). */ int in_pcbrele_rlocked(struct inpcb *inp) { struct inpcbinfo *pcbinfo; KASSERT(inp->inp_refcount > 0, ("%s: refcount 0", __func__)); INP_RLOCK_ASSERT(inp); if (refcount_release(&inp->inp_refcount) == 0) return (0); KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL", __func__)); INP_RUNLOCK(inp); pcbinfo = inp->inp_pcbinfo; uma_zfree(pcbinfo->ipi_zone, inp); return (1); } int in_pcbrele_wlocked(struct inpcb *inp) { struct inpcbinfo *pcbinfo; KASSERT(inp->inp_refcount > 0, ("%s: refcount 0", __func__)); INP_WLOCK_ASSERT(inp); if (refcount_release(&inp->inp_refcount) == 0) return (0); KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL", __func__)); INP_WUNLOCK(inp); pcbinfo = inp->inp_pcbinfo; uma_zfree(pcbinfo->ipi_zone, inp); return (1); } /* * Temporary wrapper. */ int in_pcbrele(struct inpcb *inp) { return (in_pcbrele_wlocked(inp)); } /* * Unconditionally schedule an inpcb to be freed by decrementing its * reference count, which should occur only after the inpcb has been detached * from its socket. If another thread holds a temporary reference (acquired * using in_pcbref()) then the free is deferred until that reference is * released using in_pcbrele(), but the inpcb is still unlocked. Almost all * work, including removal from global lists, is done in this context, where * the pcbinfo lock is held. */ void in_pcbfree(struct inpcb *inp) { struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL", __func__)); INP_INFO_WLOCK_ASSERT(pcbinfo); INP_WLOCK_ASSERT(inp); /* XXXRW: Do as much as possible here. */ #ifdef IPSEC if (inp->inp_sp != NULL) ipsec_delete_pcbpolicy(inp); #endif /* IPSEC */ inp->inp_gencnt = ++pcbinfo->ipi_gencnt; in_pcbremlists(inp); #ifdef INET6 if (inp->inp_vflag & INP_IPV6PROTO) { ip6_freepcbopts(inp->in6p_outputopts); if (inp->in6p_moptions != NULL) ip6_freemoptions(inp->in6p_moptions); } #endif if (inp->inp_options) (void)m_free(inp->inp_options); #ifdef INET if (inp->inp_moptions != NULL) inp_freemoptions(inp->inp_moptions); #endif inp->inp_vflag = 0; crfree(inp->inp_cred); #ifdef MAC mac_inpcb_destroy(inp); #endif if (!in_pcbrele_wlocked(inp)) INP_WUNLOCK(inp); } /* * in_pcbdrop() removes an inpcb from hashed lists, releasing its address and * port reservation, and preventing it from being returned by inpcb lookups. * * It is used by TCP to mark an inpcb as unused and avoid future packet * delivery or event notification when a socket remains open but TCP has * closed. This might occur as a result of a shutdown()-initiated TCP close * or a RST on the wire, and allows the port binding to be reused while still * maintaining the invariant that so_pcb always points to a valid inpcb until * in_pcbdetach(). * * XXXRW: Possibly in_pcbdrop() should also prevent future notifications by * in_pcbnotifyall() and in_pcbpurgeif0()? */ void in_pcbdrop(struct inpcb *inp) { - INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo); INP_WLOCK_ASSERT(inp); + /* + * XXXRW: Possibly we should protect the setting of INP_DROPPED with + * the hash lock...? + */ inp->inp_flags |= INP_DROPPED; if (inp->inp_flags & INP_INHASHLIST) { struct inpcbport *phd = inp->inp_phd; + INP_HASH_WLOCK(inp->inp_pcbinfo); LIST_REMOVE(inp, inp_hash); LIST_REMOVE(inp, inp_portlist); if (LIST_FIRST(&phd->phd_pcblist) == NULL) { LIST_REMOVE(phd, phd_hash); free(phd, M_PCB); } + INP_HASH_WUNLOCK(inp->inp_pcbinfo); inp->inp_flags &= ~INP_INHASHLIST; } } #ifdef INET /* * Common routines to return the socket addresses associated with inpcbs. */ struct sockaddr * in_sockaddr(in_port_t port, struct in_addr *addr_p) { struct sockaddr_in *sin; sin = malloc(sizeof *sin, M_SONAME, M_WAITOK | M_ZERO); sin->sin_family = AF_INET; sin->sin_len = sizeof(*sin); sin->sin_addr = *addr_p; sin->sin_port = port; return (struct sockaddr *)sin; } int in_getsockaddr(struct socket *so, struct sockaddr **nam) { struct inpcb *inp; struct in_addr addr; in_port_t port; inp = sotoinpcb(so); KASSERT(inp != NULL, ("in_getsockaddr: inp == NULL")); INP_RLOCK(inp); port = inp->inp_lport; addr = inp->inp_laddr; INP_RUNLOCK(inp); *nam = in_sockaddr(port, &addr); return 0; } int in_getpeeraddr(struct socket *so, struct sockaddr **nam) { struct inpcb *inp; struct in_addr addr; in_port_t port; inp = sotoinpcb(so); KASSERT(inp != NULL, ("in_getpeeraddr: inp == NULL")); INP_RLOCK(inp); port = inp->inp_fport; addr = inp->inp_faddr; INP_RUNLOCK(inp); *nam = in_sockaddr(port, &addr); return 0; } void in_pcbnotifyall(struct inpcbinfo *pcbinfo, struct in_addr faddr, int errno, struct inpcb *(*notify)(struct inpcb *, int)) { struct inpcb *inp, *inp_temp; INP_INFO_WLOCK(pcbinfo); LIST_FOREACH_SAFE(inp, pcbinfo->ipi_listhead, inp_list, inp_temp) { INP_WLOCK(inp); #ifdef INET6 if ((inp->inp_vflag & INP_IPV4) == 0) { INP_WUNLOCK(inp); continue; } #endif if (inp->inp_faddr.s_addr != faddr.s_addr || inp->inp_socket == NULL) { INP_WUNLOCK(inp); continue; } if ((*notify)(inp, errno)) INP_WUNLOCK(inp); } INP_INFO_WUNLOCK(pcbinfo); } void in_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp) { struct inpcb *inp; struct ip_moptions *imo; int i, gap; INP_INFO_RLOCK(pcbinfo); LIST_FOREACH(inp, pcbinfo->ipi_listhead, inp_list) { INP_WLOCK(inp); imo = inp->inp_moptions; if ((inp->inp_vflag & INP_IPV4) && imo != NULL) { /* * Unselect the outgoing interface if it is being * detached. */ if (imo->imo_multicast_ifp == ifp) imo->imo_multicast_ifp = NULL; /* * Drop multicast group membership if we joined * through the interface being detached. */ for (i = 0, gap = 0; i < imo->imo_num_memberships; i++) { if (imo->imo_membership[i]->inm_ifp == ifp) { in_delmulti(imo->imo_membership[i]); gap++; } else if (gap != 0) imo->imo_membership[i - gap] = imo->imo_membership[i]; } imo->imo_num_memberships -= gap; } INP_WUNLOCK(inp); } INP_INFO_RUNLOCK(pcbinfo); } /* - * Lookup a PCB based on the local address and port. + * Lookup a PCB based on the local address and port. Caller must hold the + * hash lock. No inpcb locks or references are acquired. */ #define INP_LOOKUP_MAPPED_PCB_COST 3 struct inpcb * in_pcblookup_local(struct inpcbinfo *pcbinfo, struct in_addr laddr, u_short lport, int lookupflags, struct ucred *cred) { struct inpcb *inp; #ifdef INET6 int matchwild = 3 + INP_LOOKUP_MAPPED_PCB_COST; #else int matchwild = 3; #endif int wildcard; KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0, ("%s: invalid lookup flags %d", __func__, lookupflags)); - INP_INFO_LOCK_ASSERT(pcbinfo); + INP_HASH_LOCK_ASSERT(pcbinfo); if ((lookupflags & INPLOOKUP_WILDCARD) == 0) { struct inpcbhead *head; /* * Look for an unconnected (wildcard foreign addr) PCB that * matches the local address and port we're looking for. */ head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport, 0, pcbinfo->ipi_hashmask)]; LIST_FOREACH(inp, head, inp_hash) { #ifdef INET6 /* XXX inp locking */ if ((inp->inp_vflag & INP_IPV4) == 0) continue; #endif if (inp->inp_faddr.s_addr == INADDR_ANY && inp->inp_laddr.s_addr == laddr.s_addr && inp->inp_lport == lport) { /* * Found? */ if (cred == NULL || prison_equal_ip4(cred->cr_prison, inp->inp_cred->cr_prison)) return (inp); } } /* * Not found. */ return (NULL); } else { struct inpcbporthead *porthash; struct inpcbport *phd; struct inpcb *match = NULL; /* * Best fit PCB lookup. * * First see if this local port is in use by looking on the * port hash list. */ porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport, pcbinfo->ipi_porthashmask)]; LIST_FOREACH(phd, porthash, phd_hash) { if (phd->phd_port == lport) break; } if (phd != NULL) { /* * Port is in use by one or more PCBs. Look for best * fit. */ LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) { wildcard = 0; if (cred != NULL && !prison_equal_ip4(inp->inp_cred->cr_prison, cred->cr_prison)) continue; #ifdef INET6 /* XXX inp locking */ if ((inp->inp_vflag & INP_IPV4) == 0) continue; /* * We never select the PCB that has * INP_IPV6 flag and is bound to :: if * we have another PCB which is bound * to 0.0.0.0. If a PCB has the * INP_IPV6 flag, then we set its cost * higher than IPv4 only PCBs. * * Note that the case only happens * when a socket is bound to ::, under * the condition that the use of the * mapped address is allowed. */ if ((inp->inp_vflag & INP_IPV6) != 0) wildcard += INP_LOOKUP_MAPPED_PCB_COST; #endif if (inp->inp_faddr.s_addr != INADDR_ANY) wildcard++; if (inp->inp_laddr.s_addr != INADDR_ANY) { if (laddr.s_addr == INADDR_ANY) wildcard++; else if (inp->inp_laddr.s_addr != laddr.s_addr) continue; } else { if (laddr.s_addr != INADDR_ANY) wildcard++; } if (wildcard < matchwild) { match = inp; matchwild = wildcard; if (matchwild == 0) break; } } } return (match); } } #undef INP_LOOKUP_MAPPED_PCB_COST /* - * Lookup PCB in hash list. + * Lookup PCB in hash list, using pcbinfo tables. This variation assumes + * that the caller has locked the hash list, and will not perform any further + * locking or reference operations on either the hash list or the connection. */ -struct inpcb * -in_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in_addr faddr, +static struct inpcb * +in_pcblookup_hash_locked(struct inpcbinfo *pcbinfo, struct in_addr faddr, u_int fport_arg, struct in_addr laddr, u_int lport_arg, int lookupflags, struct ifnet *ifp) { struct inpcbhead *head; struct inpcb *inp, *tmpinp; u_short fport = fport_arg, lport = lport_arg; KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0, ("%s: invalid lookup flags %d", __func__, lookupflags)); - INP_INFO_LOCK_ASSERT(pcbinfo); + INP_HASH_LOCK_ASSERT(pcbinfo); /* * First look for an exact match. */ tmpinp = NULL; head = &pcbinfo->ipi_hashbase[INP_PCBHASH(faddr.s_addr, lport, fport, pcbinfo->ipi_hashmask)]; LIST_FOREACH(inp, head, inp_hash) { #ifdef INET6 /* XXX inp locking */ if ((inp->inp_vflag & INP_IPV4) == 0) continue; #endif if (inp->inp_faddr.s_addr == faddr.s_addr && inp->inp_laddr.s_addr == laddr.s_addr && inp->inp_fport == fport && inp->inp_lport == lport) { /* * XXX We should be able to directly return * the inp here, without any checks. * Well unless both bound with SO_REUSEPORT? */ if (prison_flag(inp->inp_cred, PR_IP4)) return (inp); if (tmpinp == NULL) tmpinp = inp; } } if (tmpinp != NULL) return (tmpinp); /* * Then look for a wildcard match, if requested. */ if ((lookupflags & INPLOOKUP_WILDCARD) != 0) { struct inpcb *local_wild = NULL, *local_exact = NULL; #ifdef INET6 struct inpcb *local_wild_mapped = NULL; #endif struct inpcb *jail_wild = NULL; int injail; /* * Order of socket selection - we always prefer jails. * 1. jailed, non-wild. * 2. jailed, wild. * 3. non-jailed, non-wild. * 4. non-jailed, wild. */ head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport, 0, pcbinfo->ipi_hashmask)]; LIST_FOREACH(inp, head, inp_hash) { #ifdef INET6 /* XXX inp locking */ if ((inp->inp_vflag & INP_IPV4) == 0) continue; #endif if (inp->inp_faddr.s_addr != INADDR_ANY || inp->inp_lport != lport) continue; /* XXX inp locking */ if (ifp && ifp->if_type == IFT_FAITH && (inp->inp_flags & INP_FAITH) == 0) continue; injail = prison_flag(inp->inp_cred, PR_IP4); if (injail) { if (prison_check_ip4(inp->inp_cred, &laddr) != 0) continue; } else { if (local_exact != NULL) continue; } if (inp->inp_laddr.s_addr == laddr.s_addr) { if (injail) return (inp); else local_exact = inp; } else if (inp->inp_laddr.s_addr == INADDR_ANY) { #ifdef INET6 /* XXX inp locking, NULL check */ if (inp->inp_vflag & INP_IPV6PROTO) local_wild_mapped = inp; else #endif /* INET6 */ if (injail) jail_wild = inp; else local_wild = inp; } } /* LIST_FOREACH */ if (jail_wild != NULL) return (jail_wild); if (local_exact != NULL) return (local_exact); if (local_wild != NULL) return (local_wild); #ifdef INET6 if (local_wild_mapped != NULL) return (local_wild_mapped); #endif /* defined(INET6) */ } /* if ((lookupflags & INPLOOKUP_WILDCARD) != 0) */ return (NULL); } + +/* + * Lookup PCB in hash list, using pcbinfo tables. This variation locks the + * hash list lock, and will return the inpcb locked (i.e., requires + * INPLOOKUP_LOCKPCB). + */ +static struct inpcb * +in_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in_addr faddr, + u_int fport, struct in_addr laddr, u_int lport, int lookupflags, + struct ifnet *ifp) +{ + struct inpcb *inp; + + INP_HASH_RLOCK(pcbinfo); + inp = in_pcblookup_hash_locked(pcbinfo, faddr, fport, laddr, lport, + (lookupflags & ~(INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)), ifp); + if (inp != NULL) { + in_pcbref(inp); + INP_HASH_RUNLOCK(pcbinfo); + if (lookupflags & INPLOOKUP_WLOCKPCB) { + INP_WLOCK(inp); + if (in_pcbrele_wlocked(inp)) + return (NULL); + } else if (lookupflags & INPLOOKUP_RLOCKPCB) { + INP_RLOCK(inp); + if (in_pcbrele_rlocked(inp)) + return (NULL); + } else + panic("%s: locking bug", __func__); + } else + INP_HASH_RUNLOCK(pcbinfo); + return (inp); +} + +/* + * Public inpcb lookup routines, accepting a 4-tuple. + */ +struct inpcb * +in_pcblookup(struct inpcbinfo *pcbinfo, struct in_addr faddr, u_int fport, + struct in_addr laddr, u_int lport, int lookupflags, struct ifnet *ifp) +{ + + KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0, + ("%s: invalid lookup flags %d", __func__, lookupflags)); + KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0, + ("%s: LOCKPCB not set", __func__)); + + return (in_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport, + lookupflags, ifp)); +} #endif /* INET */ /* * Insert PCB onto various hash lists. */ int in_pcbinshash(struct inpcb *inp) { struct inpcbhead *pcbhash; struct inpcbporthead *pcbporthash; struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; struct inpcbport *phd; u_int32_t hashkey_faddr; - INP_INFO_WLOCK_ASSERT(pcbinfo); INP_WLOCK_ASSERT(inp); + INP_HASH_WLOCK_ASSERT(pcbinfo); + KASSERT((inp->inp_flags & INP_INHASHLIST) == 0, ("in_pcbinshash: INP_INHASHLIST")); #ifdef INET6 if (inp->inp_vflag & INP_IPV6) hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */; else #endif /* INET6 */ hashkey_faddr = inp->inp_faddr.s_addr; pcbhash = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr, inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)]; pcbporthash = &pcbinfo->ipi_porthashbase[ INP_PCBPORTHASH(inp->inp_lport, pcbinfo->ipi_porthashmask)]; /* * Go through port list and look for a head for this lport. */ LIST_FOREACH(phd, pcbporthash, phd_hash) { if (phd->phd_port == inp->inp_lport) break; } /* * If none exists, malloc one and tack it on. */ if (phd == NULL) { phd = malloc(sizeof(struct inpcbport), M_PCB, M_NOWAIT); if (phd == NULL) { return (ENOBUFS); /* XXX */ } phd->phd_port = inp->inp_lport; LIST_INIT(&phd->phd_pcblist); LIST_INSERT_HEAD(pcbporthash, phd, phd_hash); } inp->inp_phd = phd; LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist); LIST_INSERT_HEAD(pcbhash, inp, inp_hash); inp->inp_flags |= INP_INHASHLIST; return (0); } /* * Move PCB to the proper hash bucket when { faddr, fport } have been * changed. NOTE: This does not handle the case of the lport changing (the * hashed port list would have to be updated as well), so the lport must * not change after in_pcbinshash() has been called. */ void in_pcbrehash(struct inpcb *inp) { struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; struct inpcbhead *head; u_int32_t hashkey_faddr; - INP_INFO_WLOCK_ASSERT(pcbinfo); INP_WLOCK_ASSERT(inp); + INP_HASH_WLOCK_ASSERT(pcbinfo); + KASSERT(inp->inp_flags & INP_INHASHLIST, ("in_pcbrehash: !INP_INHASHLIST")); #ifdef INET6 if (inp->inp_vflag & INP_IPV6) hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */; else #endif /* INET6 */ hashkey_faddr = inp->inp_faddr.s_addr; head = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr, inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)]; LIST_REMOVE(inp, inp_hash); LIST_INSERT_HEAD(head, inp, inp_hash); } /* * Remove PCB from various lists. */ static void in_pcbremlists(struct inpcb *inp) { struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; INP_INFO_WLOCK_ASSERT(pcbinfo); INP_WLOCK_ASSERT(inp); inp->inp_gencnt = ++pcbinfo->ipi_gencnt; if (inp->inp_flags & INP_INHASHLIST) { struct inpcbport *phd = inp->inp_phd; + INP_HASH_WLOCK(pcbinfo); LIST_REMOVE(inp, inp_hash); LIST_REMOVE(inp, inp_portlist); if (LIST_FIRST(&phd->phd_pcblist) == NULL) { LIST_REMOVE(phd, phd_hash); free(phd, M_PCB); } + INP_HASH_WUNLOCK(pcbinfo); inp->inp_flags &= ~INP_INHASHLIST; } LIST_REMOVE(inp, inp_list); pcbinfo->ipi_count--; } /* * A set label operation has occurred at the socket layer, propagate the * label change into the in_pcb for the socket. */ void in_pcbsosetlabel(struct socket *so) { #ifdef MAC struct inpcb *inp; inp = sotoinpcb(so); KASSERT(inp != NULL, ("in_pcbsosetlabel: so->so_pcb == NULL")); INP_WLOCK(inp); SOCK_LOCK(so); mac_inpcb_sosetlabel(so, inp); SOCK_UNLOCK(so); INP_WUNLOCK(inp); #endif } /* * ipport_tick runs once per second, determining if random port allocation * should be continued. If more than ipport_randomcps ports have been * allocated in the last second, then we return to sequential port * allocation. We return to random allocation only once we drop below * ipport_randomcps for at least ipport_randomtime seconds. */ static void ipport_tick(void *xtp) { VNET_ITERATOR_DECL(vnet_iter); VNET_LIST_RLOCK_NOSLEEP(); VNET_FOREACH(vnet_iter) { CURVNET_SET(vnet_iter); /* XXX appease INVARIANTS here */ if (V_ipport_tcpallocs <= V_ipport_tcplastcount + V_ipport_randomcps) { if (V_ipport_stoprandom > 0) V_ipport_stoprandom--; } else V_ipport_stoprandom = V_ipport_randomtime; V_ipport_tcplastcount = V_ipport_tcpallocs; CURVNET_RESTORE(); } VNET_LIST_RUNLOCK_NOSLEEP(); callout_reset(&ipport_tick_callout, hz, ipport_tick, NULL); } static void ip_fini(void *xtp) { callout_stop(&ipport_tick_callout); } /* * The ipport_callout should start running at about the time we attach the * inet or inet6 domains. */ static void ipport_tick_init(const void *unused __unused) { /* Start ipport_tick. */ callout_init(&ipport_tick_callout, CALLOUT_MPSAFE); callout_reset(&ipport_tick_callout, 1, ipport_tick, NULL); EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL, SHUTDOWN_PRI_DEFAULT); } SYSINIT(ipport_tick_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ipport_tick_init, NULL); void inp_wlock(struct inpcb *inp) { INP_WLOCK(inp); } void inp_wunlock(struct inpcb *inp) { INP_WUNLOCK(inp); } void inp_rlock(struct inpcb *inp) { INP_RLOCK(inp); } void inp_runlock(struct inpcb *inp) { INP_RUNLOCK(inp); } #ifdef INVARIANTS void inp_lock_assert(struct inpcb *inp) { INP_WLOCK_ASSERT(inp); } void inp_unlock_assert(struct inpcb *inp) { INP_UNLOCK_ASSERT(inp); } #endif void inp_apply_all(void (*func)(struct inpcb *, void *), void *arg) { struct inpcb *inp; INP_INFO_RLOCK(&V_tcbinfo); LIST_FOREACH(inp, V_tcbinfo.ipi_listhead, inp_list) { INP_WLOCK(inp); func(inp, arg); INP_WUNLOCK(inp); } INP_INFO_RUNLOCK(&V_tcbinfo); } struct socket * inp_inpcbtosocket(struct inpcb *inp) { INP_WLOCK_ASSERT(inp); return (inp->inp_socket); } struct tcpcb * inp_inpcbtotcpcb(struct inpcb *inp) { INP_WLOCK_ASSERT(inp); return ((struct tcpcb *)inp->inp_ppcb); } int inp_ip_tos_get(const struct inpcb *inp) { return (inp->inp_ip_tos); } void inp_ip_tos_set(struct inpcb *inp, int val) { inp->inp_ip_tos = val; } void inp_4tuple_get(struct inpcb *inp, uint32_t *laddr, uint16_t *lp, uint32_t *faddr, uint16_t *fp) { INP_LOCK_ASSERT(inp); *laddr = inp->inp_laddr.s_addr; *faddr = inp->inp_faddr.s_addr; *lp = inp->inp_lport; *fp = inp->inp_fport; } struct inpcb * so_sotoinpcb(struct socket *so) { return (sotoinpcb(so)); } struct tcpcb * so_sototcpcb(struct socket *so) { return (sototcpcb(so)); } #ifdef DDB static void db_print_indent(int indent) { int i; for (i = 0; i < indent; i++) db_printf(" "); } static void db_print_inconninfo(struct in_conninfo *inc, const char *name, int indent) { char faddr_str[48], laddr_str[48]; db_print_indent(indent); db_printf("%s at %p\n", name, inc); indent += 2; #ifdef INET6 if (inc->inc_flags & INC_ISIPV6) { /* IPv6. */ ip6_sprintf(laddr_str, &inc->inc6_laddr); ip6_sprintf(faddr_str, &inc->inc6_faddr); } else { #endif /* IPv4. */ inet_ntoa_r(inc->inc_laddr, laddr_str); inet_ntoa_r(inc->inc_faddr, faddr_str); #ifdef INET6 } #endif db_print_indent(indent); db_printf("inc_laddr %s inc_lport %u\n", laddr_str, ntohs(inc->inc_lport)); db_print_indent(indent); db_printf("inc_faddr %s inc_fport %u\n", faddr_str, ntohs(inc->inc_fport)); } static void db_print_inpflags(int inp_flags) { int comma; comma = 0; if (inp_flags & INP_RECVOPTS) { db_printf("%sINP_RECVOPTS", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_RECVRETOPTS) { db_printf("%sINP_RECVRETOPTS", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_RECVDSTADDR) { db_printf("%sINP_RECVDSTADDR", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_HDRINCL) { db_printf("%sINP_HDRINCL", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_HIGHPORT) { db_printf("%sINP_HIGHPORT", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_LOWPORT) { db_printf("%sINP_LOWPORT", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_ANONPORT) { db_printf("%sINP_ANONPORT", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_RECVIF) { db_printf("%sINP_RECVIF", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_MTUDISC) { db_printf("%sINP_MTUDISC", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_FAITH) { db_printf("%sINP_FAITH", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_RECVTTL) { db_printf("%sINP_RECVTTL", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_DONTFRAG) { db_printf("%sINP_DONTFRAG", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_IPV6_V6ONLY) { db_printf("%sIN6P_IPV6_V6ONLY", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_PKTINFO) { db_printf("%sIN6P_PKTINFO", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_HOPLIMIT) { db_printf("%sIN6P_HOPLIMIT", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_HOPOPTS) { db_printf("%sIN6P_HOPOPTS", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_DSTOPTS) { db_printf("%sIN6P_DSTOPTS", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_RTHDR) { db_printf("%sIN6P_RTHDR", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_RTHDRDSTOPTS) { db_printf("%sIN6P_RTHDRDSTOPTS", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_TCLASS) { db_printf("%sIN6P_TCLASS", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_AUTOFLOWLABEL) { db_printf("%sIN6P_AUTOFLOWLABEL", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_TIMEWAIT) { db_printf("%sINP_TIMEWAIT", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_ONESBCAST) { db_printf("%sINP_ONESBCAST", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_DROPPED) { db_printf("%sINP_DROPPED", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_SOCKREF) { db_printf("%sINP_SOCKREF", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_RFC2292) { db_printf("%sIN6P_RFC2292", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_MTU) { db_printf("IN6P_MTU%s", comma ? ", " : ""); comma = 1; } } static void db_print_inpvflag(u_char inp_vflag) { int comma; comma = 0; if (inp_vflag & INP_IPV4) { db_printf("%sINP_IPV4", comma ? ", " : ""); comma = 1; } if (inp_vflag & INP_IPV6) { db_printf("%sINP_IPV6", comma ? ", " : ""); comma = 1; } if (inp_vflag & INP_IPV6PROTO) { db_printf("%sINP_IPV6PROTO", comma ? ", " : ""); comma = 1; } } static void db_print_inpcb(struct inpcb *inp, const char *name, int indent) { db_print_indent(indent); db_printf("%s at %p\n", name, inp); indent += 2; db_print_indent(indent); db_printf("inp_flow: 0x%x\n", inp->inp_flow); db_print_inconninfo(&inp->inp_inc, "inp_conninfo", indent); db_print_indent(indent); db_printf("inp_ppcb: %p inp_pcbinfo: %p inp_socket: %p\n", inp->inp_ppcb, inp->inp_pcbinfo, inp->inp_socket); db_print_indent(indent); db_printf("inp_label: %p inp_flags: 0x%x (", inp->inp_label, inp->inp_flags); db_print_inpflags(inp->inp_flags); db_printf(")\n"); db_print_indent(indent); db_printf("inp_sp: %p inp_vflag: 0x%x (", inp->inp_sp, inp->inp_vflag); db_print_inpvflag(inp->inp_vflag); db_printf(")\n"); db_print_indent(indent); db_printf("inp_ip_ttl: %d inp_ip_p: %d inp_ip_minttl: %d\n", inp->inp_ip_ttl, inp->inp_ip_p, inp->inp_ip_minttl); db_print_indent(indent); #ifdef INET6 if (inp->inp_vflag & INP_IPV6) { db_printf("in6p_options: %p in6p_outputopts: %p " "in6p_moptions: %p\n", inp->in6p_options, inp->in6p_outputopts, inp->in6p_moptions); db_printf("in6p_icmp6filt: %p in6p_cksum %d " "in6p_hops %u\n", inp->in6p_icmp6filt, inp->in6p_cksum, inp->in6p_hops); } else #endif { db_printf("inp_ip_tos: %d inp_ip_options: %p " "inp_ip_moptions: %p\n", inp->inp_ip_tos, inp->inp_options, inp->inp_moptions); } db_print_indent(indent); db_printf("inp_phd: %p inp_gencnt: %ju\n", inp->inp_phd, (uintmax_t)inp->inp_gencnt); } DB_SHOW_COMMAND(inpcb, db_show_inpcb) { struct inpcb *inp; if (!have_addr) { db_printf("usage: show inpcb \n"); return; } inp = (struct inpcb *)addr; db_print_inpcb(inp, "inpcb", 0); } #endif Index: head/sys/netinet/in_pcb.h =================================================================== --- head/sys/netinet/in_pcb.h (revision 222487) +++ head/sys/netinet/in_pcb.h (revision 222488) @@ -1,547 +1,573 @@ /*- * Copyright (c) 1982, 1986, 1990, 1993 * The Regents of the University of California. * Copyright (c) 2010-2011 Juniper Networks, Inc. * All rights reserved. * * Portions of this software were developed by Robert N. M. Watson under * contract to Juniper Networks, Inc. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)in_pcb.h 8.1 (Berkeley) 6/10/93 * $FreeBSD$ */ #ifndef _NETINET_IN_PCB_H_ #define _NETINET_IN_PCB_H_ #include #include #include #include #ifdef _KERNEL #include #include #include #endif #define in6pcb inpcb /* for KAME src sync over BSD*'s */ #define in6p_sp inp_sp /* for KAME src sync over BSD*'s */ struct inpcbpolicy; /* * struct inpcb is the common protocol control block structure used in most * IP transport protocols. * * Pointers to local and foreign host table entries, local and foreign socket * numbers, and pointers up (to a socket structure) and down (to a * protocol-specific control block) are stored here. */ LIST_HEAD(inpcbhead, inpcb); LIST_HEAD(inpcbporthead, inpcbport); typedef u_quad_t inp_gen_t; /* * PCB with AF_INET6 null bind'ed laddr can receive AF_INET input packet. * So, AF_INET6 null laddr is also used as AF_INET null laddr, by utilizing * the following structure. */ struct in_addr_4in6 { u_int32_t ia46_pad32[3]; struct in_addr ia46_addr4; }; /* * NOTE: ipv6 addrs should be 64-bit aligned, per RFC 2553. in_conninfo has * some extra padding to accomplish this. */ struct in_endpoints { u_int16_t ie_fport; /* foreign port */ u_int16_t ie_lport; /* local port */ /* protocol dependent part, local and foreign addr */ union { /* foreign host table entry */ struct in_addr_4in6 ie46_foreign; struct in6_addr ie6_foreign; } ie_dependfaddr; union { /* local host table entry */ struct in_addr_4in6 ie46_local; struct in6_addr ie6_local; } ie_dependladdr; }; #define ie_faddr ie_dependfaddr.ie46_foreign.ia46_addr4 #define ie_laddr ie_dependladdr.ie46_local.ia46_addr4 #define ie6_faddr ie_dependfaddr.ie6_foreign #define ie6_laddr ie_dependladdr.ie6_local /* * XXX The defines for inc_* are hacks and should be changed to direct * references. */ struct in_conninfo { u_int8_t inc_flags; u_int8_t inc_len; u_int16_t inc_fibnum; /* XXX was pad, 16 bits is plenty */ /* protocol dependent part */ struct in_endpoints inc_ie; }; /* * Flags for inc_flags. */ #define INC_ISIPV6 0x01 #define inc_isipv6 inc_flags /* temp compatability */ #define inc_fport inc_ie.ie_fport #define inc_lport inc_ie.ie_lport #define inc_faddr inc_ie.ie_faddr #define inc_laddr inc_ie.ie_laddr #define inc6_faddr inc_ie.ie6_faddr #define inc6_laddr inc_ie.ie6_laddr struct icmp6_filter; /*- * struct inpcb captures the network layer state for TCP, UDP, and raw IPv4 * and IPv6 sockets. In the case of TCP, further per-connection state is * hung off of inp_ppcb most of the time. Almost all fields of struct inpcb * are static after creation or protected by a per-inpcb rwlock, inp_lock. A * few fields also require the global pcbinfo lock for the inpcb to be held, * when modified, such as the global connection lists and hashes, as well as * binding information (which affects which hash a connection is on). This * model means that connections can be looked up without holding the * per-connection lock, which is important for performance when attempting to * find the connection for a packet given its IP and port tuple. Writing to * these fields that write locks be held on both the inpcb and global locks. * * Key: * (c) - Constant after initialization * (i) - Protected by the inpcb lock * (p) - Protected by the pcbinfo lock for the inpcb * (s) - Protected by another subsystem's locks * (x) - Undefined locking * * A few other notes: * * When a read lock is held, stability of the field is guaranteed; to write * to a field, a write lock must generally be held. * * netinet/netinet6-layer code should not assume that the inp_socket pointer * is safe to dereference without inp_lock being held, even for protocols * other than TCP (where the inpcb persists during TIMEWAIT even after the * socket has been freed), or there may be close(2)-related races. * * The inp_vflag field is overloaded, and would otherwise ideally be (c). */ struct inpcb { LIST_ENTRY(inpcb) inp_hash; /* (i/p) hash list */ LIST_ENTRY(inpcb) inp_list; /* (i/p) list for all PCBs for proto */ void *inp_ppcb; /* (i) pointer to per-protocol pcb */ struct inpcbinfo *inp_pcbinfo; /* (c) PCB list info */ struct socket *inp_socket; /* (i) back pointer to socket */ struct ucred *inp_cred; /* (c) cache of socket cred */ u_int32_t inp_flow; /* (i) IPv6 flow information */ int inp_flags; /* (i) generic IP/datagram flags */ int inp_flags2; /* (i) generic IP/datagram flags #2*/ u_char inp_vflag; /* (i) IP version flag (v4/v6) */ u_char inp_ip_ttl; /* (i) time to live proto */ u_char inp_ip_p; /* (c) protocol proto */ u_char inp_ip_minttl; /* (i) minimum TTL or drop */ uint32_t inp_flowid; /* (x) flow id / queue id */ u_int inp_refcount; /* (i) refcount */ void *inp_pspare[4]; /* (x) rtentry / general use */ u_int inp_ispare[4]; /* general use */ /* Local and foreign ports, local and foreign addr. */ struct in_conninfo inp_inc; /* (i/p) list for PCB's local port */ /* MAC and IPSEC policy information. */ struct label *inp_label; /* (i) MAC label */ struct inpcbpolicy *inp_sp; /* (s) for IPSEC */ /* Protocol-dependent part; options. */ struct { u_char inp4_ip_tos; /* (i) type of service proto */ struct mbuf *inp4_options; /* (i) IP options */ struct ip_moptions *inp4_moptions; /* (i) IP mcast options */ } inp_depend4; struct { /* (i) IP options */ struct mbuf *inp6_options; /* (i) IP6 options for outgoing packets */ struct ip6_pktopts *inp6_outputopts; /* (i) IP multicast options */ struct ip6_moptions *inp6_moptions; /* (i) ICMPv6 code type filter */ struct icmp6_filter *inp6_icmp6filt; /* (i) IPV6_CHECKSUM setsockopt */ int inp6_cksum; short inp6_hops; } inp_depend6; LIST_ENTRY(inpcb) inp_portlist; /* (i/p) */ struct inpcbport *inp_phd; /* (i/p) head of this list */ #define inp_zero_size offsetof(struct inpcb, inp_gencnt) inp_gen_t inp_gencnt; /* (c) generation count */ struct llentry *inp_lle; /* cached L2 information */ struct rtentry *inp_rt; /* cached L3 information */ struct rwlock inp_lock; }; #define inp_fport inp_inc.inc_fport #define inp_lport inp_inc.inc_lport #define inp_faddr inp_inc.inc_faddr #define inp_laddr inp_inc.inc_laddr #define inp_ip_tos inp_depend4.inp4_ip_tos #define inp_options inp_depend4.inp4_options #define inp_moptions inp_depend4.inp4_moptions #define in6p_faddr inp_inc.inc6_faddr #define in6p_laddr inp_inc.inc6_laddr #define in6p_hops inp_depend6.inp6_hops /* default hop limit */ #define in6p_flowinfo inp_flow #define in6p_options inp_depend6.inp6_options #define in6p_outputopts inp_depend6.inp6_outputopts #define in6p_moptions inp_depend6.inp6_moptions #define in6p_icmp6filt inp_depend6.inp6_icmp6filt #define in6p_cksum inp_depend6.inp6_cksum #define inp_vnet inp_pcbinfo->ipi_vnet /* * The range of the generation count, as used in this implementation, is 9e19. * We would have to create 300 billion connections per second for this number * to roll over in a year. This seems sufficiently unlikely that we simply * don't concern ourselves with that possibility. */ /* * Interface exported to userland by various protocols which use inpcbs. Hack * alert -- only define if struct xsocket is in scope. */ #ifdef _SYS_SOCKETVAR_H_ struct xinpcb { size_t xi_len; /* length of this structure */ struct inpcb xi_inp; struct xsocket xi_socket; u_quad_t xi_alignment_hack; }; struct xinpgen { size_t xig_len; /* length of this structure */ u_int xig_count; /* number of PCBs at this time */ inp_gen_t xig_gen; /* generation count at this time */ so_gen_t xig_sogen; /* socket generation count at this time */ }; #endif /* _SYS_SOCKETVAR_H_ */ struct inpcbport { LIST_ENTRY(inpcbport) phd_hash; struct inpcbhead phd_pcblist; u_short phd_port; }; /*- * Global data structure for each high-level protocol (UDP, TCP, ...) in both * IPv4 and IPv6. Holds inpcb lists and information for managing them. * - * Each pcbinfo is protected by ipi_lock, covering mutable global fields (such - * as the global pcb list) and hashed lookup tables. The lock order is: + * Each pcbinfo is protected by two locks: ipi_lock and ipi_hash_lock, + * the former covering mutable global fields (such as the global pcb list), + * and the latter covering the hashed lookup tables. The lock order is: * - * ipi_lock (before) inpcb locks + * ipi_lock (before) inpcb locks (before) ipi_hash_lock * * Locking key: * * (c) Constant or nearly constant after initialisation * (g) Locked by ipi_lock - * (h) Read using either ipi_lock or inpcb lock; write requires both. + * (h) Read using either ipi_hash_lock or inpcb lock; write requires both. * (x) Synchronisation properties poorly defined */ struct inpcbinfo { /* - * Global lock protecting global inpcb list, inpcb count, hash tables, - * etc. + * Global lock protecting global inpcb list, inpcb count, etc. */ struct rwlock ipi_lock; /* * Global list of inpcbs on the protocol. */ struct inpcbhead *ipi_listhead; /* (g) */ u_int ipi_count; /* (g) */ /* * Generation count -- incremented each time a connection is allocated * or freed. */ u_quad_t ipi_gencnt; /* (g) */ /* * Fields associated with port lookup and allocation. */ u_short ipi_lastport; /* (x) */ u_short ipi_lastlow; /* (x) */ u_short ipi_lasthi; /* (x) */ /* * UMA zone from which inpcbs are allocated for this protocol. */ struct uma_zone *ipi_zone; /* (c) */ /* + * Global lock protecting hash lookup tables. + */ + struct rwlock ipi_hash_lock; + + /* * Global hash of inpcbs, hashed by local and foreign addresses and * port numbers. */ - struct inpcbhead *ipi_hashbase; /* (g) */ - u_long ipi_hashmask; /* (g) */ + struct inpcbhead *ipi_hashbase; /* (h) */ + u_long ipi_hashmask; /* (h) */ /* * Global hash of inpcbs, hashed by only local port number. */ - struct inpcbporthead *ipi_porthashbase; /* (g) */ - u_long ipi_porthashmask; /* (g) */ + struct inpcbporthead *ipi_porthashbase; /* (h) */ + u_long ipi_porthashmask; /* (h) */ /* * Pointer to network stack instance */ struct vnet *ipi_vnet; /* (c) */ /* * general use 2 */ void *ipi_pspare[2]; }; #define INP_LOCK_INIT(inp, d, t) \ rw_init_flags(&(inp)->inp_lock, (t), RW_RECURSE | RW_DUPOK) #define INP_LOCK_DESTROY(inp) rw_destroy(&(inp)->inp_lock) #define INP_RLOCK(inp) rw_rlock(&(inp)->inp_lock) #define INP_WLOCK(inp) rw_wlock(&(inp)->inp_lock) #define INP_TRY_RLOCK(inp) rw_try_rlock(&(inp)->inp_lock) #define INP_TRY_WLOCK(inp) rw_try_wlock(&(inp)->inp_lock) #define INP_RUNLOCK(inp) rw_runlock(&(inp)->inp_lock) #define INP_WUNLOCK(inp) rw_wunlock(&(inp)->inp_lock) #define INP_TRY_UPGRADE(inp) rw_try_upgrade(&(inp)->inp_lock) #define INP_DOWNGRADE(inp) rw_downgrade(&(inp)->inp_lock) #define INP_WLOCKED(inp) rw_wowned(&(inp)->inp_lock) #define INP_LOCK_ASSERT(inp) rw_assert(&(inp)->inp_lock, RA_LOCKED) #define INP_RLOCK_ASSERT(inp) rw_assert(&(inp)->inp_lock, RA_RLOCKED) #define INP_WLOCK_ASSERT(inp) rw_assert(&(inp)->inp_lock, RA_WLOCKED) #define INP_UNLOCK_ASSERT(inp) rw_assert(&(inp)->inp_lock, RA_UNLOCKED) #ifdef _KERNEL /* * These locking functions are for inpcb consumers outside of sys/netinet, * more specifically, they were added for the benefit of TOE drivers. The * macros are reserved for use by the stack. */ void inp_wlock(struct inpcb *); void inp_wunlock(struct inpcb *); void inp_rlock(struct inpcb *); void inp_runlock(struct inpcb *); #ifdef INVARIANTS void inp_lock_assert(struct inpcb *); void inp_unlock_assert(struct inpcb *); #else static __inline void inp_lock_assert(struct inpcb *inp __unused) { } static __inline void inp_unlock_assert(struct inpcb *inp __unused) { } #endif void inp_apply_all(void (*func)(struct inpcb *, void *), void *arg); int inp_ip_tos_get(const struct inpcb *inp); void inp_ip_tos_set(struct inpcb *inp, int val); struct socket * inp_inpcbtosocket(struct inpcb *inp); struct tcpcb * inp_inpcbtotcpcb(struct inpcb *inp); void inp_4tuple_get(struct inpcb *inp, uint32_t *laddr, uint16_t *lp, uint32_t *faddr, uint16_t *fp); #endif /* _KERNEL */ #define INP_INFO_LOCK_INIT(ipi, d) \ rw_init_flags(&(ipi)->ipi_lock, (d), RW_RECURSE) #define INP_INFO_LOCK_DESTROY(ipi) rw_destroy(&(ipi)->ipi_lock) #define INP_INFO_RLOCK(ipi) rw_rlock(&(ipi)->ipi_lock) #define INP_INFO_WLOCK(ipi) rw_wlock(&(ipi)->ipi_lock) #define INP_INFO_TRY_RLOCK(ipi) rw_try_rlock(&(ipi)->ipi_lock) #define INP_INFO_TRY_WLOCK(ipi) rw_try_wlock(&(ipi)->ipi_lock) #define INP_INFO_TRY_UPGRADE(ipi) rw_try_upgrade(&(ipi)->ipi_lock) #define INP_INFO_RUNLOCK(ipi) rw_runlock(&(ipi)->ipi_lock) #define INP_INFO_WUNLOCK(ipi) rw_wunlock(&(ipi)->ipi_lock) #define INP_INFO_LOCK_ASSERT(ipi) rw_assert(&(ipi)->ipi_lock, RA_LOCKED) #define INP_INFO_RLOCK_ASSERT(ipi) rw_assert(&(ipi)->ipi_lock, RA_RLOCKED) #define INP_INFO_WLOCK_ASSERT(ipi) rw_assert(&(ipi)->ipi_lock, RA_WLOCKED) #define INP_INFO_UNLOCK_ASSERT(ipi) rw_assert(&(ipi)->ipi_lock, RA_UNLOCKED) +#define INP_HASH_LOCK_INIT(ipi, d) \ + rw_init_flags(&(ipi)->ipi_hash_lock, (d), 0) +#define INP_HASH_LOCK_DESTROY(ipi) rw_destroy(&(ipi)->ipi_hash_lock) +#define INP_HASH_RLOCK(ipi) rw_rlock(&(ipi)->ipi_hash_lock) +#define INP_HASH_WLOCK(ipi) rw_wlock(&(ipi)->ipi_hash_lock) +#define INP_HASH_RUNLOCK(ipi) rw_runlock(&(ipi)->ipi_hash_lock) +#define INP_HASH_WUNLOCK(ipi) rw_wunlock(&(ipi)->ipi_hash_lock) +#define INP_HASH_LOCK_ASSERT(ipi) rw_assert(&(ipi)->ipi_hash_lock, \ + RA_LOCKED) +#define INP_HASH_WLOCK_ASSERT(ipi) rw_assert(&(ipi)->ipi_hash_lock, \ + RA_WLOCKED) + #define INP_PCBHASH(faddr, lport, fport, mask) \ (((faddr) ^ ((faddr) >> 16) ^ ntohs((lport) ^ (fport))) & (mask)) #define INP_PCBPORTHASH(lport, mask) \ (ntohs((lport)) & (mask)) /* * Flags for inp_vflags -- historically version flags only */ #define INP_IPV4 0x1 #define INP_IPV6 0x2 #define INP_IPV6PROTO 0x4 /* opened under IPv6 protocol */ /* * Flags for inp_flags. */ #define INP_RECVOPTS 0x00000001 /* receive incoming IP options */ #define INP_RECVRETOPTS 0x00000002 /* receive IP options for reply */ #define INP_RECVDSTADDR 0x00000004 /* receive IP dst address */ #define INP_HDRINCL 0x00000008 /* user supplies entire IP header */ #define INP_HIGHPORT 0x00000010 /* user wants "high" port binding */ #define INP_LOWPORT 0x00000020 /* user wants "low" port binding */ #define INP_ANONPORT 0x00000040 /* port chosen for user */ #define INP_RECVIF 0x00000080 /* receive incoming interface */ #define INP_MTUDISC 0x00000100 /* user can do MTU discovery */ #define INP_FAITH 0x00000200 /* accept FAITH'ed connections */ #define INP_RECVTTL 0x00000400 /* receive incoming IP TTL */ #define INP_DONTFRAG 0x00000800 /* don't fragment packet */ #define INP_BINDANY 0x00001000 /* allow bind to any address */ #define INP_INHASHLIST 0x00002000 /* in_pcbinshash() has been called */ #define IN6P_IPV6_V6ONLY 0x00008000 /* restrict AF_INET6 socket for v6 */ #define IN6P_PKTINFO 0x00010000 /* receive IP6 dst and I/F */ #define IN6P_HOPLIMIT 0x00020000 /* receive hoplimit */ #define IN6P_HOPOPTS 0x00040000 /* receive hop-by-hop options */ #define IN6P_DSTOPTS 0x00080000 /* receive dst options after rthdr */ #define IN6P_RTHDR 0x00100000 /* receive routing header */ #define IN6P_RTHDRDSTOPTS 0x00200000 /* receive dstoptions before rthdr */ #define IN6P_TCLASS 0x00400000 /* receive traffic class value */ #define IN6P_AUTOFLOWLABEL 0x00800000 /* attach flowlabel automatically */ #define INP_TIMEWAIT 0x01000000 /* in TIMEWAIT, ppcb is tcptw */ #define INP_ONESBCAST 0x02000000 /* send all-ones broadcast */ #define INP_DROPPED 0x04000000 /* protocol drop flag */ #define INP_SOCKREF 0x08000000 /* strong socket reference */ #define INP_SW_FLOWID 0x10000000 /* software generated flow id */ #define INP_HW_FLOWID 0x20000000 /* hardware generated flow id */ #define IN6P_RFC2292 0x40000000 /* used RFC2292 API on the socket */ #define IN6P_MTU 0x80000000 /* receive path MTU */ #define INP_CONTROLOPTS (INP_RECVOPTS|INP_RECVRETOPTS|INP_RECVDSTADDR|\ INP_RECVIF|INP_RECVTTL|\ IN6P_PKTINFO|IN6P_HOPLIMIT|IN6P_HOPOPTS|\ IN6P_DSTOPTS|IN6P_RTHDR|IN6P_RTHDRDSTOPTS|\ IN6P_TCLASS|IN6P_AUTOFLOWLABEL|IN6P_RFC2292|\ IN6P_MTU) /* * Flags for inp_flags2. */ #define INP_LLE_VALID 0x00000001 /* cached lle is valid */ #define INP_RT_VALID 0x00000002 /* cached rtentry is valid */ -#define INPLOOKUP_WILDCARD 1 +/* + * Flags passed to in_pcblookup*() functions. + */ +#define INPLOOKUP_WILDCARD 0x00000001 /* Allow wildcard sockets. */ +#define INPLOOKUP_RLOCKPCB 0x00000002 /* Return inpcb read-locked. */ +#define INPLOOKUP_WLOCKPCB 0x00000004 /* Return inpcb write-locked. */ + +#define INPLOOKUP_MASK (INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB | \ + INPLOOKUP_WLOCKPCB) + #define sotoinpcb(so) ((struct inpcb *)(so)->so_pcb) #define sotoin6pcb(so) sotoinpcb(so) /* for KAME src sync over BSD*'s */ #define INP_SOCKAF(so) so->so_proto->pr_domain->dom_family #define INP_CHECK_SOCKAF(so, af) (INP_SOCKAF(so) == af) #ifdef _KERNEL VNET_DECLARE(int, ipport_reservedhigh); VNET_DECLARE(int, ipport_reservedlow); VNET_DECLARE(int, ipport_lowfirstauto); VNET_DECLARE(int, ipport_lowlastauto); VNET_DECLARE(int, ipport_firstauto); VNET_DECLARE(int, ipport_lastauto); VNET_DECLARE(int, ipport_hifirstauto); VNET_DECLARE(int, ipport_hilastauto); VNET_DECLARE(int, ipport_randomized); VNET_DECLARE(int, ipport_randomcps); VNET_DECLARE(int, ipport_randomtime); VNET_DECLARE(int, ipport_stoprandom); VNET_DECLARE(int, ipport_tcpallocs); #define V_ipport_reservedhigh VNET(ipport_reservedhigh) #define V_ipport_reservedlow VNET(ipport_reservedlow) #define V_ipport_lowfirstauto VNET(ipport_lowfirstauto) #define V_ipport_lowlastauto VNET(ipport_lowlastauto) #define V_ipport_firstauto VNET(ipport_firstauto) #define V_ipport_lastauto VNET(ipport_lastauto) #define V_ipport_hifirstauto VNET(ipport_hifirstauto) #define V_ipport_hilastauto VNET(ipport_hilastauto) #define V_ipport_randomized VNET(ipport_randomized) #define V_ipport_randomcps VNET(ipport_randomcps) #define V_ipport_randomtime VNET(ipport_randomtime) #define V_ipport_stoprandom VNET(ipport_stoprandom) #define V_ipport_tcpallocs VNET(ipport_tcpallocs) void in_pcbinfo_destroy(struct inpcbinfo *); void in_pcbinfo_init(struct inpcbinfo *, const char *, struct inpcbhead *, int, int, char *, uma_init, uma_fini, uint32_t); void in_pcbpurgeif0(struct inpcbinfo *, struct ifnet *); int in_pcballoc(struct socket *, struct inpcbinfo *); int in_pcbbind(struct inpcb *, struct sockaddr *, struct ucred *); int in_pcb_lport(struct inpcb *, struct in_addr *, u_short *, struct ucred *, int); int in_pcbbind_setup(struct inpcb *, struct sockaddr *, in_addr_t *, u_short *, struct ucred *); int in_pcbconnect(struct inpcb *, struct sockaddr *, struct ucred *); int in_pcbconnect_setup(struct inpcb *, struct sockaddr *, in_addr_t *, u_short *, in_addr_t *, u_short *, struct inpcb **, struct ucred *); void in_pcbdetach(struct inpcb *); void in_pcbdisconnect(struct inpcb *); void in_pcbdrop(struct inpcb *); void in_pcbfree(struct inpcb *); int in_pcbinshash(struct inpcb *); struct inpcb * in_pcblookup_local(struct inpcbinfo *, struct in_addr, u_short, int, struct ucred *); struct inpcb * - in_pcblookup_hash(struct inpcbinfo *, struct in_addr, u_int, + in_pcblookup(struct inpcbinfo *, struct in_addr, u_int, struct in_addr, u_int, int, struct ifnet *); void in_pcbnotifyall(struct inpcbinfo *pcbinfo, struct in_addr, int, struct inpcb *(*)(struct inpcb *, int)); void in_pcbref(struct inpcb *); void in_pcbrehash(struct inpcb *); int in_pcbrele(struct inpcb *); int in_pcbrele_rlocked(struct inpcb *); int in_pcbrele_wlocked(struct inpcb *); void in_pcbsetsolabel(struct socket *so); int in_getpeeraddr(struct socket *so, struct sockaddr **nam); int in_getsockaddr(struct socket *so, struct sockaddr **nam); struct sockaddr * in_sockaddr(in_port_t port, struct in_addr *addr); void in_pcbsosetlabel(struct socket *so); #endif /* _KERNEL */ #endif /* !_NETINET_IN_PCB_H_ */ Index: head/sys/netinet/ip_divert.c =================================================================== --- head/sys/netinet/ip_divert.c (revision 222487) +++ head/sys/netinet/ip_divert.c (revision 222488) @@ -1,794 +1,794 @@ /*- * Copyright (c) 1982, 1986, 1988, 1993 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #if !defined(KLD_MODULE) #include "opt_inet.h" #include "opt_sctp.h" #ifndef INET #error "IPDIVERT requires INET." #endif #endif #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef SCTP #include #endif #include /* * 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 or other packet filters, * see the divert(4) manpage for features. * Packets are selected by the packet filter and tagged with an * MTAG_IPFW_RULE tag carrying the 'divert port' number (as set by * the packet filter) and information on the matching filter rule for * subsequent reinjection. The divert_port is used to put the packet * on the corresponding divert socket, while the rule number is passed * up (at least partially) as the sin_port in the struct sockaddr. * * Packets written to the divert socket carry in sin_addr a * destination address, and in sin_port the number of the filter rule * after which to continue processing. * If the destination address is INADDR_ANY, the packet is treated as * as outgoing and sent to ip_output(); otherwise it is treated as * incoming and sent to ip_input(). * Further, sin_zero carries some information on the interface, * which can be used in the reinject -- see comments in the code. * * On reinjection, processing in ip_input() and ip_output() * will be exactly the same as for the original packet, except that * packet filter processing will start at the rule number after the one * written in the sin_port (ipfw does not allow a rule #0, so sin_port=0 * will apply the entire ruleset to the packet). */ /* Internal variables. */ static VNET_DEFINE(struct inpcbhead, divcb); static VNET_DEFINE(struct inpcbinfo, divcbinfo); #define V_divcb VNET(divcb) #define V_divcbinfo VNET(divcbinfo) static u_long div_sendspace = DIVSNDQ; /* XXX sysctl ? */ static u_long div_recvspace = DIVRCVQ; /* XXX sysctl ? */ static eventhandler_tag ip_divert_event_tag; /* * Initialize divert connection block queue. */ static void div_zone_change(void *tag) { uma_zone_set_max(V_divcbinfo.ipi_zone, maxsockets); } static int div_inpcb_init(void *mem, int size, int flags) { struct inpcb *inp = mem; INP_LOCK_INIT(inp, "inp", "divinp"); return (0); } static void div_inpcb_fini(void *mem, int size) { struct inpcb *inp = mem; INP_LOCK_DESTROY(inp); } static void div_init(void) { /* * XXX We don't use the hash list for divert IP, but it's easier to * allocate one-entry hash lists than it is to check all over the * place for hashbase == NULL. */ in_pcbinfo_init(&V_divcbinfo, "div", &V_divcb, 1, 1, "divcb", div_inpcb_init, div_inpcb_fini, UMA_ZONE_NOFREE); } static void div_destroy(void) { in_pcbinfo_destroy(&V_divcbinfo); } /* * IPPROTO_DIVERT is not in the real IP protocol number space; this * function should never be called. Just in case, drop any packets. */ static void div_input(struct mbuf *m, int off) { KMOD_IPSTAT_INC(ips_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 divert_packet(struct mbuf *m, int incoming) { struct ip *ip; struct inpcb *inp; struct socket *sa; u_int16_t nport; struct sockaddr_in divsrc; struct m_tag *mtag; mtag = m_tag_locate(m, MTAG_IPFW_RULE, 0, NULL); if (mtag == NULL) { m_freem(m); return; } /* Assure header */ if (m->m_len < sizeof(struct ip) && (m = m_pullup(m, sizeof(struct ip))) == 0) return; ip = mtod(m, struct ip *); /* Delayed checksums are currently not compatible with divert. */ if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { ip->ip_len = ntohs(ip->ip_len); in_delayed_cksum(m); m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; ip->ip_len = htons(ip->ip_len); } #ifdef SCTP if (m->m_pkthdr.csum_flags & CSUM_SCTP) { ip->ip_len = ntohs(ip->ip_len); sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2)); m->m_pkthdr.csum_flags &= ~CSUM_SCTP; ip->ip_len = htons(ip->ip_len); } #endif bzero(&divsrc, sizeof(divsrc)); divsrc.sin_len = sizeof(divsrc); divsrc.sin_family = AF_INET; /* record matching rule, in host format */ divsrc.sin_port = ((struct ipfw_rule_ref *)(mtag+1))->rulenum; /* * Record receive interface address, if any. * But only for incoming packets. */ if (incoming) { struct ifaddr *ifa; struct ifnet *ifp; /* Sanity check */ M_ASSERTPKTHDR(m); /* Find IP address for receive interface */ ifp = m->m_pkthdr.rcvif; if_addr_rlock(ifp); TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { if (ifa->ifa_addr->sa_family != AF_INET) continue; divsrc.sin_addr = ((struct sockaddr_in *) ifa->ifa_addr)->sin_addr; break; } if_addr_runlock(ifp); } /* * Record the incoming interface name whenever we have one. */ if (m->m_pkthdr.rcvif) { /* * Hide the actual interface name in there in the * sin_zero array. XXX This needs to be moved to a * different sockaddr type for divert, e.g. * sockaddr_div with multiple fields like * sockaddr_dl. Presently we have only 7 bytes * but that will do for now as most interfaces * are 4 or less + 2 or less bytes for unit. * There is probably a faster way of doing this, * possibly taking it from the sockaddr_dl on the iface. * This solves the problem of a P2P link and a LAN interface * having the same address, which can result in the wrong * interface being assigned to the packet when fed back * into the divert socket. Theoretically if the daemon saves * and re-uses the sockaddr_in as suggested in the man pages, * this iface name will come along for the ride. * (see div_output for the other half of this.) */ strlcpy(divsrc.sin_zero, m->m_pkthdr.rcvif->if_xname, sizeof(divsrc.sin_zero)); } /* Put packet on socket queue, if any */ sa = NULL; nport = htons((u_int16_t)(((struct ipfw_rule_ref *)(mtag+1))->info)); INP_INFO_RLOCK(&V_divcbinfo); LIST_FOREACH(inp, &V_divcb, inp_list) { /* XXX why does only one socket match? */ if (inp->inp_lport == nport) { INP_RLOCK(inp); sa = inp->inp_socket; SOCKBUF_LOCK(&sa->so_rcv); if (sbappendaddr_locked(&sa->so_rcv, (struct sockaddr *)&divsrc, m, (struct mbuf *)0) == 0) { SOCKBUF_UNLOCK(&sa->so_rcv); sa = NULL; /* force mbuf reclaim below */ } else sorwakeup_locked(sa); INP_RUNLOCK(inp); break; } } INP_INFO_RUNLOCK(&V_divcbinfo); if (sa == NULL) { m_freem(m); KMOD_IPSTAT_INC(ips_noproto); KMOD_IPSTAT_DEC(ips_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 div_output(struct socket *so, struct mbuf *m, struct sockaddr_in *sin, struct mbuf *control) { struct m_tag *mtag; struct ipfw_rule_ref *dt; int error = 0; struct mbuf *options; /* * An mbuf may hasn't come from userland, but we pretend * that it has. */ m->m_pkthdr.rcvif = NULL; m->m_nextpkt = NULL; M_SETFIB(m, so->so_fibnum); if (control) m_freem(control); /* XXX */ mtag = m_tag_locate(m, MTAG_IPFW_RULE, 0, NULL); if (mtag == NULL) { /* this should be normal */ mtag = m_tag_alloc(MTAG_IPFW_RULE, 0, sizeof(struct ipfw_rule_ref), M_NOWAIT | M_ZERO); if (mtag == NULL) { error = ENOBUFS; goto cantsend; } m_tag_prepend(m, mtag); } dt = (struct ipfw_rule_ref *)(mtag+1); /* Loopback avoidance and state recovery */ if (sin) { int i; /* set the starting point. We provide a non-zero slot, * but a non_matching chain_id to skip that info and use * the rulenum/rule_id. */ dt->slot = 1; /* dummy, chain_id is invalid */ dt->chain_id = 0; dt->rulenum = sin->sin_port+1; /* host format ? */ dt->rule_id = 0; /* * Find receive interface with the given name, stuffed * (if it exists) in the sin_zero[] field. * The name is user supplied data so don't trust its size * or that it is zero terminated. */ for (i = 0; i < sizeof(sin->sin_zero) && sin->sin_zero[i]; i++) ; if ( i > 0 && i < sizeof(sin->sin_zero)) m->m_pkthdr.rcvif = ifunit(sin->sin_zero); } /* Reinject packet into the system as incoming or outgoing */ if (!sin || sin->sin_addr.s_addr == 0) { struct ip *const ip = mtod(m, struct ip *); struct inpcb *inp; dt->info |= IPFW_IS_DIVERT | IPFW_INFO_OUT; inp = sotoinpcb(so); INP_RLOCK(inp); /* * Don't allow both user specified and setsockopt options, * and don't allow packet length sizes that will crash */ if (((ip->ip_hl != (sizeof (*ip) >> 2)) && inp->inp_options) || ((u_short)ntohs(ip->ip_len) > m->m_pkthdr.len)) { error = EINVAL; INP_RUNLOCK(inp); m_freem(m); } else { /* Convert fields to host order for ip_output() */ ip->ip_len = ntohs(ip->ip_len); ip->ip_off = ntohs(ip->ip_off); /* Send packet to output processing */ KMOD_IPSTAT_INC(ips_rawout); /* XXX */ #ifdef MAC mac_inpcb_create_mbuf(inp, m); #endif /* * Get ready to inject the packet into ip_output(). * Just in case socket options were specified on the * divert socket, we duplicate them. This is done * to avoid having to hold the PCB locks over the call * to ip_output(), as doing this results in a number of * lock ordering complexities. * * Note that we set the multicast options argument for * ip_output() to NULL since it should be invariant that * they are not present. */ KASSERT(inp->inp_moptions == NULL, ("multicast options set on a divert socket")); options = NULL; /* * XXXCSJP: It is unclear to me whether or not it makes * sense for divert sockets to have options. However, * for now we will duplicate them with the INP locks * held so we can use them in ip_output() without * requring a reference to the pcb. */ if (inp->inp_options != NULL) { options = m_dup(inp->inp_options, M_DONTWAIT); if (options == NULL) error = ENOBUFS; } INP_RUNLOCK(inp); if (error == ENOBUFS) { m_freem(m); return (error); } error = ip_output(m, options, NULL, ((so->so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0) | IP_ALLOWBROADCAST | IP_RAWOUTPUT, NULL, NULL); if (options != NULL) m_freem(options); } } else { dt->info |= IPFW_IS_DIVERT | IPFW_INFO_IN; 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; bzero(sin->sin_zero, sizeof(sin->sin_zero)); sin->sin_port = 0; ifa = ifa_ifwithaddr((struct sockaddr *) sin); if (ifa == NULL) { error = EADDRNOTAVAIL; goto cantsend; } m->m_pkthdr.rcvif = ifa->ifa_ifp; ifa_free(ifa); } #ifdef MAC mac_socket_create_mbuf(so, m); #endif /* Send packet to input processing via netisr */ netisr_queue_src(NETISR_IP, (uintptr_t)so, m); } return error; cantsend: m_freem(m); return error; } static int div_attach(struct socket *so, int proto, struct thread *td) { struct inpcb *inp; int error; inp = sotoinpcb(so); KASSERT(inp == NULL, ("div_attach: inp != NULL")); if (td != NULL) { error = priv_check(td, PRIV_NETINET_DIVERT); if (error) return (error); } error = soreserve(so, div_sendspace, div_recvspace); if (error) return error; INP_INFO_WLOCK(&V_divcbinfo); error = in_pcballoc(so, &V_divcbinfo); if (error) { INP_INFO_WUNLOCK(&V_divcbinfo); return error; } inp = (struct inpcb *)so->so_pcb; INP_INFO_WUNLOCK(&V_divcbinfo); inp->inp_ip_p = proto; inp->inp_vflag |= INP_IPV4; inp->inp_flags |= INP_HDRINCL; INP_WUNLOCK(inp); return 0; } static void div_detach(struct socket *so) { struct inpcb *inp; inp = sotoinpcb(so); KASSERT(inp != NULL, ("div_detach: inp == NULL")); INP_INFO_WLOCK(&V_divcbinfo); INP_WLOCK(inp); in_pcbdetach(inp); in_pcbfree(inp); INP_INFO_WUNLOCK(&V_divcbinfo); } static int div_bind(struct socket *so, struct sockaddr *nam, struct thread *td) { struct inpcb *inp; int error; inp = sotoinpcb(so); KASSERT(inp != NULL, ("div_bind: inp == NULL")); /* in_pcbbind assumes that nam is a sockaddr_in * 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_INET) return EAFNOSUPPORT; ((struct sockaddr_in *)nam)->sin_addr.s_addr = INADDR_ANY; INP_INFO_WLOCK(&V_divcbinfo); INP_WLOCK(inp); error = in_pcbbind(inp, nam, td->td_ucred); INP_WUNLOCK(inp); INP_INFO_WUNLOCK(&V_divcbinfo); return error; } static int div_shutdown(struct socket *so) { struct inpcb *inp; inp = sotoinpcb(so); KASSERT(inp != NULL, ("div_shutdown: inp == NULL")); INP_WLOCK(inp); socantsendmore(so); INP_WUNLOCK(inp); return 0; } static int div_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, struct mbuf *control, struct thread *td) { /* Packet must have a header (but that's about it) */ if (m->m_len < sizeof (struct ip) && (m = m_pullup(m, sizeof (struct ip))) == 0) { KMOD_IPSTAT_INC(ips_toosmall); m_freem(m); return EINVAL; } /* Send packet */ return div_output(so, m, (struct sockaddr_in *)nam, control); } static void div_ctlinput(int cmd, struct sockaddr *sa, void *vip) { struct in_addr faddr; faddr = ((struct sockaddr_in *)sa)->sin_addr; if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY) return; if (PRC_IS_REDIRECT(cmd)) return; } static int div_pcblist(SYSCTL_HANDLER_ARGS) { int error, i, n; struct inpcb *inp, **inp_list; inp_gen_t gencnt; struct xinpgen xig; /* * The process of preparing the TCB list is too time-consuming and * resource-intensive to repeat twice on every request. */ if (req->oldptr == 0) { n = V_divcbinfo.ipi_count; n += imax(n / 8, 10); req->oldidx = 2 * (sizeof xig) + n * sizeof(struct xinpcb); return 0; } if (req->newptr != 0) return EPERM; /* * OK, now we're committed to doing something. */ INP_INFO_RLOCK(&V_divcbinfo); gencnt = V_divcbinfo.ipi_gencnt; n = V_divcbinfo.ipi_count; INP_INFO_RUNLOCK(&V_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; inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK); if (inp_list == 0) return ENOMEM; INP_INFO_RLOCK(&V_divcbinfo); for (inp = LIST_FIRST(V_divcbinfo.ipi_listhead), i = 0; inp && i < n; inp = LIST_NEXT(inp, inp_list)) { INP_WLOCK(inp); if (inp->inp_gencnt <= gencnt && cr_canseeinpcb(req->td->td_ucred, inp) == 0) { in_pcbref(inp); inp_list[i++] = inp; } INP_WUNLOCK(inp); } INP_INFO_RUNLOCK(&V_divcbinfo); n = i; error = 0; for (i = 0; i < n; i++) { inp = inp_list[i]; INP_RLOCK(inp); if (inp->inp_gencnt <= gencnt) { struct xinpcb xi; bzero(&xi, sizeof(xi)); xi.xi_len = sizeof xi; /* XXX should avoid extra copy */ bcopy(inp, &xi.xi_inp, sizeof *inp); if (inp->inp_socket) sotoxsocket(inp->inp_socket, &xi.xi_socket); INP_RUNLOCK(inp); error = SYSCTL_OUT(req, &xi, sizeof xi); } else INP_RUNLOCK(inp); } INP_INFO_WLOCK(&V_divcbinfo); for (i = 0; i < n; i++) { inp = inp_list[i]; - INP_WLOCK(inp); - if (!in_pcbrele(inp)) - INP_WUNLOCK(inp); + INP_RLOCK(inp); + if (!in_pcbrele_rlocked(inp)) + INP_RUNLOCK(inp); } INP_INFO_WUNLOCK(&V_divcbinfo); if (!error) { /* * Give the user an updated idea of our state. * If the generation differs from what we told * her before, she knows that something happened * while we were processing this request, and it * might be necessary to retry. */ INP_INFO_RLOCK(&V_divcbinfo); xig.xig_gen = V_divcbinfo.ipi_gencnt; xig.xig_sogen = so_gencnt; xig.xig_count = V_divcbinfo.ipi_count; INP_INFO_RUNLOCK(&V_divcbinfo); error = SYSCTL_OUT(req, &xig, sizeof xig); } free(inp_list, M_TEMP); return error; } #ifdef SYSCTL_NODE SYSCTL_NODE(_net_inet, IPPROTO_DIVERT, divert, CTLFLAG_RW, 0, "IPDIVERT"); SYSCTL_PROC(_net_inet_divert, OID_AUTO, pcblist, CTLTYPE_OPAQUE | CTLFLAG_RD, NULL, 0, div_pcblist, "S,xinpcb", "List of active divert sockets"); #endif struct pr_usrreqs div_usrreqs = { .pru_attach = div_attach, .pru_bind = div_bind, .pru_control = in_control, .pru_detach = div_detach, .pru_peeraddr = in_getpeeraddr, .pru_send = div_send, .pru_shutdown = div_shutdown, .pru_sockaddr = in_getsockaddr, .pru_sosetlabel = in_pcbsosetlabel }; struct protosw div_protosw = { .pr_type = SOCK_RAW, .pr_protocol = IPPROTO_DIVERT, .pr_flags = PR_ATOMIC|PR_ADDR, .pr_input = div_input, .pr_ctlinput = div_ctlinput, .pr_ctloutput = ip_ctloutput, .pr_init = div_init, #ifdef VIMAGE .pr_destroy = div_destroy, #endif .pr_usrreqs = &div_usrreqs }; static int div_modevent(module_t mod, int type, void *unused) { int err = 0; #ifndef VIMAGE int n; #endif 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_INET, &div_protosw); if (err != 0) return (err); ip_divert_ptr = divert_packet; ip_divert_event_tag = EVENTHANDLER_REGISTER(maxsockets_change, div_zone_change, NULL, EVENTHANDLER_PRI_ANY); 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: #ifdef VIMAGE err = EPERM; break; #else /* * 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(&V_divcbinfo); n = V_divcbinfo.ipi_count; if (n != 0) { err = EBUSY; INP_INFO_WUNLOCK(&V_divcbinfo); break; } ip_divert_ptr = NULL; err = pf_proto_unregister(PF_INET, IPPROTO_DIVERT, SOCK_RAW); INP_INFO_WUNLOCK(&V_divcbinfo); div_destroy(); EVENTHANDLER_DEREGISTER(maxsockets_change, ip_divert_event_tag); break; #endif /* !VIMAGE */ default: err = EOPNOTSUPP; break; } return err; } static moduledata_t ipdivertmod = { "ipdivert", div_modevent, 0 }; DECLARE_MODULE(ipdivert, ipdivertmod, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY); MODULE_DEPEND(ipdivert, ipfw, 2, 2, 2); MODULE_VERSION(ipdivert, 1); Index: head/sys/netinet/ipfw/ip_fw2.c =================================================================== --- head/sys/netinet/ipfw/ip_fw2.c (revision 222487) +++ head/sys/netinet/ipfw/ip_fw2.c (revision 222488) @@ -1,2546 +1,2558 @@ /*- * Copyright (c) 2002-2009 Luigi Rizzo, Universita` di Pisa * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); /* * The FreeBSD IP packet firewall, main file */ #if !defined(KLD_MODULE) #include "opt_ipfw.h" #include "opt_ipdivert.h" #include "opt_ipdn.h" #include "opt_inet.h" #ifndef INET #error IPFIREWALL requires INET. #endif /* INET */ #endif #include "opt_inet6.h" #include "opt_ipsec.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include /* for ETHERTYPE_IP */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef INET6 #include #include #endif #include /* XXX for in_cksum */ #ifdef MAC #include #endif /* * static variables followed by global ones. * All ipfw global variables are here. */ /* ipfw_vnet_ready controls when we are open for business */ static VNET_DEFINE(int, ipfw_vnet_ready) = 0; #define V_ipfw_vnet_ready VNET(ipfw_vnet_ready) static VNET_DEFINE(int, fw_deny_unknown_exthdrs); #define V_fw_deny_unknown_exthdrs VNET(fw_deny_unknown_exthdrs) #ifdef IPFIREWALL_DEFAULT_TO_ACCEPT static int default_to_accept = 1; #else static int default_to_accept; #endif VNET_DEFINE(int, autoinc_step); VNET_DEFINE(int, fw_one_pass) = 1; /* * Each rule belongs to one of 32 different sets (0..31). * The variable set_disable contains one bit per set. * If the bit is set, all rules in the corresponding set * are disabled. Set RESVD_SET(31) is reserved for the default rule * and rules that are not deleted by the flush command, * and CANNOT be disabled. * Rules in set RESVD_SET can only be deleted individually. */ VNET_DEFINE(u_int32_t, set_disable); #define V_set_disable VNET(set_disable) VNET_DEFINE(int, fw_verbose); /* counter for ipfw_log(NULL...) */ VNET_DEFINE(u_int64_t, norule_counter); VNET_DEFINE(int, verbose_limit); /* layer3_chain contains the list of rules for layer 3 */ VNET_DEFINE(struct ip_fw_chain, layer3_chain); ipfw_nat_t *ipfw_nat_ptr = NULL; struct cfg_nat *(*lookup_nat_ptr)(struct nat_list *, int); ipfw_nat_cfg_t *ipfw_nat_cfg_ptr; ipfw_nat_cfg_t *ipfw_nat_del_ptr; ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr; ipfw_nat_cfg_t *ipfw_nat_get_log_ptr; #ifdef SYSCTL_NODE uint32_t dummy_def = IPFW_DEFAULT_RULE; uint32_t dummy_tables_max = IPFW_TABLES_MAX; SYSBEGIN(f3) SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, one_pass, CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_one_pass), 0, "Only do a single pass through ipfw when using dummynet(4)"); SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, CTLFLAG_RW, &VNET_NAME(autoinc_step), 0, "Rule number auto-increment step"); SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, verbose, CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_verbose), 0, "Log matches to ipfw rules"); SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_RW, &VNET_NAME(verbose_limit), 0, "Set upper limit of matches of ipfw rules logged"); SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD, &dummy_def, 0, "The default/max possible rule number."); SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, tables_max, CTLFLAG_RD, &dummy_tables_max, 0, "The maximum number of tables."); SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, default_to_accept, CTLFLAG_RDTUN, &default_to_accept, 0, "Make the default rule accept all packets."); TUNABLE_INT("net.inet.ip.fw.default_to_accept", &default_to_accept); SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, static_count, CTLFLAG_RD, &VNET_NAME(layer3_chain.n_rules), 0, "Number of static rules"); #ifdef INET6 SYSCTL_DECL(_net_inet6_ip6); SYSCTL_NODE(_net_inet6_ip6, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); SYSCTL_VNET_INT(_net_inet6_ip6_fw, OID_AUTO, deny_unknown_exthdrs, CTLFLAG_RW | CTLFLAG_SECURE, &VNET_NAME(fw_deny_unknown_exthdrs), 0, "Deny packets with unknown IPv6 Extension Headers"); #endif /* INET6 */ SYSEND #endif /* SYSCTL_NODE */ /* * Some macros used in the various matching options. * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T * Other macros just cast void * into the appropriate type */ #define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) #define TCP(p) ((struct tcphdr *)(p)) #define SCTP(p) ((struct sctphdr *)(p)) #define UDP(p) ((struct udphdr *)(p)) #define ICMP(p) ((struct icmphdr *)(p)) #define ICMP6(p) ((struct icmp6_hdr *)(p)) static __inline int icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd) { int type = icmp->icmp_type; return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<icmp_type; return (type <= ICMP_MAXTYPE && (TT & (1<arg1 or cmd->d[0]. * * We scan options and store the bits we find set. We succeed if * * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear * * The code is sometimes optimized not to store additional variables. */ static int flags_match(ipfw_insn *cmd, u_int8_t bits) { u_char want_clear; bits = ~bits; if ( ((cmd->arg1 & 0xff) & bits) != 0) return 0; /* some bits we want set were clear */ want_clear = (cmd->arg1 >> 8) & 0xff; if ( (want_clear & bits) != want_clear) return 0; /* some bits we want clear were set */ return 1; } static int ipopts_match(struct ip *ip, ipfw_insn *cmd) { int optlen, bits = 0; u_char *cp = (u_char *)(ip + 1); int x = (ip->ip_hl << 2) - sizeof (struct ip); for (; x > 0; x -= optlen, cp += optlen) { int opt = cp[IPOPT_OPTVAL]; if (opt == IPOPT_EOL) break; if (opt == IPOPT_NOP) optlen = 1; else { optlen = cp[IPOPT_OLEN]; if (optlen <= 0 || optlen > x) return 0; /* invalid or truncated */ } switch (opt) { default: break; case IPOPT_LSRR: bits |= IP_FW_IPOPT_LSRR; break; case IPOPT_SSRR: bits |= IP_FW_IPOPT_SSRR; break; case IPOPT_RR: bits |= IP_FW_IPOPT_RR; break; case IPOPT_TS: bits |= IP_FW_IPOPT_TS; break; } } return (flags_match(cmd, bits)); } static int tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd) { int optlen, bits = 0; u_char *cp = (u_char *)(tcp + 1); int x = (tcp->th_off << 2) - sizeof(struct tcphdr); for (; x > 0; x -= optlen, cp += optlen) { int opt = cp[0]; if (opt == TCPOPT_EOL) break; if (opt == TCPOPT_NOP) optlen = 1; else { optlen = cp[1]; if (optlen <= 0) break; } switch (opt) { default: break; case TCPOPT_MAXSEG: bits |= IP_FW_TCPOPT_MSS; break; case TCPOPT_WINDOW: bits |= IP_FW_TCPOPT_WINDOW; break; case TCPOPT_SACK_PERMITTED: case TCPOPT_SACK: bits |= IP_FW_TCPOPT_SACK; break; case TCPOPT_TIMESTAMP: bits |= IP_FW_TCPOPT_TS; break; } } return (flags_match(cmd, bits)); } static int iface_match(struct ifnet *ifp, ipfw_insn_if *cmd) { if (ifp == NULL) /* no iface with this packet, match fails */ return 0; /* Check by name or by IP address */ if (cmd->name[0] != '\0') { /* match by name */ /* Check name */ if (cmd->p.glob) { if (fnmatch(cmd->name, ifp->if_xname, 0) == 0) return(1); } else { if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0) return(1); } } else { #ifdef __FreeBSD__ /* and OSX too ? */ struct ifaddr *ia; if_addr_rlock(ifp); TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { if (ia->ifa_addr->sa_family != AF_INET) continue; if (cmd->p.ip.s_addr == ((struct sockaddr_in *) (ia->ifa_addr))->sin_addr.s_addr) { if_addr_runlock(ifp); return(1); /* match */ } } if_addr_runlock(ifp); #endif /* __FreeBSD__ */ } return(0); /* no match, fail ... */ } /* * The verify_path function checks if a route to the src exists and * if it is reachable via ifp (when provided). * * The 'verrevpath' option checks that the interface that an IP packet * arrives on is the same interface that traffic destined for the * packet's source address would be routed out of. * The 'versrcreach' option just checks that the source address is * reachable via any route (except default) in the routing table. * These two are a measure to block forged packets. This is also * commonly known as "anti-spoofing" or Unicast Reverse Path * Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs * is purposely reminiscent of the Cisco IOS command, * * ip verify unicast reverse-path * ip verify unicast source reachable-via any * * which implements the same functionality. But note that the syntax * is misleading, and the check may be performed on all IP packets * whether unicast, multicast, or broadcast. */ static int verify_path(struct in_addr src, struct ifnet *ifp, u_int fib) { #ifndef __FreeBSD__ return 0; #else struct route ro; struct sockaddr_in *dst; bzero(&ro, sizeof(ro)); dst = (struct sockaddr_in *)&(ro.ro_dst); dst->sin_family = AF_INET; dst->sin_len = sizeof(*dst); dst->sin_addr = src; in_rtalloc_ign(&ro, 0, fib); if (ro.ro_rt == NULL) return 0; /* * If ifp is provided, check for equality with rtentry. * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, * in order to pass packets injected back by if_simloop(): * if useloopback == 1 routing entry (via lo0) for our own address * may exist, so we need to handle routing assymetry. */ if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { RTFREE(ro.ro_rt); return 0; } /* if no ifp provided, check if rtentry is not default route */ if (ifp == NULL && satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) { RTFREE(ro.ro_rt); return 0; } /* or if this is a blackhole/reject route */ if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { RTFREE(ro.ro_rt); return 0; } /* found valid route */ RTFREE(ro.ro_rt); return 1; #endif /* __FreeBSD__ */ } #ifdef INET6 /* * ipv6 specific rules here... */ static __inline int icmp6type_match (int type, ipfw_insn_u32 *cmd) { return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) ); } static int flow6id_match( int curr_flow, ipfw_insn_u32 *cmd ) { int i; for (i=0; i <= cmd->o.arg1; ++i ) if (curr_flow == cmd->d[i] ) return 1; return 0; } /* support for IP6_*_ME opcodes */ static int search_ip6_addr_net (struct in6_addr * ip6_addr) { struct ifnet *mdc; struct ifaddr *mdc2; struct in6_ifaddr *fdm; struct in6_addr copia; TAILQ_FOREACH(mdc, &V_ifnet, if_link) { if_addr_rlock(mdc); TAILQ_FOREACH(mdc2, &mdc->if_addrhead, ifa_link) { if (mdc2->ifa_addr->sa_family == AF_INET6) { fdm = (struct in6_ifaddr *)mdc2; copia = fdm->ia_addr.sin6_addr; /* need for leaving scope_id in the sock_addr */ in6_clearscope(&copia); if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia)) { if_addr_runlock(mdc); return 1; } } } if_addr_runlock(mdc); } return 0; } static int verify_path6(struct in6_addr *src, struct ifnet *ifp) { struct route_in6 ro; struct sockaddr_in6 *dst; bzero(&ro, sizeof(ro)); dst = (struct sockaddr_in6 * )&(ro.ro_dst); dst->sin6_family = AF_INET6; dst->sin6_len = sizeof(*dst); dst->sin6_addr = *src; /* XXX MRT 0 for ipv6 at this time */ rtalloc_ign((struct route *)&ro, 0); if (ro.ro_rt == NULL) return 0; /* * if ifp is provided, check for equality with rtentry * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, * to support the case of sending packets to an address of our own. * (where the former interface is the first argument of if_simloop() * (=ifp), the latter is lo0) */ if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { RTFREE(ro.ro_rt); return 0; } /* if no ifp provided, check if rtentry is not default route */ if (ifp == NULL && IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) { RTFREE(ro.ro_rt); return 0; } /* or if this is a blackhole/reject route */ if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { RTFREE(ro.ro_rt); return 0; } /* found valid route */ RTFREE(ro.ro_rt); return 1; } static int is_icmp6_query(int icmp6_type) { if ((icmp6_type <= ICMP6_MAXTYPE) && (icmp6_type == ICMP6_ECHO_REQUEST || icmp6_type == ICMP6_MEMBERSHIP_QUERY || icmp6_type == ICMP6_WRUREQUEST || icmp6_type == ICMP6_FQDN_QUERY || icmp6_type == ICMP6_NI_QUERY)) return (1); return (0); } static void send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6) { struct mbuf *m; m = args->m; if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) { struct tcphdr *tcp; tcp = (struct tcphdr *)((char *)ip6 + hlen); if ((tcp->th_flags & TH_RST) == 0) { struct mbuf *m0; m0 = ipfw_send_pkt(args->m, &(args->f_id), ntohl(tcp->th_seq), ntohl(tcp->th_ack), tcp->th_flags | TH_RST); if (m0 != NULL) ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL); } FREE_PKT(m); } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */ #if 0 /* * Unlike above, the mbufs need to line up with the ip6 hdr, * as the contents are read. We need to m_adj() the * needed amount. * The mbuf will however be thrown away so we can adjust it. * Remember we did an m_pullup on it already so we * can make some assumptions about contiguousness. */ if (args->L3offset) m_adj(m, args->L3offset); #endif icmp6_error(m, ICMP6_DST_UNREACH, code, 0); } else FREE_PKT(m); args->m = NULL; } #endif /* INET6 */ /* * sends a reject message, consuming the mbuf passed as an argument. */ static void send_reject(struct ip_fw_args *args, int code, int iplen, struct ip *ip) { #if 0 /* XXX When ip is not guaranteed to be at mtod() we will * need to account for this */ * The mbuf will however be thrown away so we can adjust it. * Remember we did an m_pullup on it already so we * can make some assumptions about contiguousness. */ if (args->L3offset) m_adj(m, args->L3offset); #endif if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */ /* We need the IP header in host order for icmp_error(). */ SET_HOST_IPLEN(ip); icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); } else if (args->f_id.proto == IPPROTO_TCP) { struct tcphdr *const tcp = L3HDR(struct tcphdr, mtod(args->m, struct ip *)); if ( (tcp->th_flags & TH_RST) == 0) { struct mbuf *m; m = ipfw_send_pkt(args->m, &(args->f_id), ntohl(tcp->th_seq), ntohl(tcp->th_ack), tcp->th_flags | TH_RST); if (m != NULL) ip_output(m, NULL, NULL, 0, NULL, NULL); } FREE_PKT(args->m); } else FREE_PKT(args->m); args->m = NULL; } /* * Support for uid/gid/jail lookup. These tests are expensive * (because we may need to look into the list of active sockets) * so we cache the results. ugid_lookupp is 0 if we have not * yet done a lookup, 1 if we succeeded, and -1 if we tried * and failed. The function always returns the match value. * We could actually spare the variable and use *uc, setting * it to '(void *)check_uidgid if we have no info, NULL if * we tried and failed, or any other value if successful. */ static int check_uidgid(ipfw_insn_u32 *insn, int proto, struct ifnet *oif, struct in_addr dst_ip, u_int16_t dst_port, struct in_addr src_ip, u_int16_t src_port, int *ugid_lookupp, struct ucred **uc, struct inpcb *inp) { #ifndef __FreeBSD__ return cred_check(insn, proto, oif, dst_ip, dst_port, src_ip, src_port, (struct bsd_ucred *)uc, ugid_lookupp, ((struct mbuf *)inp)->m_skb); #else /* FreeBSD */ struct inpcbinfo *pi; - int wildcard; + int lookupflags; struct inpcb *pcb; int match; /* * Check to see if the UDP or TCP stack supplied us with * the PCB. If so, rather then holding a lock and looking * up the PCB, we can use the one that was supplied. */ if (inp && *ugid_lookupp == 0) { INP_LOCK_ASSERT(inp); if (inp->inp_socket != NULL) { *uc = crhold(inp->inp_cred); *ugid_lookupp = 1; } else *ugid_lookupp = -1; } /* * If we have already been here and the packet has no * PCB entry associated with it, then we can safely * assume that this is a no match. */ if (*ugid_lookupp == -1) return (0); if (proto == IPPROTO_TCP) { - wildcard = 0; + lookupflags = 0; pi = &V_tcbinfo; } else if (proto == IPPROTO_UDP) { - wildcard = INPLOOKUP_WILDCARD; + lookupflags = INPLOOKUP_WILDCARD; pi = &V_udbinfo; } else return 0; + lookupflags |= INPLOOKUP_RLOCKPCB; match = 0; if (*ugid_lookupp == 0) { - INP_INFO_RLOCK(pi); pcb = (oif) ? - in_pcblookup_hash(pi, + in_pcblookup(pi, dst_ip, htons(dst_port), src_ip, htons(src_port), - wildcard, oif) : - in_pcblookup_hash(pi, + lookupflags, oif) : + in_pcblookup(pi, src_ip, htons(src_port), dst_ip, htons(dst_port), - wildcard, NULL); + lookupflags, NULL); if (pcb != NULL) { + INP_RLOCK_ASSERT(pcb); *uc = crhold(pcb->inp_cred); *ugid_lookupp = 1; + INP_RUNLOCK(pcb); } - INP_INFO_RUNLOCK(pi); if (*ugid_lookupp == 0) { /* * We tried and failed, set the variable to -1 * so we will not try again on this packet. */ *ugid_lookupp = -1; return (0); } } if (insn->o.opcode == O_UID) match = ((*uc)->cr_uid == (uid_t)insn->d[0]); else if (insn->o.opcode == O_GID) match = groupmember((gid_t)insn->d[0], *uc); else if (insn->o.opcode == O_JAIL) match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]); return match; #endif /* __FreeBSD__ */ } /* * Helper function to set args with info on the rule after the matching * one. slot is precise, whereas we guess rule_id as they are * assigned sequentially. */ static inline void set_match(struct ip_fw_args *args, int slot, struct ip_fw_chain *chain) { args->rule.chain_id = chain->id; args->rule.slot = slot + 1; /* we use 0 as a marker */ args->rule.rule_id = 1 + chain->map[slot]->id; args->rule.rulenum = chain->map[slot]->rulenum; } /* * The main check routine for the firewall. * * All arguments are in args so we can modify them and return them * back to the caller. * * Parameters: * * args->m (in/out) The packet; we set to NULL when/if we nuke it. * Starts with the IP header. * args->eh (in) Mac header if present, NULL for layer3 packet. * args->L3offset Number of bytes bypassed if we came from L2. * e.g. often sizeof(eh) ** NOTYET ** * args->oif Outgoing interface, NULL if packet is incoming. * The incoming interface is in the mbuf. (in) * args->divert_rule (in/out) * Skip up to the first rule past this rule number; * upon return, non-zero port number for divert or tee. * * args->rule Pointer to the last matching rule (in/out) * args->next_hop Socket we are forwarding to (out). * args->f_id Addresses grabbed from the packet (out) * args->rule.info a cookie depending on rule action * * Return value: * * IP_FW_PASS the packet must be accepted * IP_FW_DENY the packet must be dropped * IP_FW_DIVERT divert packet, port in m_tag * IP_FW_TEE tee packet, port in m_tag * IP_FW_DUMMYNET to dummynet, pipe in args->cookie * IP_FW_NETGRAPH into netgraph, cookie args->cookie * args->rule contains the matching rule, * args->rule.info has additional information. * */ int ipfw_chk(struct ip_fw_args *args) { /* * Local variables holding state while processing a packet: * * IMPORTANT NOTE: to speed up the processing of rules, there * are some assumption on the values of the variables, which * are documented here. Should you change them, please check * the implementation of the various instructions to make sure * that they still work. * * args->eh The MAC header. It is non-null for a layer2 * packet, it is NULL for a layer-3 packet. * **notyet** * args->L3offset Offset in the packet to the L3 (IP or equiv.) header. * * m | args->m Pointer to the mbuf, as received from the caller. * It may change if ipfw_chk() does an m_pullup, or if it * consumes the packet because it calls send_reject(). * XXX This has to change, so that ipfw_chk() never modifies * or consumes the buffer. * ip is the beginning of the ip(4 or 6) header. * Calculated by adding the L3offset to the start of data. * (Until we start using L3offset, the packet is * supposed to start with the ip header). */ struct mbuf *m = args->m; struct ip *ip = mtod(m, struct ip *); /* * For rules which contain uid/gid or jail constraints, cache * a copy of the users credentials after the pcb lookup has been * executed. This will speed up the processing of rules with * these types of constraints, as well as decrease contention * on pcb related locks. */ #ifndef __FreeBSD__ struct bsd_ucred ucred_cache; #else struct ucred *ucred_cache = NULL; #endif int ucred_lookup = 0; /* * oif | args->oif If NULL, ipfw_chk has been called on the * inbound path (ether_input, ip_input). * If non-NULL, ipfw_chk has been called on the outbound path * (ether_output, ip_output). */ struct ifnet *oif = args->oif; int f_pos = 0; /* index of current rule in the array */ int retval = 0; /* * hlen The length of the IP header. */ u_int hlen = 0; /* hlen >0 means we have an IP pkt */ /* * offset The offset of a fragment. offset != 0 means that * we have a fragment at this offset of an IPv4 packet. * offset == 0 means that (if this is an IPv4 packet) * this is the first or only fragment. * For IPv6 offset == 0 means there is no Fragment Header. * If offset != 0 for IPv6 always use correct mask to * get the correct offset because we add IP6F_MORE_FRAG * to be able to dectect the first fragment which would * otherwise have offset = 0. */ u_short offset = 0; /* * Local copies of addresses. They are only valid if we have * an IP packet. * * proto The protocol. Set to 0 for non-ip packets, * or to the protocol read from the packet otherwise. * proto != 0 means that we have an IPv4 packet. * * src_port, dst_port port numbers, in HOST format. Only * valid for TCP and UDP packets. * * src_ip, dst_ip ip addresses, in NETWORK format. * Only valid for IPv4 packets. */ uint8_t proto; uint16_t src_port = 0, dst_port = 0; /* NOTE: host format */ struct in_addr src_ip, dst_ip; /* NOTE: network format */ uint16_t iplen=0; int pktlen; uint16_t etype = 0; /* Host order stored ether type */ /* * dyn_dir = MATCH_UNKNOWN when rules unchecked, * MATCH_NONE when checked and not matched (q = NULL), * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) */ int dyn_dir = MATCH_UNKNOWN; ipfw_dyn_rule *q = NULL; struct ip_fw_chain *chain = &V_layer3_chain; /* * We store in ulp a pointer to the upper layer protocol header. * In the ipv4 case this is easy to determine from the header, * but for ipv6 we might have some additional headers in the middle. * ulp is NULL if not found. */ void *ulp = NULL; /* upper layer protocol pointer. */ /* XXX ipv6 variables */ int is_ipv6 = 0; uint8_t icmp6_type = 0; uint16_t ext_hd = 0; /* bits vector for extension header filtering */ /* end of ipv6 variables */ int is_ipv4 = 0; int done = 0; /* flag to exit the outer loop */ if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready)) return (IP_FW_PASS); /* accept */ dst_ip.s_addr = 0; /* make sure it is initialized */ src_ip.s_addr = 0; /* make sure it is initialized */ pktlen = m->m_pkthdr.len; args->f_id.fib = M_GETFIB(m); /* note mbuf not altered) */ proto = args->f_id.proto = 0; /* mark f_id invalid */ /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */ /* * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous, * then it sets p to point at the offset "len" in the mbuf. WARNING: the * pointer might become stale after other pullups (but we never use it * this way). */ #define PULLUP_TO(_len, p, T) PULLUP_LEN(_len, p, sizeof(T)) #define PULLUP_LEN(_len, p, T) \ do { \ int x = (_len) + T; \ if ((m)->m_len < x) { \ args->m = m = m_pullup(m, x); \ if (m == NULL) \ goto pullup_failed; \ } \ p = (mtod(m, char *) + (_len)); \ } while (0) /* * if we have an ether header, */ if (args->eh) etype = ntohs(args->eh->ether_type); /* Identify IP packets and fill up variables. */ if (pktlen >= sizeof(struct ip6_hdr) && (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) { struct ip6_hdr *ip6 = (struct ip6_hdr *)ip; is_ipv6 = 1; args->f_id.addr_type = 6; hlen = sizeof(struct ip6_hdr); proto = ip6->ip6_nxt; /* Search extension headers to find upper layer protocols */ while (ulp == NULL) { switch (proto) { case IPPROTO_ICMPV6: PULLUP_TO(hlen, ulp, struct icmp6_hdr); icmp6_type = ICMP6(ulp)->icmp6_type; break; case IPPROTO_TCP: PULLUP_TO(hlen, ulp, struct tcphdr); dst_port = TCP(ulp)->th_dport; src_port = TCP(ulp)->th_sport; /* save flags for dynamic rules */ args->f_id._flags = TCP(ulp)->th_flags; break; case IPPROTO_SCTP: PULLUP_TO(hlen, ulp, struct sctphdr); src_port = SCTP(ulp)->src_port; dst_port = SCTP(ulp)->dest_port; break; case IPPROTO_UDP: PULLUP_TO(hlen, ulp, struct udphdr); dst_port = UDP(ulp)->uh_dport; src_port = UDP(ulp)->uh_sport; break; case IPPROTO_HOPOPTS: /* RFC 2460 */ PULLUP_TO(hlen, ulp, struct ip6_hbh); ext_hd |= EXT_HOPOPTS; hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; ulp = NULL; break; case IPPROTO_ROUTING: /* RFC 2460 */ PULLUP_TO(hlen, ulp, struct ip6_rthdr); switch (((struct ip6_rthdr *)ulp)->ip6r_type) { case 0: ext_hd |= EXT_RTHDR0; break; case 2: ext_hd |= EXT_RTHDR2; break; default: printf("IPFW2: IPV6 - Unknown Routing " "Header type(%d)\n", ((struct ip6_rthdr *)ulp)->ip6r_type); if (V_fw_deny_unknown_exthdrs) return (IP_FW_DENY); break; } ext_hd |= EXT_ROUTING; hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3; proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt; ulp = NULL; break; case IPPROTO_FRAGMENT: /* RFC 2460 */ PULLUP_TO(hlen, ulp, struct ip6_frag); ext_hd |= EXT_FRAGMENT; hlen += sizeof (struct ip6_frag); proto = ((struct ip6_frag *)ulp)->ip6f_nxt; offset = ((struct ip6_frag *)ulp)->ip6f_offlg & IP6F_OFF_MASK; /* Add IP6F_MORE_FRAG for offset of first * fragment to be != 0. */ offset |= ((struct ip6_frag *)ulp)->ip6f_offlg & IP6F_MORE_FRAG; if (offset == 0) { printf("IPFW2: IPV6 - Invalid Fragment " "Header\n"); if (V_fw_deny_unknown_exthdrs) return (IP_FW_DENY); break; } args->f_id.extra = ntohl(((struct ip6_frag *)ulp)->ip6f_ident); ulp = NULL; break; case IPPROTO_DSTOPTS: /* RFC 2460 */ PULLUP_TO(hlen, ulp, struct ip6_hbh); ext_hd |= EXT_DSTOPTS; hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; ulp = NULL; break; case IPPROTO_AH: /* RFC 2402 */ PULLUP_TO(hlen, ulp, struct ip6_ext); ext_hd |= EXT_AH; hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2; proto = ((struct ip6_ext *)ulp)->ip6e_nxt; ulp = NULL; break; case IPPROTO_ESP: /* RFC 2406 */ PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */ /* Anything past Seq# is variable length and * data past this ext. header is encrypted. */ ext_hd |= EXT_ESP; break; case IPPROTO_NONE: /* RFC 2460 */ /* * Packet ends here, and IPv6 header has * already been pulled up. If ip6e_len!=0 * then octets must be ignored. */ ulp = ip; /* non-NULL to get out of loop. */ break; case IPPROTO_OSPFIGP: /* XXX OSPF header check? */ PULLUP_TO(hlen, ulp, struct ip6_ext); break; case IPPROTO_PIM: /* XXX PIM header check? */ PULLUP_TO(hlen, ulp, struct pim); break; case IPPROTO_CARP: PULLUP_TO(hlen, ulp, struct carp_header); if (((struct carp_header *)ulp)->carp_version != CARP_VERSION) return (IP_FW_DENY); if (((struct carp_header *)ulp)->carp_type != CARP_ADVERTISEMENT) return (IP_FW_DENY); break; case IPPROTO_IPV6: /* RFC 2893 */ PULLUP_TO(hlen, ulp, struct ip6_hdr); break; case IPPROTO_IPV4: /* RFC 2893 */ PULLUP_TO(hlen, ulp, struct ip); break; default: printf("IPFW2: IPV6 - Unknown Extension " "Header(%d), ext_hd=%x\n", proto, ext_hd); if (V_fw_deny_unknown_exthdrs) return (IP_FW_DENY); PULLUP_TO(hlen, ulp, struct ip6_ext); break; } /*switch */ } ip = mtod(m, struct ip *); ip6 = (struct ip6_hdr *)ip; args->f_id.src_ip6 = ip6->ip6_src; args->f_id.dst_ip6 = ip6->ip6_dst; args->f_id.src_ip = 0; args->f_id.dst_ip = 0; args->f_id.flow_id6 = ntohl(ip6->ip6_flow); } else if (pktlen >= sizeof(struct ip) && (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) { is_ipv4 = 1; hlen = ip->ip_hl << 2; args->f_id.addr_type = 4; /* * Collect parameters into local variables for faster matching. */ proto = ip->ip_p; src_ip = ip->ip_src; dst_ip = ip->ip_dst; offset = ntohs(ip->ip_off) & IP_OFFMASK; iplen = ntohs(ip->ip_len); pktlen = iplen < pktlen ? iplen : pktlen; if (offset == 0) { switch (proto) { case IPPROTO_TCP: PULLUP_TO(hlen, ulp, struct tcphdr); dst_port = TCP(ulp)->th_dport; src_port = TCP(ulp)->th_sport; /* save flags for dynamic rules */ args->f_id._flags = TCP(ulp)->th_flags; break; case IPPROTO_SCTP: PULLUP_TO(hlen, ulp, struct sctphdr); src_port = SCTP(ulp)->src_port; dst_port = SCTP(ulp)->dest_port; break; case IPPROTO_UDP: PULLUP_TO(hlen, ulp, struct udphdr); dst_port = UDP(ulp)->uh_dport; src_port = UDP(ulp)->uh_sport; break; case IPPROTO_ICMP: PULLUP_TO(hlen, ulp, struct icmphdr); //args->f_id.flags = ICMP(ulp)->icmp_type; break; default: break; } } ip = mtod(m, struct ip *); args->f_id.src_ip = ntohl(src_ip.s_addr); args->f_id.dst_ip = ntohl(dst_ip.s_addr); } #undef PULLUP_TO if (proto) { /* we may have port numbers, store them */ args->f_id.proto = proto; args->f_id.src_port = src_port = ntohs(src_port); args->f_id.dst_port = dst_port = ntohs(dst_port); } IPFW_RLOCK(chain); if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */ IPFW_RUNLOCK(chain); return (IP_FW_PASS); /* accept */ } if (args->rule.slot) { /* * Packet has already been tagged as a result of a previous * match on rule args->rule aka args->rule_id (PIPE, QUEUE, * REASS, NETGRAPH, DIVERT/TEE...) * Validate the slot and continue from the next one * if still present, otherwise do a lookup. */ f_pos = (args->rule.chain_id == chain->id) ? args->rule.slot : ipfw_find_rule(chain, args->rule.rulenum, args->rule.rule_id); } else { f_pos = 0; } /* * Now scan the rules, and parse microinstructions for each rule. * We have two nested loops and an inner switch. Sometimes we * need to break out of one or both loops, or re-enter one of * the loops with updated variables. Loop variables are: * * f_pos (outer loop) points to the current rule. * On output it points to the matching rule. * done (outer loop) is used as a flag to break the loop. * l (inner loop) residual length of current rule. * cmd points to the current microinstruction. * * We break the inner loop by setting l=0 and possibly * cmdlen=0 if we don't want to advance cmd. * We break the outer loop by setting done=1 * We can restart the inner loop by setting l>0 and f_pos, f, cmd * as needed. */ for (; f_pos < chain->n_rules; f_pos++) { ipfw_insn *cmd; uint32_t tablearg = 0; int l, cmdlen, skip_or; /* skip rest of OR block */ struct ip_fw *f; f = chain->map[f_pos]; if (V_set_disable & (1 << f->set) ) continue; skip_or = 0; for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; l -= cmdlen, cmd += cmdlen) { int match; /* * check_body is a jump target used when we find a * CHECK_STATE, and need to jump to the body of * the target rule. */ /* check_body: */ cmdlen = F_LEN(cmd); /* * An OR block (insn_1 || .. || insn_n) has the * F_OR bit set in all but the last instruction. * The first match will set "skip_or", and cause * the following instructions to be skipped until * past the one with the F_OR bit clear. */ if (skip_or) { /* skip this instruction */ if ((cmd->len & F_OR) == 0) skip_or = 0; /* next one is good */ continue; } match = 0; /* set to 1 if we succeed */ switch (cmd->opcode) { /* * The first set of opcodes compares the packet's * fields with some pattern, setting 'match' if a * match is found. At the end of the loop there is * logic to deal with F_NOT and F_OR flags associated * with the opcode. */ case O_NOP: match = 1; break; case O_FORWARD_MAC: printf("ipfw: opcode %d unimplemented\n", cmd->opcode); break; case O_GID: case O_UID: case O_JAIL: /* * We only check offset == 0 && proto != 0, * as this ensures that we have a * packet with the ports info. */ if (offset!=0) break; if (is_ipv6) /* XXX to be fixed later */ break; if (proto == IPPROTO_TCP || proto == IPPROTO_UDP) match = check_uidgid( (ipfw_insn_u32 *)cmd, proto, oif, dst_ip, dst_port, src_ip, src_port, &ucred_lookup, #ifdef __FreeBSD__ &ucred_cache, args->inp); #else (void *)&ucred_cache, (struct inpcb *)args->m); #endif break; case O_RECV: match = iface_match(m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); break; case O_XMIT: match = iface_match(oif, (ipfw_insn_if *)cmd); break; case O_VIA: match = iface_match(oif ? oif : m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); break; case O_MACADDR2: if (args->eh != NULL) { /* have MAC header */ u_int32_t *want = (u_int32_t *) ((ipfw_insn_mac *)cmd)->addr; u_int32_t *mask = (u_int32_t *) ((ipfw_insn_mac *)cmd)->mask; u_int32_t *hdr = (u_int32_t *)args->eh; match = ( want[0] == (hdr[0] & mask[0]) && want[1] == (hdr[1] & mask[1]) && want[2] == (hdr[2] & mask[2]) ); } break; case O_MAC_TYPE: if (args->eh != NULL) { u_int16_t *p = ((ipfw_insn_u16 *)cmd)->ports; int i; for (i = cmdlen - 1; !match && i>0; i--, p += 2) match = (etype >= p[0] && etype <= p[1]); } break; case O_FRAG: match = (offset != 0); break; case O_IN: /* "out" is "not in" */ match = (oif == NULL); break; case O_LAYER2: match = (args->eh != NULL); break; case O_DIVERTED: { /* For diverted packets, args->rule.info * contains the divert port (in host format) * reason and direction. */ uint32_t i = args->rule.info; match = (i&IPFW_IS_MASK) == IPFW_IS_DIVERT && cmd->arg1 & ((i & IPFW_INFO_IN) ? 1 : 2); } break; case O_PROTO: /* * We do not allow an arg of 0 so the * check of "proto" only suffices. */ match = (proto == cmd->arg1); break; case O_IP_SRC: match = is_ipv4 && (((ipfw_insn_ip *)cmd)->addr.s_addr == src_ip.s_addr); break; case O_IP_SRC_LOOKUP: case O_IP_DST_LOOKUP: if (is_ipv4) { uint32_t key = (cmd->opcode == O_IP_DST_LOOKUP) ? dst_ip.s_addr : src_ip.s_addr; uint32_t v = 0; if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) { /* generic lookup. The key must be * in 32bit big-endian format. */ v = ((ipfw_insn_u32 *)cmd)->d[1]; if (v == 0) key = dst_ip.s_addr; else if (v == 1) key = src_ip.s_addr; else if (v == 6) /* dscp */ key = (ip->ip_tos >> 2) & 0x3f; else if (offset != 0) break; else if (proto != IPPROTO_TCP && proto != IPPROTO_UDP) break; else if (v == 2) key = htonl(dst_port); else if (v == 3) key = htonl(src_port); else if (v == 4 || v == 5) { check_uidgid( (ipfw_insn_u32 *)cmd, proto, oif, dst_ip, dst_port, src_ip, src_port, &ucred_lookup, #ifdef __FreeBSD__ &ucred_cache, args->inp); if (v == 4 /* O_UID */) key = ucred_cache->cr_uid; else if (v == 5 /* O_JAIL */) key = ucred_cache->cr_prison->pr_id; #else /* !__FreeBSD__ */ (void *)&ucred_cache, (struct inpcb *)args->m); if (v ==4 /* O_UID */) key = ucred_cache.uid; else if (v == 5 /* O_JAIL */) key = ucred_cache.xid; #endif /* !__FreeBSD__ */ key = htonl(key); } else break; } match = ipfw_lookup_table(chain, cmd->arg1, key, &v); if (!match) break; if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) match = ((ipfw_insn_u32 *)cmd)->d[0] == v; else tablearg = v; } break; case O_IP_SRC_MASK: case O_IP_DST_MASK: if (is_ipv4) { uint32_t a = (cmd->opcode == O_IP_DST_MASK) ? dst_ip.s_addr : src_ip.s_addr; uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; int i = cmdlen-1; for (; !match && i>0; i-= 2, p+= 2) match = (p[0] == (a & p[1])); } break; case O_IP_SRC_ME: if (is_ipv4) { struct ifnet *tif; INADDR_TO_IFP(src_ip, tif); match = (tif != NULL); break; } #ifdef INET6 /* FALLTHROUGH */ case O_IP6_SRC_ME: match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6); #endif break; case O_IP_DST_SET: case O_IP_SRC_SET: if (is_ipv4) { u_int32_t *d = (u_int32_t *)(cmd+1); u_int32_t addr = cmd->opcode == O_IP_DST_SET ? args->f_id.dst_ip : args->f_id.src_ip; if (addr < d[0]) break; addr -= d[0]; /* subtract base */ match = (addr < cmd->arg1) && ( d[ 1 + (addr>>5)] & (1<<(addr & 0x1f)) ); } break; case O_IP_DST: match = is_ipv4 && (((ipfw_insn_ip *)cmd)->addr.s_addr == dst_ip.s_addr); break; case O_IP_DST_ME: if (is_ipv4) { struct ifnet *tif; INADDR_TO_IFP(dst_ip, tif); match = (tif != NULL); break; } #ifdef INET6 /* FALLTHROUGH */ case O_IP6_DST_ME: match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6); #endif break; case O_IP_SRCPORT: case O_IP_DSTPORT: /* * offset == 0 && proto != 0 is enough * to guarantee that we have a * packet with port info. */ if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) && offset == 0) { u_int16_t x = (cmd->opcode == O_IP_SRCPORT) ? src_port : dst_port ; u_int16_t *p = ((ipfw_insn_u16 *)cmd)->ports; int i; for (i = cmdlen - 1; !match && i>0; i--, p += 2) match = (x>=p[0] && x<=p[1]); } break; case O_ICMPTYPE: match = (offset == 0 && proto==IPPROTO_ICMP && icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) ); break; #ifdef INET6 case O_ICMP6TYPE: match = is_ipv6 && offset == 0 && proto==IPPROTO_ICMPV6 && icmp6type_match( ICMP6(ulp)->icmp6_type, (ipfw_insn_u32 *)cmd); break; #endif /* INET6 */ case O_IPOPT: match = (is_ipv4 && ipopts_match(ip, cmd) ); break; case O_IPVER: match = (is_ipv4 && cmd->arg1 == ip->ip_v); break; case O_IPID: case O_IPLEN: case O_IPTTL: if (is_ipv4) { /* only for IP packets */ uint16_t x; uint16_t *p; int i; if (cmd->opcode == O_IPLEN) x = iplen; else if (cmd->opcode == O_IPTTL) x = ip->ip_ttl; else /* must be IPID */ x = ntohs(ip->ip_id); if (cmdlen == 1) { match = (cmd->arg1 == x); break; } /* otherwise we have ranges */ p = ((ipfw_insn_u16 *)cmd)->ports; i = cmdlen - 1; for (; !match && i>0; i--, p += 2) match = (x >= p[0] && x <= p[1]); } break; case O_IPPRECEDENCE: match = (is_ipv4 && (cmd->arg1 == (ip->ip_tos & 0xe0)) ); break; case O_IPTOS: match = (is_ipv4 && flags_match(cmd, ip->ip_tos)); break; case O_TCPDATALEN: if (proto == IPPROTO_TCP && offset == 0) { struct tcphdr *tcp; uint16_t x; uint16_t *p; int i; tcp = TCP(ulp); x = iplen - ((ip->ip_hl + tcp->th_off) << 2); if (cmdlen == 1) { match = (cmd->arg1 == x); break; } /* otherwise we have ranges */ p = ((ipfw_insn_u16 *)cmd)->ports; i = cmdlen - 1; for (; !match && i>0; i--, p += 2) match = (x >= p[0] && x <= p[1]); } break; case O_TCPFLAGS: match = (proto == IPPROTO_TCP && offset == 0 && flags_match(cmd, TCP(ulp)->th_flags)); break; case O_TCPOPTS: PULLUP_LEN(hlen, ulp, (TCP(ulp)->th_off << 2)); match = (proto == IPPROTO_TCP && offset == 0 && tcpopts_match(TCP(ulp), cmd)); break; case O_TCPSEQ: match = (proto == IPPROTO_TCP && offset == 0 && ((ipfw_insn_u32 *)cmd)->d[0] == TCP(ulp)->th_seq); break; case O_TCPACK: match = (proto == IPPROTO_TCP && offset == 0 && ((ipfw_insn_u32 *)cmd)->d[0] == TCP(ulp)->th_ack); break; case O_TCPWIN: match = (proto == IPPROTO_TCP && offset == 0 && cmd->arg1 == TCP(ulp)->th_win); break; case O_ESTAB: /* reject packets which have SYN only */ /* XXX should i also check for TH_ACK ? */ match = (proto == IPPROTO_TCP && offset == 0 && (TCP(ulp)->th_flags & (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); break; case O_ALTQ: { struct pf_mtag *at; ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; match = 1; at = pf_find_mtag(m); if (at != NULL && at->qid != 0) break; at = pf_get_mtag(m); if (at == NULL) { /* * Let the packet fall back to the * default ALTQ. */ break; } at->qid = altq->qid; if (is_ipv4) at->af = AF_INET; else at->af = AF_LINK; at->hdr = ip; break; } case O_LOG: ipfw_log(f, hlen, args, m, oif, offset, tablearg, ip); match = 1; break; case O_PROB: match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); break; case O_VERREVPATH: /* Outgoing packets automatically pass/match */ match = ((oif != NULL) || (m->m_pkthdr.rcvif == NULL) || ( #ifdef INET6 is_ipv6 ? verify_path6(&(args->f_id.src_ip6), m->m_pkthdr.rcvif) : #endif verify_path(src_ip, m->m_pkthdr.rcvif, args->f_id.fib))); break; case O_VERSRCREACH: /* Outgoing packets automatically pass/match */ match = (hlen > 0 && ((oif != NULL) || #ifdef INET6 is_ipv6 ? verify_path6(&(args->f_id.src_ip6), NULL) : #endif verify_path(src_ip, NULL, args->f_id.fib))); break; case O_ANTISPOOF: /* Outgoing packets automatically pass/match */ if (oif == NULL && hlen > 0 && ( (is_ipv4 && in_localaddr(src_ip)) #ifdef INET6 || (is_ipv6 && in6_localaddr(&(args->f_id.src_ip6))) #endif )) match = #ifdef INET6 is_ipv6 ? verify_path6( &(args->f_id.src_ip6), m->m_pkthdr.rcvif) : #endif verify_path(src_ip, m->m_pkthdr.rcvif, args->f_id.fib); else match = 1; break; case O_IPSEC: #ifdef IPSEC match = (m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); #endif /* otherwise no match */ break; #ifdef INET6 case O_IP6_SRC: match = is_ipv6 && IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6, &((ipfw_insn_ip6 *)cmd)->addr6); break; case O_IP6_DST: match = is_ipv6 && IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6, &((ipfw_insn_ip6 *)cmd)->addr6); break; case O_IP6_SRC_MASK: case O_IP6_DST_MASK: if (is_ipv6) { int i = cmdlen - 1; struct in6_addr p; struct in6_addr *d = &((ipfw_insn_ip6 *)cmd)->addr6; for (; !match && i > 0; d += 2, i -= F_INSN_SIZE(struct in6_addr) * 2) { p = (cmd->opcode == O_IP6_SRC_MASK) ? args->f_id.src_ip6: args->f_id.dst_ip6; APPLY_MASK(&p, &d[1]); match = IN6_ARE_ADDR_EQUAL(&d[0], &p); } } break; case O_FLOW6ID: match = is_ipv6 && flow6id_match(args->f_id.flow_id6, (ipfw_insn_u32 *) cmd); break; case O_EXT_HDR: match = is_ipv6 && (ext_hd & ((ipfw_insn *) cmd)->arg1); break; case O_IP6: match = is_ipv6; break; #endif case O_IP4: match = is_ipv4; break; case O_TAG: { struct m_tag *mtag; uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? tablearg : cmd->arg1; /* Packet is already tagged with this tag? */ mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL); /* We have `untag' action when F_NOT flag is * present. And we must remove this mtag from * mbuf and reset `match' to zero (`match' will * be inversed later). * Otherwise we should allocate new mtag and * push it into mbuf. */ if (cmd->len & F_NOT) { /* `untag' action */ if (mtag != NULL) m_tag_delete(m, mtag); match = 0; } else { if (mtag == NULL) { mtag = m_tag_alloc( MTAG_IPFW, tag, 0, M_NOWAIT); if (mtag != NULL) m_tag_prepend(m, mtag); } match = 1; } break; } case O_FIB: /* try match the specified fib */ if (args->f_id.fib == cmd->arg1) match = 1; break; case O_SOCKARG: { struct inpcb *inp = args->inp; struct inpcbinfo *pi; if (is_ipv6) /* XXX can we remove this ? */ break; if (proto == IPPROTO_TCP) pi = &V_tcbinfo; else if (proto == IPPROTO_UDP) pi = &V_udbinfo; else break; + /* + * XXXRW: so_user_cookie should almost + * certainly be inp_user_cookie? + */ + /* For incomming packet, lookup up the inpcb using the src/dest ip/port tuple */ if (inp == NULL) { - INP_INFO_RLOCK(pi); - inp = in_pcblookup_hash(pi, + inp = in_pcblookup(pi, src_ip, htons(src_port), dst_ip, htons(dst_port), - 0, NULL); - INP_INFO_RUNLOCK(pi); - } - - if (inp && inp->inp_socket) { - tablearg = inp->inp_socket->so_user_cookie; - if (tablearg) - match = 1; + INPLOOKUP_RLOCKPCB, NULL); + if (inp != NULL) { + tablearg = + inp->inp_socket->so_user_cookie; + if (tablearg) + match = 1; + INP_RUNLOCK(inp); + } + } else { + if (inp->inp_socket) { + tablearg = + inp->inp_socket->so_user_cookie; + if (tablearg) + match = 1; + } } break; } case O_TAGGED: { struct m_tag *mtag; uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? tablearg : cmd->arg1; if (cmdlen == 1) { match = m_tag_locate(m, MTAG_IPFW, tag, NULL) != NULL; break; } /* we have ranges */ for (mtag = m_tag_first(m); mtag != NULL && !match; mtag = m_tag_next(m, mtag)) { uint16_t *p; int i; if (mtag->m_tag_cookie != MTAG_IPFW) continue; p = ((ipfw_insn_u16 *)cmd)->ports; i = cmdlen - 1; for(; !match && i > 0; i--, p += 2) match = mtag->m_tag_id >= p[0] && mtag->m_tag_id <= p[1]; } break; } /* * The second set of opcodes represents 'actions', * i.e. the terminal part of a rule once the packet * matches all previous patterns. * Typically there is only one action for each rule, * and the opcode is stored at the end of the rule * (but there are exceptions -- see below). * * In general, here we set retval and terminate the * outer loop (would be a 'break 3' in some language, * but we need to set l=0, done=1) * * Exceptions: * O_COUNT and O_SKIPTO actions: * instead of terminating, we jump to the next rule * (setting l=0), or to the SKIPTO target (setting * f/f_len, cmd and l as needed), respectively. * * O_TAG, O_LOG and O_ALTQ action parameters: * perform some action and set match = 1; * * O_LIMIT and O_KEEP_STATE: these opcodes are * not real 'actions', and are stored right * before the 'action' part of the rule. * These opcodes try to install an entry in the * state tables; if successful, we continue with * the next opcode (match=1; break;), otherwise * the packet must be dropped (set retval, * break loops with l=0, done=1) * * O_PROBE_STATE and O_CHECK_STATE: these opcodes * cause a lookup of the state table, and a jump * to the 'action' part of the parent rule * if an entry is found, or * (CHECK_STATE only) a jump to the next rule if * the entry is not found. * The result of the lookup is cached so that * further instances of these opcodes become NOPs. * The jump to the next rule is done by setting * l=0, cmdlen=0. */ case O_LIMIT: case O_KEEP_STATE: if (ipfw_install_state(f, (ipfw_insn_limit *)cmd, args, tablearg)) { /* error or limit violation */ retval = IP_FW_DENY; l = 0; /* exit inner loop */ done = 1; /* exit outer loop */ } match = 1; break; case O_PROBE_STATE: case O_CHECK_STATE: /* * dynamic rules are checked at the first * keep-state or check-state occurrence, * with the result being stored in dyn_dir. * The compiler introduces a PROBE_STATE * instruction for us when we have a * KEEP_STATE (because PROBE_STATE needs * to be run first). */ if (dyn_dir == MATCH_UNKNOWN && (q = ipfw_lookup_dyn_rule(&args->f_id, &dyn_dir, proto == IPPROTO_TCP ? TCP(ulp) : NULL)) != NULL) { /* * Found dynamic entry, update stats * and jump to the 'action' part of * the parent rule by setting * f, cmd, l and clearing cmdlen. */ q->pcnt++; q->bcnt += pktlen; /* XXX we would like to have f_pos * readily accessible in the dynamic * rule, instead of having to * lookup q->rule. */ f = q->rule; f_pos = ipfw_find_rule(chain, f->rulenum, f->id); cmd = ACTION_PTR(f); l = f->cmd_len - f->act_ofs; ipfw_dyn_unlock(); cmdlen = 0; match = 1; break; } /* * Dynamic entry not found. If CHECK_STATE, * skip to next rule, if PROBE_STATE just * ignore and continue with next opcode. */ if (cmd->opcode == O_CHECK_STATE) l = 0; /* exit inner loop */ match = 1; break; case O_ACCEPT: retval = 0; /* accept */ l = 0; /* exit inner loop */ done = 1; /* exit outer loop */ break; case O_PIPE: case O_QUEUE: set_match(args, f_pos, chain); args->rule.info = (cmd->arg1 == IP_FW_TABLEARG) ? tablearg : cmd->arg1; if (cmd->opcode == O_PIPE) args->rule.info |= IPFW_IS_PIPE; if (V_fw_one_pass) args->rule.info |= IPFW_ONEPASS; retval = IP_FW_DUMMYNET; l = 0; /* exit inner loop */ done = 1; /* exit outer loop */ break; case O_DIVERT: case O_TEE: if (args->eh) /* not on layer 2 */ break; /* otherwise this is terminal */ l = 0; /* exit inner loop */ done = 1; /* exit outer loop */ retval = (cmd->opcode == O_DIVERT) ? IP_FW_DIVERT : IP_FW_TEE; set_match(args, f_pos, chain); args->rule.info = (cmd->arg1 == IP_FW_TABLEARG) ? tablearg : cmd->arg1; break; case O_COUNT: f->pcnt++; /* update stats */ f->bcnt += pktlen; f->timestamp = time_uptime; l = 0; /* exit inner loop */ break; case O_SKIPTO: f->pcnt++; /* update stats */ f->bcnt += pktlen; f->timestamp = time_uptime; /* If possible use cached f_pos (in f->next_rule), * whose version is written in f->next_rule * (horrible hacks to avoid changing the ABI). */ if (cmd->arg1 != IP_FW_TABLEARG && (uintptr_t)f->x_next == chain->id) { f_pos = (uintptr_t)f->next_rule; } else { int i = (cmd->arg1 == IP_FW_TABLEARG) ? tablearg : cmd->arg1; /* make sure we do not jump backward */ if (i <= f->rulenum) i = f->rulenum + 1; f_pos = ipfw_find_rule(chain, i, 0); /* update the cache */ if (cmd->arg1 != IP_FW_TABLEARG) { f->next_rule = (void *)(uintptr_t)f_pos; f->x_next = (void *)(uintptr_t)chain->id; } } /* * Skip disabled rules, and re-enter * the inner loop with the correct * f_pos, f, l and cmd. * Also clear cmdlen and skip_or */ for (; f_pos < chain->n_rules - 1 && (V_set_disable & (1 << chain->map[f_pos]->set)); f_pos++) ; /* Re-enter the inner loop at the skipto rule. */ f = chain->map[f_pos]; l = f->cmd_len; cmd = f->cmd; match = 1; cmdlen = 0; skip_or = 0; continue; break; /* not reached */ case O_REJECT: /* * Drop the packet and send a reject notice * if the packet is not ICMP (or is an ICMP * query), and it is not multicast/broadcast. */ if (hlen > 0 && is_ipv4 && offset == 0 && (proto != IPPROTO_ICMP || is_icmp_query(ICMP(ulp))) && !(m->m_flags & (M_BCAST|M_MCAST)) && !IN_MULTICAST(ntohl(dst_ip.s_addr))) { send_reject(args, cmd->arg1, iplen, ip); m = args->m; } /* FALLTHROUGH */ #ifdef INET6 case O_UNREACH6: if (hlen > 0 && is_ipv6 && ((offset & IP6F_OFF_MASK) == 0) && (proto != IPPROTO_ICMPV6 || (is_icmp6_query(icmp6_type) == 1)) && !(m->m_flags & (M_BCAST|M_MCAST)) && !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) { send_reject6( args, cmd->arg1, hlen, (struct ip6_hdr *)ip); m = args->m; } /* FALLTHROUGH */ #endif case O_DENY: retval = IP_FW_DENY; l = 0; /* exit inner loop */ done = 1; /* exit outer loop */ break; case O_FORWARD_IP: if (args->eh) /* not valid on layer2 pkts */ break; if (!q || dyn_dir == MATCH_FORWARD) { struct sockaddr_in *sa; sa = &(((ipfw_insn_sa *)cmd)->sa); if (sa->sin_addr.s_addr == INADDR_ANY) { bcopy(sa, &args->hopstore, sizeof(*sa)); args->hopstore.sin_addr.s_addr = htonl(tablearg); args->next_hop = &args->hopstore; } else { args->next_hop = sa; } } retval = IP_FW_PASS; l = 0; /* exit inner loop */ done = 1; /* exit outer loop */ break; case O_NETGRAPH: case O_NGTEE: set_match(args, f_pos, chain); args->rule.info = (cmd->arg1 == IP_FW_TABLEARG) ? tablearg : cmd->arg1; if (V_fw_one_pass) args->rule.info |= IPFW_ONEPASS; retval = (cmd->opcode == O_NETGRAPH) ? IP_FW_NETGRAPH : IP_FW_NGTEE; l = 0; /* exit inner loop */ done = 1; /* exit outer loop */ break; case O_SETFIB: { uint32_t fib; f->pcnt++; /* update stats */ f->bcnt += pktlen; f->timestamp = time_uptime; fib = (cmd->arg1 == IP_FW_TABLEARG) ? tablearg: cmd->arg1; if (fib >= rt_numfibs) fib = 0; M_SETFIB(m, fib); args->f_id.fib = fib; l = 0; /* exit inner loop */ break; } case O_NAT: if (!IPFW_NAT_LOADED) { retval = IP_FW_DENY; } else { struct cfg_nat *t; int nat_id; set_match(args, f_pos, chain); t = ((ipfw_insn_nat *)cmd)->nat; if (t == NULL) { nat_id = (cmd->arg1 == IP_FW_TABLEARG) ? tablearg : cmd->arg1; t = (*lookup_nat_ptr)(&chain->nat, nat_id); if (t == NULL) { retval = IP_FW_DENY; l = 0; /* exit inner loop */ done = 1; /* exit outer loop */ break; } if (cmd->arg1 != IP_FW_TABLEARG) ((ipfw_insn_nat *)cmd)->nat = t; } retval = ipfw_nat_ptr(args, t, m); } l = 0; /* exit inner loop */ done = 1; /* exit outer loop */ break; case O_REASS: { int ip_off; f->pcnt++; f->bcnt += pktlen; l = 0; /* in any case exit inner loop */ ip_off = ntohs(ip->ip_off); /* if not fragmented, go to next rule */ if ((ip_off & (IP_MF | IP_OFFMASK)) == 0) break; /* * ip_reass() expects len & off in host * byte order. */ SET_HOST_IPLEN(ip); args->m = m = ip_reass(m); /* * do IP header checksum fixup. */ if (m == NULL) { /* fragment got swallowed */ retval = IP_FW_DENY; } else { /* good, packet complete */ int hlen; ip = mtod(m, struct ip *); hlen = ip->ip_hl << 2; SET_NET_IPLEN(ip); ip->ip_sum = 0; if (hlen == sizeof(struct ip)) ip->ip_sum = in_cksum_hdr(ip); else ip->ip_sum = in_cksum(m, hlen); retval = IP_FW_REASS; set_match(args, f_pos, chain); } done = 1; /* exit outer loop */ break; } default: panic("-- unknown opcode %d\n", cmd->opcode); } /* end of switch() on opcodes */ /* * if we get here with l=0, then match is irrelevant. */ if (cmd->len & F_NOT) match = !match; if (match) { if (cmd->len & F_OR) skip_or = 1; } else { if (!(cmd->len & F_OR)) /* not an OR block, */ break; /* try next rule */ } } /* end of inner loop, scan opcodes */ #undef PULLUP_LEN if (done) break; /* next_rule:; */ /* try next rule */ } /* end of outer for, scan rules */ if (done) { struct ip_fw *rule = chain->map[f_pos]; /* Update statistics */ rule->pcnt++; rule->bcnt += pktlen; rule->timestamp = time_uptime; } else { retval = IP_FW_DENY; printf("ipfw: ouch!, skip past end of rules, denying packet\n"); } IPFW_RUNLOCK(chain); #ifdef __FreeBSD__ if (ucred_cache != NULL) crfree(ucred_cache); #endif return (retval); pullup_failed: if (V_fw_verbose) printf("ipfw: pullup failed\n"); return (IP_FW_DENY); } /* * Module and VNET glue */ /* * Stuff that must be initialised only on boot or module load */ static int ipfw_init(void) { int error = 0; ipfw_dyn_attach(); /* * Only print out this stuff the first time around, * when called from the sysinit code. */ printf("ipfw2 " #ifdef INET6 "(+ipv6) " #endif "initialized, divert %s, nat %s, " "rule-based forwarding " #ifdef IPFIREWALL_FORWARD "enabled, " #else "disabled, " #endif "default to %s, logging ", #ifdef IPDIVERT "enabled", #else "loadable", #endif #ifdef IPFIREWALL_NAT "enabled", #else "loadable", #endif default_to_accept ? "accept" : "deny"); /* * Note: V_xxx variables can be accessed here but the vnet specific * initializer may not have been called yet for the VIMAGE case. * Tuneables will have been processed. We will print out values for * the default vnet. * XXX This should all be rationalized AFTER 8.0 */ if (V_fw_verbose == 0) printf("disabled\n"); else if (V_verbose_limit == 0) printf("unlimited\n"); else printf("limited to %d packets/entry by default\n", V_verbose_limit); ipfw_log_bpf(1); /* init */ return (error); } /* * Called for the removal of the last instance only on module unload. */ static void ipfw_destroy(void) { ipfw_log_bpf(0); /* uninit */ ipfw_dyn_detach(); printf("IP firewall unloaded\n"); } /* * Stuff that must be initialized for every instance * (including the first of course). */ static int vnet_ipfw_init(const void *unused) { int error; struct ip_fw *rule = NULL; struct ip_fw_chain *chain; chain = &V_layer3_chain; /* First set up some values that are compile time options */ V_autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ V_fw_deny_unknown_exthdrs = 1; #ifdef IPFIREWALL_VERBOSE V_fw_verbose = 1; #endif #ifdef IPFIREWALL_VERBOSE_LIMIT V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT; #endif #ifdef IPFIREWALL_NAT LIST_INIT(&chain->nat); #endif /* insert the default rule and create the initial map */ chain->n_rules = 1; chain->static_len = sizeof(struct ip_fw); chain->map = malloc(sizeof(struct ip_fw *), M_IPFW, M_NOWAIT | M_ZERO); if (chain->map) rule = malloc(chain->static_len, M_IPFW, M_NOWAIT | M_ZERO); if (rule == NULL) { if (chain->map) free(chain->map, M_IPFW); printf("ipfw2: ENOSPC initializing default rule " "(support disabled)\n"); return (ENOSPC); } error = ipfw_init_tables(chain); if (error) { panic("init_tables"); /* XXX Marko fix this ! */ } /* fill and insert the default rule */ rule->act_ofs = 0; rule->rulenum = IPFW_DEFAULT_RULE; rule->cmd_len = 1; rule->set = RESVD_SET; rule->cmd[0].len = 1; rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY; chain->rules = chain->default_rule = chain->map[0] = rule; chain->id = rule->id = 1; IPFW_LOCK_INIT(chain); ipfw_dyn_init(); /* First set up some values that are compile time options */ V_ipfw_vnet_ready = 1; /* Open for business */ /* * Hook the sockopt handler, and the layer2 (V_ip_fw_chk_ptr) * and pfil hooks for ipv4 and ipv6. Even if the latter two fail * we still keep the module alive because the sockopt and * layer2 paths are still useful. * ipfw[6]_hook return 0 on success, ENOENT on failure, * so we can ignore the exact return value and just set a flag. * * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so * changes in the underlying (per-vnet) variables trigger * immediate hook()/unhook() calls. * In layer2 we have the same behaviour, except that V_ether_ipfw * is checked on each packet because there are no pfil hooks. */ V_ip_fw_ctl_ptr = ipfw_ctl; V_ip_fw_chk_ptr = ipfw_chk; error = ipfw_attach_hooks(1); return (error); } /* * Called for the removal of each instance. */ static int vnet_ipfw_uninit(const void *unused) { struct ip_fw *reap, *rule; struct ip_fw_chain *chain = &V_layer3_chain; int i; V_ipfw_vnet_ready = 0; /* tell new callers to go away */ /* * disconnect from ipv4, ipv6, layer2 and sockopt. * Then grab, release and grab again the WLOCK so we make * sure the update is propagated and nobody will be in. */ (void)ipfw_attach_hooks(0 /* detach */); V_ip_fw_chk_ptr = NULL; V_ip_fw_ctl_ptr = NULL; IPFW_UH_WLOCK(chain); IPFW_UH_WUNLOCK(chain); IPFW_UH_WLOCK(chain); IPFW_WLOCK(chain); IPFW_WUNLOCK(chain); IPFW_WLOCK(chain); ipfw_dyn_uninit(0); /* run the callout_drain */ ipfw_destroy_tables(chain); reap = NULL; for (i = 0; i < chain->n_rules; i++) { rule = chain->map[i]; rule->x_next = reap; reap = rule; } if (chain->map) free(chain->map, M_IPFW); IPFW_WUNLOCK(chain); IPFW_UH_WUNLOCK(chain); if (reap != NULL) ipfw_reap_rules(reap); IPFW_LOCK_DESTROY(chain); ipfw_dyn_uninit(1); /* free the remaining parts */ return 0; } /* * Module event handler. * In general we have the choice of handling most of these events by the * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to * use the SYSINIT handlers as they are more capable of expressing the * flow of control during module and vnet operations, so this is just * a skeleton. Note there is no SYSINIT equivalent of the module * SHUTDOWN handler, but we don't have anything to do in that case anyhow. */ static int ipfw_modevent(module_t mod, int type, void *unused) { int err = 0; switch (type) { case MOD_LOAD: /* Called once at module load or * system boot if compiled in. */ break; case MOD_QUIESCE: /* Called before unload. May veto unloading. */ break; case MOD_UNLOAD: /* Called during unload. */ break; case MOD_SHUTDOWN: /* Called during system shutdown. */ break; default: err = EOPNOTSUPP; break; } return err; } static moduledata_t ipfwmod = { "ipfw", ipfw_modevent, 0 }; /* Define startup order. */ #define IPFW_SI_SUB_FIREWALL SI_SUB_PROTO_IFATTACHDOMAIN #define IPFW_MODEVENT_ORDER (SI_ORDER_ANY - 255) /* On boot slot in here. */ #define IPFW_MODULE_ORDER (IPFW_MODEVENT_ORDER + 1) /* A little later. */ #define IPFW_VNET_ORDER (IPFW_MODEVENT_ORDER + 2) /* Later still. */ DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER); MODULE_VERSION(ipfw, 2); /* should declare some dependencies here */ /* * Starting up. Done in order after ipfwmod() has been called. * VNET_SYSINIT is also called for each existing vnet and each new vnet. */ SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER, ipfw_init, NULL); VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER, vnet_ipfw_init, NULL); /* * Closing up shop. These are done in REVERSE ORDER, but still * after ipfwmod() has been called. Not called on reboot. * VNET_SYSUNINIT is also called for each exiting vnet as it exits. * or when the module is unloaded. */ SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER, ipfw_destroy, NULL); VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER, vnet_ipfw_uninit, NULL); /* end of file */ Index: head/sys/netinet/raw_ip.c =================================================================== --- head/sys/netinet/raw_ip.c (revision 222487) +++ head/sys/netinet/raw_ip.c (revision 222488) @@ -1,1124 +1,1115 @@ /*- * 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. * * @(#)raw_ip.c 8.7 (Berkeley) 5/15/95 */ #include __FBSDID("$FreeBSD$"); #include "opt_inet.h" #include "opt_inet6.h" #include "opt_ipsec.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef IPSEC #include #endif /*IPSEC*/ #include VNET_DEFINE(int, ip_defttl) = IPDEFTTL; SYSCTL_VNET_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW, &VNET_NAME(ip_defttl), 0, "Maximum TTL on IP packets"); VNET_DEFINE(struct inpcbhead, ripcb); VNET_DEFINE(struct inpcbinfo, ripcbinfo); #define V_ripcb VNET(ripcb) #define V_ripcbinfo VNET(ripcbinfo) /* * Control and data hooks for ipfw, dummynet, divert and so on. * The data hooks are not used here but it is convenient * to keep them all in one place. */ VNET_DEFINE(ip_fw_chk_ptr_t, ip_fw_chk_ptr) = NULL; VNET_DEFINE(ip_fw_ctl_ptr_t, ip_fw_ctl_ptr) = NULL; int (*ip_dn_ctl_ptr)(struct sockopt *); int (*ip_dn_io_ptr)(struct mbuf **, int, struct ip_fw_args *); void (*ip_divert_ptr)(struct mbuf *, int); int (*ng_ipfw_input_p)(struct mbuf **, int, struct ip_fw_args *, int); #ifdef INET /* * Hooks for multicast routing. They all default to NULL, so leave them not * initialized and rely on BSS being set to 0. */ /* * The socket used to communicate with the multicast routing daemon. */ VNET_DEFINE(struct socket *, ip_mrouter); /* * The various mrouter and rsvp functions. */ int (*ip_mrouter_set)(struct socket *, struct sockopt *); int (*ip_mrouter_get)(struct socket *, struct sockopt *); int (*ip_mrouter_done)(void); int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *, struct ip_moptions *); int (*mrt_ioctl)(u_long, caddr_t, int); int (*legal_vif_num)(int); u_long (*ip_mcast_src)(int); void (*rsvp_input_p)(struct mbuf *m, int off); int (*ip_rsvp_vif)(struct socket *, struct sockopt *); void (*ip_rsvp_force_done)(struct socket *); #endif /* INET */ u_long rip_sendspace = 9216; SYSCTL_ULONG(_net_inet_raw, OID_AUTO, maxdgram, CTLFLAG_RW, &rip_sendspace, 0, "Maximum outgoing raw IP datagram size"); u_long rip_recvspace = 9216; SYSCTL_ULONG(_net_inet_raw, OID_AUTO, recvspace, CTLFLAG_RW, &rip_recvspace, 0, "Maximum space for incoming raw IP datagrams"); /* * Hash functions */ #define INP_PCBHASH_RAW_SIZE 256 #define INP_PCBHASH_RAW(proto, laddr, faddr, mask) \ (((proto) + (laddr) + (faddr)) % (mask) + 1) #ifdef INET static void rip_inshash(struct inpcb *inp) { struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; struct inpcbhead *pcbhash; int hash; INP_INFO_WLOCK_ASSERT(pcbinfo); INP_WLOCK_ASSERT(inp); if (inp->inp_ip_p != 0 && inp->inp_laddr.s_addr != INADDR_ANY && inp->inp_faddr.s_addr != INADDR_ANY) { hash = INP_PCBHASH_RAW(inp->inp_ip_p, inp->inp_laddr.s_addr, inp->inp_faddr.s_addr, pcbinfo->ipi_hashmask); } else hash = 0; pcbhash = &pcbinfo->ipi_hashbase[hash]; LIST_INSERT_HEAD(pcbhash, inp, inp_hash); } static void rip_delhash(struct inpcb *inp) { INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo); INP_WLOCK_ASSERT(inp); LIST_REMOVE(inp, inp_hash); } #endif /* INET */ /* * Raw interface to IP protocol. */ /* * Initialize raw connection block q. */ static void rip_zone_change(void *tag) { uma_zone_set_max(V_ripcbinfo.ipi_zone, maxsockets); } static int rip_inpcb_init(void *mem, int size, int flags) { struct inpcb *inp = mem; INP_LOCK_INIT(inp, "inp", "rawinp"); return (0); } void rip_init(void) { in_pcbinfo_init(&V_ripcbinfo, "rip", &V_ripcb, INP_PCBHASH_RAW_SIZE, 1, "ripcb", rip_inpcb_init, NULL, UMA_ZONE_NOFREE); EVENTHANDLER_REGISTER(maxsockets_change, rip_zone_change, NULL, EVENTHANDLER_PRI_ANY); } #ifdef VIMAGE void rip_destroy(void) { in_pcbinfo_destroy(&V_ripcbinfo); } #endif #ifdef INET static int rip_append(struct inpcb *last, struct ip *ip, struct mbuf *n, struct sockaddr_in *ripsrc) { int policyfail = 0; - INP_RLOCK_ASSERT(last); + INP_LOCK_ASSERT(last); #ifdef IPSEC /* check AH/ESP integrity. */ if (ipsec4_in_reject(n, last)) { policyfail = 1; } #endif /* IPSEC */ #ifdef MAC if (!policyfail && mac_inpcb_check_deliver(last, n) != 0) policyfail = 1; #endif /* Check the minimum TTL for socket. */ if (last->inp_ip_minttl && last->inp_ip_minttl > ip->ip_ttl) policyfail = 1; if (!policyfail) { struct mbuf *opts = NULL; struct socket *so; so = last->inp_socket; if ((last->inp_flags & INP_CONTROLOPTS) || (so->so_options & (SO_TIMESTAMP | SO_BINTIME))) ip_savecontrol(last, &opts, ip, n); SOCKBUF_LOCK(&so->so_rcv); if (sbappendaddr_locked(&so->so_rcv, (struct sockaddr *)ripsrc, n, opts) == 0) { /* should notify about lost packet */ m_freem(n); if (opts) m_freem(opts); SOCKBUF_UNLOCK(&so->so_rcv); } else sorwakeup_locked(so); } else m_freem(n); return (policyfail); } /* * Setup generic address and protocol structures for raw_input routine, then * pass them along with mbuf chain. */ void rip_input(struct mbuf *m, int off) { struct ifnet *ifp; struct ip *ip = mtod(m, struct ip *); int proto = ip->ip_p; struct inpcb *inp, *last; struct sockaddr_in ripsrc; int hash; bzero(&ripsrc, sizeof(ripsrc)); ripsrc.sin_len = sizeof(ripsrc); ripsrc.sin_family = AF_INET; ripsrc.sin_addr = ip->ip_src; last = NULL; ifp = m->m_pkthdr.rcvif; hash = INP_PCBHASH_RAW(proto, ip->ip_src.s_addr, ip->ip_dst.s_addr, V_ripcbinfo.ipi_hashmask); INP_INFO_RLOCK(&V_ripcbinfo); LIST_FOREACH(inp, &V_ripcbinfo.ipi_hashbase[hash], inp_hash) { if (inp->inp_ip_p != proto) continue; #ifdef INET6 /* XXX inp locking */ if ((inp->inp_vflag & INP_IPV4) == 0) continue; #endif if (inp->inp_laddr.s_addr != ip->ip_dst.s_addr) continue; if (inp->inp_faddr.s_addr != ip->ip_src.s_addr) continue; if (jailed_without_vnet(inp->inp_cred)) { /* * XXX: If faddr was bound to multicast group, * jailed raw socket will drop datagram. */ if (prison_check_ip4(inp->inp_cred, &ip->ip_dst) != 0) continue; } if (last != NULL) { struct mbuf *n; n = m_copy(m, 0, (int)M_COPYALL); if (n != NULL) (void) rip_append(last, ip, n, &ripsrc); /* XXX count dropped packet */ INP_RUNLOCK(last); } INP_RLOCK(inp); last = inp; } LIST_FOREACH(inp, &V_ripcbinfo.ipi_hashbase[0], inp_hash) { if (inp->inp_ip_p && inp->inp_ip_p != proto) continue; #ifdef INET6 /* XXX inp locking */ if ((inp->inp_vflag & INP_IPV4) == 0) continue; #endif if (!in_nullhost(inp->inp_laddr) && !in_hosteq(inp->inp_laddr, ip->ip_dst)) continue; if (!in_nullhost(inp->inp_faddr) && !in_hosteq(inp->inp_faddr, ip->ip_src)) continue; if (jailed_without_vnet(inp->inp_cred)) { /* * Allow raw socket in jail to receive multicast; * assume process had PRIV_NETINET_RAW at attach, * and fall through into normal filter path if so. */ if (!IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) && prison_check_ip4(inp->inp_cred, &ip->ip_dst) != 0) continue; } /* * If this raw socket has multicast state, and we * have received a multicast, check if this socket * should receive it, as multicast filtering is now * the responsibility of the transport layer. */ if (inp->inp_moptions != NULL && IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { /* * If the incoming datagram is for IGMP, allow it * through unconditionally to the raw socket. * * In the case of IGMPv2, we may not have explicitly * joined the group, and may have set IFF_ALLMULTI * on the interface. imo_multi_filter() may discard * control traffic we actually need to see. * * Userland multicast routing daemons should continue * filter the control traffic appropriately. */ int blocked; blocked = MCAST_PASS; if (proto != IPPROTO_IGMP) { struct sockaddr_in group; bzero(&group, sizeof(struct sockaddr_in)); group.sin_len = sizeof(struct sockaddr_in); group.sin_family = AF_INET; group.sin_addr = ip->ip_dst; blocked = imo_multi_filter(inp->inp_moptions, ifp, (struct sockaddr *)&group, (struct sockaddr *)&ripsrc); } if (blocked != MCAST_PASS) { IPSTAT_INC(ips_notmember); continue; } } if (last != NULL) { struct mbuf *n; n = m_copy(m, 0, (int)M_COPYALL); if (n != NULL) (void) rip_append(last, ip, n, &ripsrc); /* XXX count dropped packet */ INP_RUNLOCK(last); } INP_RLOCK(inp); last = inp; } INP_INFO_RUNLOCK(&V_ripcbinfo); if (last != NULL) { if (rip_append(last, ip, m, &ripsrc) != 0) IPSTAT_INC(ips_delivered); INP_RUNLOCK(last); } else { m_freem(m); IPSTAT_INC(ips_noproto); IPSTAT_DEC(ips_delivered); } } /* * Generate IP header and pass packet to ip_output. Tack on options user may * have setup with control call. */ int rip_output(struct mbuf *m, struct socket *so, u_long dst) { struct ip *ip; int error; struct inpcb *inp = sotoinpcb(so); int flags = ((so->so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0) | IP_ALLOWBROADCAST; /* * If the user handed us a complete IP packet, use it. Otherwise, * allocate an mbuf for a header and fill it in. */ if ((inp->inp_flags & INP_HDRINCL) == 0) { if (m->m_pkthdr.len + sizeof(struct ip) > IP_MAXPACKET) { m_freem(m); return(EMSGSIZE); } M_PREPEND(m, sizeof(struct ip), M_DONTWAIT); if (m == NULL) return(ENOBUFS); INP_RLOCK(inp); ip = mtod(m, struct ip *); ip->ip_tos = inp->inp_ip_tos; if (inp->inp_flags & INP_DONTFRAG) ip->ip_off = IP_DF; else ip->ip_off = 0; ip->ip_p = inp->inp_ip_p; ip->ip_len = m->m_pkthdr.len; ip->ip_src = inp->inp_laddr; if (jailed(inp->inp_cred)) { /* * prison_local_ip4() would be good enough but would * let a source of INADDR_ANY pass, which we do not * want to see from jails. We do not go through the * pain of in_pcbladdr() for raw sockets. */ if (ip->ip_src.s_addr == INADDR_ANY) error = prison_get_ip4(inp->inp_cred, &ip->ip_src); else error = prison_local_ip4(inp->inp_cred, &ip->ip_src); if (error != 0) { INP_RUNLOCK(inp); m_freem(m); return (error); } } ip->ip_dst.s_addr = dst; ip->ip_ttl = inp->inp_ip_ttl; } else { if (m->m_pkthdr.len > IP_MAXPACKET) { m_freem(m); return(EMSGSIZE); } INP_RLOCK(inp); ip = mtod(m, struct ip *); error = prison_check_ip4(inp->inp_cred, &ip->ip_src); if (error != 0) { INP_RUNLOCK(inp); m_freem(m); return (error); } /* * Don't allow both user specified and setsockopt options, * and don't allow packet length sizes that will crash. */ if (((ip->ip_hl != (sizeof (*ip) >> 2)) && inp->inp_options) || (ip->ip_len > m->m_pkthdr.len) || (ip->ip_len < (ip->ip_hl << 2))) { INP_RUNLOCK(inp); m_freem(m); return (EINVAL); } if (ip->ip_id == 0) ip->ip_id = ip_newid(); /* * XXX prevent ip_output from overwriting header fields. */ flags |= IP_RAWOUTPUT; IPSTAT_INC(ips_rawout); } if (inp->inp_flags & INP_ONESBCAST) flags |= IP_SENDONES; #ifdef MAC mac_inpcb_create_mbuf(inp, m); #endif error = ip_output(m, inp->inp_options, NULL, flags, inp->inp_moptions, inp); INP_RUNLOCK(inp); return (error); } /* * Raw IP socket option processing. * * IMPORTANT NOTE regarding access control: Traditionally, raw sockets could * only be created by a privileged process, and as such, socket option * operations to manage system properties on any raw socket were allowed to * take place without explicit additional access control checks. However, * raw sockets can now also be created in jail(), and therefore explicit * checks are now required. Likewise, raw sockets can be used by a process * after it gives up privilege, so some caution is required. For options * passed down to the IP layer via ip_ctloutput(), checks are assumed to be * performed in ip_ctloutput() and therefore no check occurs here. * Unilaterally checking priv_check() here breaks normal IP socket option * operations on raw sockets. * * When adding new socket options here, make sure to add access control * checks here as necessary. */ int rip_ctloutput(struct socket *so, struct sockopt *sopt) { struct inpcb *inp = sotoinpcb(so); int error, optval; if (sopt->sopt_level != IPPROTO_IP) { if ((sopt->sopt_level == SOL_SOCKET) && (sopt->sopt_name == SO_SETFIB)) { inp->inp_inc.inc_fibnum = so->so_fibnum; return (0); } return (EINVAL); } error = 0; switch (sopt->sopt_dir) { case SOPT_GET: switch (sopt->sopt_name) { case IP_HDRINCL: optval = inp->inp_flags & INP_HDRINCL; error = sooptcopyout(sopt, &optval, sizeof optval); break; case IP_FW3: /* generic ipfw v.3 functions */ case IP_FW_ADD: /* ADD actually returns the body... */ case IP_FW_GET: case IP_FW_TABLE_GETSIZE: case IP_FW_TABLE_LIST: case IP_FW_NAT_GET_CONFIG: case IP_FW_NAT_GET_LOG: if (V_ip_fw_ctl_ptr != NULL) error = V_ip_fw_ctl_ptr(sopt); else error = ENOPROTOOPT; break; case IP_DUMMYNET3: /* generic dummynet v.3 functions */ case IP_DUMMYNET_GET: if (ip_dn_ctl_ptr != NULL) error = ip_dn_ctl_ptr(sopt); else error = ENOPROTOOPT; break ; case MRT_INIT: case MRT_DONE: case MRT_ADD_VIF: case MRT_DEL_VIF: case MRT_ADD_MFC: case MRT_DEL_MFC: case MRT_VERSION: case MRT_ASSERT: case MRT_API_SUPPORT: case MRT_API_CONFIG: case MRT_ADD_BW_UPCALL: case MRT_DEL_BW_UPCALL: error = priv_check(curthread, PRIV_NETINET_MROUTE); if (error != 0) return (error); error = ip_mrouter_get ? ip_mrouter_get(so, sopt) : EOPNOTSUPP; break; default: error = ip_ctloutput(so, sopt); break; } break; case SOPT_SET: switch (sopt->sopt_name) { case IP_HDRINCL: error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); if (error) break; if (optval) inp->inp_flags |= INP_HDRINCL; else inp->inp_flags &= ~INP_HDRINCL; break; case IP_FW3: /* generic ipfw v.3 functions */ case IP_FW_ADD: case IP_FW_DEL: case IP_FW_FLUSH: case IP_FW_ZERO: case IP_FW_RESETLOG: case IP_FW_TABLE_ADD: case IP_FW_TABLE_DEL: case IP_FW_TABLE_FLUSH: case IP_FW_NAT_CFG: case IP_FW_NAT_DEL: if (V_ip_fw_ctl_ptr != NULL) error = V_ip_fw_ctl_ptr(sopt); else error = ENOPROTOOPT; break; case IP_DUMMYNET3: /* generic dummynet v.3 functions */ case IP_DUMMYNET_CONFIGURE: case IP_DUMMYNET_DEL: case IP_DUMMYNET_FLUSH: if (ip_dn_ctl_ptr != NULL) error = ip_dn_ctl_ptr(sopt); else error = ENOPROTOOPT ; break ; case IP_RSVP_ON: error = priv_check(curthread, PRIV_NETINET_MROUTE); if (error != 0) return (error); error = ip_rsvp_init(so); break; case IP_RSVP_OFF: error = priv_check(curthread, PRIV_NETINET_MROUTE); if (error != 0) return (error); error = ip_rsvp_done(); break; case IP_RSVP_VIF_ON: case IP_RSVP_VIF_OFF: error = priv_check(curthread, PRIV_NETINET_MROUTE); if (error != 0) return (error); error = ip_rsvp_vif ? ip_rsvp_vif(so, sopt) : EINVAL; break; case MRT_INIT: case MRT_DONE: case MRT_ADD_VIF: case MRT_DEL_VIF: case MRT_ADD_MFC: case MRT_DEL_MFC: case MRT_VERSION: case MRT_ASSERT: case MRT_API_SUPPORT: case MRT_API_CONFIG: case MRT_ADD_BW_UPCALL: case MRT_DEL_BW_UPCALL: error = priv_check(curthread, PRIV_NETINET_MROUTE); if (error != 0) return (error); error = ip_mrouter_set ? ip_mrouter_set(so, sopt) : EOPNOTSUPP; break; default: error = ip_ctloutput(so, sopt); break; } break; } return (error); } /* * This function exists solely to receive the PRC_IFDOWN messages which are * sent by if_down(). It looks for an ifaddr whose ifa_addr is sa, and calls * in_ifadown() to remove all routes corresponding to that address. It also * receives the PRC_IFUP messages from if_up() and reinstalls the interface * routes. */ void rip_ctlinput(int cmd, struct sockaddr *sa, void *vip) { struct in_ifaddr *ia; struct ifnet *ifp; int err; int flags; switch (cmd) { case PRC_IFDOWN: IN_IFADDR_RLOCK(); TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { if (ia->ia_ifa.ifa_addr == sa && (ia->ia_flags & IFA_ROUTE)) { ifa_ref(&ia->ia_ifa); IN_IFADDR_RUNLOCK(); /* * in_ifscrub kills the interface route. */ in_ifscrub(ia->ia_ifp, ia, 0); /* * in_ifadown gets rid of all the rest of the * routes. This is not quite the right thing * to do, but at least if we are running a * routing process they will come back. */ in_ifadown(&ia->ia_ifa, 0); ifa_free(&ia->ia_ifa); break; } } if (ia == NULL) /* If ia matched, already unlocked. */ IN_IFADDR_RUNLOCK(); break; case PRC_IFUP: IN_IFADDR_RLOCK(); TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { if (ia->ia_ifa.ifa_addr == sa) break; } if (ia == NULL || (ia->ia_flags & IFA_ROUTE)) { IN_IFADDR_RUNLOCK(); return; } ifa_ref(&ia->ia_ifa); IN_IFADDR_RUNLOCK(); flags = RTF_UP; ifp = ia->ia_ifa.ifa_ifp; if ((ifp->if_flags & IFF_LOOPBACK) || (ifp->if_flags & IFF_POINTOPOINT)) flags |= RTF_HOST; err = ifa_del_loopback_route((struct ifaddr *)ia, sa); if (err == 0) ia->ia_flags &= ~IFA_RTSELF; err = rtinit(&ia->ia_ifa, RTM_ADD, flags); if (err == 0) ia->ia_flags |= IFA_ROUTE; err = ifa_add_loopback_route((struct ifaddr *)ia, sa); if (err == 0) ia->ia_flags |= IFA_RTSELF; ifa_free(&ia->ia_ifa); break; } } static int rip_attach(struct socket *so, int proto, struct thread *td) { struct inpcb *inp; int error; inp = sotoinpcb(so); KASSERT(inp == NULL, ("rip_attach: inp != NULL")); error = priv_check(td, PRIV_NETINET_RAW); if (error) return (error); if (proto >= IPPROTO_MAX || proto < 0) return EPROTONOSUPPORT; error = soreserve(so, rip_sendspace, rip_recvspace); if (error) return (error); INP_INFO_WLOCK(&V_ripcbinfo); error = in_pcballoc(so, &V_ripcbinfo); if (error) { INP_INFO_WUNLOCK(&V_ripcbinfo); return (error); } inp = (struct inpcb *)so->so_pcb; inp->inp_vflag |= INP_IPV4; inp->inp_ip_p = proto; inp->inp_ip_ttl = V_ip_defttl; rip_inshash(inp); INP_INFO_WUNLOCK(&V_ripcbinfo); INP_WUNLOCK(inp); return (0); } static void rip_detach(struct socket *so) { struct inpcb *inp; inp = sotoinpcb(so); KASSERT(inp != NULL, ("rip_detach: inp == NULL")); KASSERT(inp->inp_faddr.s_addr == INADDR_ANY, ("rip_detach: not closed")); INP_INFO_WLOCK(&V_ripcbinfo); INP_WLOCK(inp); rip_delhash(inp); if (so == V_ip_mrouter && ip_mrouter_done) ip_mrouter_done(); if (ip_rsvp_force_done) ip_rsvp_force_done(so); if (so == V_ip_rsvpd) ip_rsvp_done(); in_pcbdetach(inp); in_pcbfree(inp); INP_INFO_WUNLOCK(&V_ripcbinfo); } static void rip_dodisconnect(struct socket *so, struct inpcb *inp) { + struct inpcbinfo *pcbinfo; - INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo); - INP_WLOCK_ASSERT(inp); - + pcbinfo = inp->inp_pcbinfo; + INP_INFO_WLOCK(pcbinfo); + INP_WLOCK(inp); rip_delhash(inp); inp->inp_faddr.s_addr = INADDR_ANY; rip_inshash(inp); SOCK_LOCK(so); so->so_state &= ~SS_ISCONNECTED; SOCK_UNLOCK(so); + INP_WUNLOCK(inp); + INP_INFO_WUNLOCK(pcbinfo); } static void rip_abort(struct socket *so) { struct inpcb *inp; inp = sotoinpcb(so); KASSERT(inp != NULL, ("rip_abort: inp == NULL")); - INP_INFO_WLOCK(&V_ripcbinfo); - INP_WLOCK(inp); rip_dodisconnect(so, inp); - INP_WUNLOCK(inp); - INP_INFO_WUNLOCK(&V_ripcbinfo); } static void rip_close(struct socket *so) { struct inpcb *inp; inp = sotoinpcb(so); KASSERT(inp != NULL, ("rip_close: inp == NULL")); - INP_INFO_WLOCK(&V_ripcbinfo); - INP_WLOCK(inp); rip_dodisconnect(so, inp); - INP_WUNLOCK(inp); - INP_INFO_WUNLOCK(&V_ripcbinfo); } static int rip_disconnect(struct socket *so) { struct inpcb *inp; if ((so->so_state & SS_ISCONNECTED) == 0) return (ENOTCONN); inp = sotoinpcb(so); KASSERT(inp != NULL, ("rip_disconnect: inp == NULL")); - INP_INFO_WLOCK(&V_ripcbinfo); - INP_WLOCK(inp); rip_dodisconnect(so, inp); - INP_WUNLOCK(inp); - INP_INFO_WUNLOCK(&V_ripcbinfo); return (0); } static int rip_bind(struct socket *so, struct sockaddr *nam, struct thread *td) { struct sockaddr_in *addr = (struct sockaddr_in *)nam; struct inpcb *inp; int error; if (nam->sa_len != sizeof(*addr)) return (EINVAL); error = prison_check_ip4(td->td_ucred, &addr->sin_addr); if (error != 0) return (error); inp = sotoinpcb(so); KASSERT(inp != NULL, ("rip_bind: inp == NULL")); if (TAILQ_EMPTY(&V_ifnet) || (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) || (addr->sin_addr.s_addr && (inp->inp_flags & INP_BINDANY) == 0 && ifa_ifwithaddr_check((struct sockaddr *)addr) == 0)) return (EADDRNOTAVAIL); INP_INFO_WLOCK(&V_ripcbinfo); INP_WLOCK(inp); rip_delhash(inp); inp->inp_laddr = addr->sin_addr; rip_inshash(inp); INP_WUNLOCK(inp); INP_INFO_WUNLOCK(&V_ripcbinfo); return (0); } static int rip_connect(struct socket *so, struct sockaddr *nam, struct thread *td) { struct sockaddr_in *addr = (struct sockaddr_in *)nam; struct inpcb *inp; if (nam->sa_len != sizeof(*addr)) return (EINVAL); if (TAILQ_EMPTY(&V_ifnet)) return (EADDRNOTAVAIL); if (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) return (EAFNOSUPPORT); inp = sotoinpcb(so); KASSERT(inp != NULL, ("rip_connect: inp == NULL")); INP_INFO_WLOCK(&V_ripcbinfo); INP_WLOCK(inp); rip_delhash(inp); inp->inp_faddr = addr->sin_addr; rip_inshash(inp); soisconnected(so); INP_WUNLOCK(inp); INP_INFO_WUNLOCK(&V_ripcbinfo); return (0); } static int rip_shutdown(struct socket *so) { struct inpcb *inp; inp = sotoinpcb(so); KASSERT(inp != NULL, ("rip_shutdown: inp == NULL")); INP_WLOCK(inp); socantsendmore(so); INP_WUNLOCK(inp); return (0); } static int rip_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, struct mbuf *control, struct thread *td) { struct inpcb *inp; u_long dst; inp = sotoinpcb(so); KASSERT(inp != NULL, ("rip_send: inp == NULL")); /* * Note: 'dst' reads below are unlocked. */ if (so->so_state & SS_ISCONNECTED) { if (nam) { m_freem(m); return (EISCONN); } dst = inp->inp_faddr.s_addr; /* Unlocked read. */ } else { if (nam == NULL) { m_freem(m); return (ENOTCONN); } dst = ((struct sockaddr_in *)nam)->sin_addr.s_addr; } return (rip_output(m, so, dst)); } #endif /* INET */ static int rip_pcblist(SYSCTL_HANDLER_ARGS) { int error, i, n; struct inpcb *inp, **inp_list; inp_gen_t gencnt; struct xinpgen xig; /* * The process of preparing the TCB list is too time-consuming and * resource-intensive to repeat twice on every request. */ if (req->oldptr == 0) { n = V_ripcbinfo.ipi_count; n += imax(n / 8, 10); req->oldidx = 2 * (sizeof xig) + n * sizeof(struct xinpcb); return (0); } if (req->newptr != 0) return (EPERM); /* * OK, now we're committed to doing something. */ INP_INFO_RLOCK(&V_ripcbinfo); gencnt = V_ripcbinfo.ipi_gencnt; n = V_ripcbinfo.ipi_count; INP_INFO_RUNLOCK(&V_ripcbinfo); 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); inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK); if (inp_list == 0) return (ENOMEM); INP_INFO_RLOCK(&V_ripcbinfo); for (inp = LIST_FIRST(V_ripcbinfo.ipi_listhead), i = 0; inp && i < n; inp = LIST_NEXT(inp, inp_list)) { INP_WLOCK(inp); if (inp->inp_gencnt <= gencnt && cr_canseeinpcb(req->td->td_ucred, inp) == 0) { in_pcbref(inp); inp_list[i++] = inp; } INP_WUNLOCK(inp); } INP_INFO_RUNLOCK(&V_ripcbinfo); n = i; error = 0; for (i = 0; i < n; i++) { inp = inp_list[i]; INP_RLOCK(inp); if (inp->inp_gencnt <= gencnt) { struct xinpcb xi; bzero(&xi, sizeof(xi)); xi.xi_len = sizeof xi; /* XXX should avoid extra copy */ bcopy(inp, &xi.xi_inp, sizeof *inp); if (inp->inp_socket) sotoxsocket(inp->inp_socket, &xi.xi_socket); INP_RUNLOCK(inp); error = SYSCTL_OUT(req, &xi, sizeof xi); } else INP_RUNLOCK(inp); } INP_INFO_WLOCK(&V_ripcbinfo); for (i = 0; i < n; i++) { inp = inp_list[i]; - INP_WLOCK(inp); - if (!in_pcbrele(inp)) - INP_WUNLOCK(inp); + INP_RLOCK(inp); + if (!in_pcbrele_rlocked(inp)) + INP_RUNLOCK(inp); } INP_INFO_WUNLOCK(&V_ripcbinfo); if (!error) { /* * Give the user an updated idea of our state. If the * generation differs from what we told her before, she knows * that something happened while we were processing this * request, and it might be necessary to retry. */ INP_INFO_RLOCK(&V_ripcbinfo); xig.xig_gen = V_ripcbinfo.ipi_gencnt; xig.xig_sogen = so_gencnt; xig.xig_count = V_ripcbinfo.ipi_count; INP_INFO_RUNLOCK(&V_ripcbinfo); error = SYSCTL_OUT(req, &xig, sizeof xig); } free(inp_list, M_TEMP); return (error); } SYSCTL_PROC(_net_inet_raw, OID_AUTO/*XXX*/, pcblist, CTLTYPE_OPAQUE | CTLFLAG_RD, NULL, 0, rip_pcblist, "S,xinpcb", "List of active raw IP sockets"); #ifdef INET struct pr_usrreqs rip_usrreqs = { .pru_abort = rip_abort, .pru_attach = rip_attach, .pru_bind = rip_bind, .pru_connect = rip_connect, .pru_control = in_control, .pru_detach = rip_detach, .pru_disconnect = rip_disconnect, .pru_peeraddr = in_getpeeraddr, .pru_send = rip_send, .pru_shutdown = rip_shutdown, .pru_sockaddr = in_getsockaddr, .pru_sosetlabel = in_pcbsosetlabel, .pru_close = rip_close, }; #endif /* INET */ Index: head/sys/netinet/siftr.c =================================================================== --- head/sys/netinet/siftr.c (revision 222487) +++ head/sys/netinet/siftr.c (revision 222488) @@ -1,1554 +1,1548 @@ /*- * Copyright (c) 2007-2009 * Swinburne University of Technology, Melbourne, Australia. * Copyright (c) 2009-2010, The FreeBSD Foundation * All rights reserved. * * Portions of this software were developed at the Centre for Advanced * Internet Architectures, Swinburne University of Technology, Melbourne, * Australia by Lawrence Stewart under sponsorship from the FreeBSD Foundation. * * 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 AUTHORS 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 AUTHORS 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. */ /****************************************************** * Statistical Information For TCP Research (SIFTR) * * A FreeBSD kernel module that adds very basic intrumentation to the * TCP stack, allowing internal stats to be recorded to a log file * for experimental, debugging and performance analysis purposes. * * SIFTR was first released in 2007 by James Healy and Lawrence Stewart whilst * working on the NewTCP research project at Swinburne University of * Technology's Centre for Advanced Internet Architectures, Melbourne, * Australia, which was made possible in part by a grant from the Cisco * University Research Program Fund at Community Foundation Silicon Valley. * More details are available at: * http://caia.swin.edu.au/urp/newtcp/ * * Work on SIFTR v1.2.x was sponsored by the FreeBSD Foundation as part of * the "Enhancing the FreeBSD TCP Implementation" project 2008-2009. * More details are available at: * http://www.freebsdfoundation.org/ * http://caia.swin.edu.au/freebsd/etcp09/ * * Lawrence Stewart is the current maintainer, and all contact regarding * SIFTR should be directed to him via email: lastewart@swin.edu.au * * Initial release date: June 2007 * Most recent update: September 2010 ******************************************************/ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef SIFTR_IPV6 #include #include #endif /* SIFTR_IPV6 */ #include /* * Three digit version number refers to X.Y.Z where: * X is the major version number * Y is bumped to mark backwards incompatible changes * Z is bumped to mark backwards compatible changes */ #define V_MAJOR 1 #define V_BACKBREAK 2 #define V_BACKCOMPAT 4 #define MODVERSION __CONCAT(V_MAJOR, __CONCAT(V_BACKBREAK, V_BACKCOMPAT)) #define MODVERSION_STR __XSTRING(V_MAJOR) "." __XSTRING(V_BACKBREAK) "." \ __XSTRING(V_BACKCOMPAT) #define HOOK 0 #define UNHOOK 1 #define SIFTR_EXPECTED_MAX_TCP_FLOWS 65536 #define SYS_NAME "FreeBSD" #define PACKET_TAG_SIFTR 100 #define PACKET_COOKIE_SIFTR 21749576 #define SIFTR_LOG_FILE_MODE 0644 #define SIFTR_DISABLE 0 #define SIFTR_ENABLE 1 /* * Hard upper limit on the length of log messages. Bump this up if you add new * data fields such that the line length could exceed the below value. */ #define MAX_LOG_MSG_LEN 200 /* XXX: Make this a sysctl tunable. */ #define SIFTR_ALQ_BUFLEN (1000*MAX_LOG_MSG_LEN) /* * 1 byte for IP version * IPv4: src/dst IP (4+4) + src/dst port (2+2) = 12 bytes * IPv6: src/dst IP (16+16) + src/dst port (2+2) = 36 bytes */ #ifdef SIFTR_IPV6 #define FLOW_KEY_LEN 37 #else #define FLOW_KEY_LEN 13 #endif #ifdef SIFTR_IPV6 #define SIFTR_IPMODE 6 #else #define SIFTR_IPMODE 4 #endif /* useful macros */ #define CAST_PTR_INT(X) (*((int*)(X))) #define UPPER_SHORT(X) (((X) & 0xFFFF0000) >> 16) #define LOWER_SHORT(X) ((X) & 0x0000FFFF) #define FIRST_OCTET(X) (((X) & 0xFF000000) >> 24) #define SECOND_OCTET(X) (((X) & 0x00FF0000) >> 16) #define THIRD_OCTET(X) (((X) & 0x0000FF00) >> 8) #define FOURTH_OCTET(X) ((X) & 0x000000FF) static MALLOC_DEFINE(M_SIFTR, "siftr", "dynamic memory used by SIFTR"); static MALLOC_DEFINE(M_SIFTR_PKTNODE, "siftr_pktnode", "SIFTR pkt_node struct"); static MALLOC_DEFINE(M_SIFTR_HASHNODE, "siftr_hashnode", "SIFTR flow_hash_node struct"); /* Used as links in the pkt manager queue. */ struct pkt_node { /* Timestamp of pkt as noted in the pfil hook. */ struct timeval tval; /* Direction pkt is travelling; either PFIL_IN or PFIL_OUT. */ uint8_t direction; /* IP version pkt_node relates to; either INP_IPV4 or INP_IPV6. */ uint8_t ipver; /* Hash of the pkt which triggered the log message. */ uint32_t hash; /* Local/foreign IP address. */ #ifdef SIFTR_IPV6 uint32_t ip_laddr[4]; uint32_t ip_faddr[4]; #else uint8_t ip_laddr[4]; uint8_t ip_faddr[4]; #endif /* Local TCP port. */ uint16_t tcp_localport; /* Foreign TCP port. */ uint16_t tcp_foreignport; /* Congestion Window (bytes). */ u_long snd_cwnd; /* Sending Window (bytes). */ u_long snd_wnd; /* Receive Window (bytes). */ u_long rcv_wnd; /* Unused (was: Bandwidth Controlled Window (bytes)). */ u_long snd_bwnd; /* Slow Start Threshold (bytes). */ u_long snd_ssthresh; /* Current state of the TCP FSM. */ int conn_state; /* Max Segment Size (bytes). */ u_int max_seg_size; /* * Smoothed RTT stored as found in the TCP control block * in units of (TCP_RTT_SCALE*hz). */ int smoothed_rtt; /* Is SACK enabled? */ u_char sack_enabled; /* Window scaling for snd window. */ u_char snd_scale; /* Window scaling for recv window. */ u_char rcv_scale; /* TCP control block flags. */ u_int flags; /* Retransmit timeout length. */ int rxt_length; /* Size of the TCP send buffer in bytes. */ u_int snd_buf_hiwater; /* Current num bytes in the send socket buffer. */ u_int snd_buf_cc; /* Size of the TCP receive buffer in bytes. */ u_int rcv_buf_hiwater; /* Current num bytes in the receive socket buffer. */ u_int rcv_buf_cc; /* Number of bytes inflight that we are waiting on ACKs for. */ u_int sent_inflight_bytes; /* Number of segments currently in the reassembly queue. */ int t_segqlen; /* Link to next pkt_node in the list. */ STAILQ_ENTRY(pkt_node) nodes; }; struct flow_hash_node { uint16_t counter; uint8_t key[FLOW_KEY_LEN]; LIST_ENTRY(flow_hash_node) nodes; }; struct siftr_stats { /* # TCP pkts seen by the SIFTR PFIL hooks, including any skipped. */ uint64_t n_in; uint64_t n_out; /* # pkts skipped due to failed malloc calls. */ uint32_t nskip_in_malloc; uint32_t nskip_out_malloc; /* # pkts skipped due to failed mtx acquisition. */ uint32_t nskip_in_mtx; uint32_t nskip_out_mtx; /* # pkts skipped due to failed inpcb lookups. */ uint32_t nskip_in_inpcb; uint32_t nskip_out_inpcb; /* # pkts skipped due to failed tcpcb lookups. */ uint32_t nskip_in_tcpcb; uint32_t nskip_out_tcpcb; /* # pkts skipped due to stack reinjection. */ uint32_t nskip_in_dejavu; uint32_t nskip_out_dejavu; }; static DPCPU_DEFINE(struct siftr_stats, ss); static volatile unsigned int siftr_exit_pkt_manager_thread = 0; static unsigned int siftr_enabled = 0; static unsigned int siftr_pkts_per_log = 1; static unsigned int siftr_generate_hashes = 0; /* static unsigned int siftr_binary_log = 0; */ static char siftr_logfile[PATH_MAX] = "/var/log/siftr.log"; static u_long siftr_hashmask; STAILQ_HEAD(pkthead, pkt_node) pkt_queue = STAILQ_HEAD_INITIALIZER(pkt_queue); LIST_HEAD(listhead, flow_hash_node) *counter_hash; static int wait_for_pkt; static struct alq *siftr_alq = NULL; static struct mtx siftr_pkt_queue_mtx; static struct mtx siftr_pkt_mgr_mtx; static struct thread *siftr_pkt_manager_thr = NULL; /* * pfil.h defines PFIL_IN as 1 and PFIL_OUT as 2, * which we use as an index into this array. */ static char direction[3] = {'\0', 'i','o'}; /* Required function prototypes. */ static int siftr_sysctl_enabled_handler(SYSCTL_HANDLER_ARGS); static int siftr_sysctl_logfile_name_handler(SYSCTL_HANDLER_ARGS); /* Declare the net.inet.siftr sysctl tree and populate it. */ SYSCTL_DECL(_net_inet_siftr); SYSCTL_NODE(_net_inet, OID_AUTO, siftr, CTLFLAG_RW, NULL, "siftr related settings"); SYSCTL_PROC(_net_inet_siftr, OID_AUTO, enabled, CTLTYPE_UINT|CTLFLAG_RW, &siftr_enabled, 0, &siftr_sysctl_enabled_handler, "IU", "switch siftr module operations on/off"); SYSCTL_PROC(_net_inet_siftr, OID_AUTO, logfile, CTLTYPE_STRING|CTLFLAG_RW, &siftr_logfile, sizeof(siftr_logfile), &siftr_sysctl_logfile_name_handler, "A", "file to save siftr log messages to"); SYSCTL_UINT(_net_inet_siftr, OID_AUTO, ppl, CTLFLAG_RW, &siftr_pkts_per_log, 1, "number of packets between generating a log message"); SYSCTL_UINT(_net_inet_siftr, OID_AUTO, genhashes, CTLFLAG_RW, &siftr_generate_hashes, 0, "enable packet hash generation"); /* XXX: TODO SYSCTL_UINT(_net_inet_siftr, OID_AUTO, binary, CTLFLAG_RW, &siftr_binary_log, 0, "write log files in binary instead of ascii"); */ /* Begin functions. */ static void siftr_process_pkt(struct pkt_node * pkt_node) { struct flow_hash_node *hash_node; struct listhead *counter_list; struct siftr_stats *ss; struct ale *log_buf; uint8_t key[FLOW_KEY_LEN]; uint8_t found_match, key_offset; hash_node = NULL; ss = DPCPU_PTR(ss); found_match = 0; key_offset = 1; /* * Create the key that will be used to create a hash index * into our hash table. Our key consists of: * ipversion, localip, localport, foreignip, foreignport */ key[0] = pkt_node->ipver; memcpy(key + key_offset, &pkt_node->ip_laddr, sizeof(pkt_node->ip_laddr)); key_offset += sizeof(pkt_node->ip_laddr); memcpy(key + key_offset, &pkt_node->tcp_localport, sizeof(pkt_node->tcp_localport)); key_offset += sizeof(pkt_node->tcp_localport); memcpy(key + key_offset, &pkt_node->ip_faddr, sizeof(pkt_node->ip_faddr)); key_offset += sizeof(pkt_node->ip_faddr); memcpy(key + key_offset, &pkt_node->tcp_foreignport, sizeof(pkt_node->tcp_foreignport)); counter_list = counter_hash + (hash32_buf(key, sizeof(key), 0) & siftr_hashmask); /* * If the list is not empty i.e. the hash index has * been used by another flow previously. */ if (LIST_FIRST(counter_list) != NULL) { /* * Loop through the hash nodes in the list. * There should normally only be 1 hash node in the list, * except if there have been collisions at the hash index * computed by hash32_buf(). */ LIST_FOREACH(hash_node, counter_list, nodes) { /* * Check if the key for the pkt we are currently * processing is the same as the key stored in the * hash node we are currently processing. * If they are the same, then we've found the * hash node that stores the counter for the flow * the pkt belongs to. */ if (memcmp(hash_node->key, key, sizeof(key)) == 0) { found_match = 1; break; } } } /* If this flow hash hasn't been seen before or we have a collision. */ if (hash_node == NULL || !found_match) { /* Create a new hash node to store the flow's counter. */ hash_node = malloc(sizeof(struct flow_hash_node), M_SIFTR_HASHNODE, M_WAITOK); if (hash_node != NULL) { /* Initialise our new hash node list entry. */ hash_node->counter = 0; memcpy(hash_node->key, key, sizeof(key)); LIST_INSERT_HEAD(counter_list, hash_node, nodes); } else { /* Malloc failed. */ if (pkt_node->direction == PFIL_IN) ss->nskip_in_malloc++; else ss->nskip_out_malloc++; return; } } else if (siftr_pkts_per_log > 1) { /* * Taking the remainder of the counter divided * by the current value of siftr_pkts_per_log * and storing that in counter provides a neat * way to modulate the frequency of log * messages being written to the log file. */ hash_node->counter = (hash_node->counter + 1) % siftr_pkts_per_log; /* * If we have not seen enough packets since the last time * we wrote a log message for this connection, return. */ if (hash_node->counter > 0) return; } log_buf = alq_getn(siftr_alq, MAX_LOG_MSG_LEN, ALQ_WAITOK); if (log_buf == NULL) return; /* Should only happen if the ALQ is shutting down. */ #ifdef SIFTR_IPV6 pkt_node->ip_laddr[3] = ntohl(pkt_node->ip_laddr[3]); pkt_node->ip_faddr[3] = ntohl(pkt_node->ip_faddr[3]); if (pkt_node->ipver == INP_IPV6) { /* IPv6 packet */ pkt_node->ip_laddr[0] = ntohl(pkt_node->ip_laddr[0]); pkt_node->ip_laddr[1] = ntohl(pkt_node->ip_laddr[1]); pkt_node->ip_laddr[2] = ntohl(pkt_node->ip_laddr[2]); pkt_node->ip_faddr[0] = ntohl(pkt_node->ip_faddr[0]); pkt_node->ip_faddr[1] = ntohl(pkt_node->ip_faddr[1]); pkt_node->ip_faddr[2] = ntohl(pkt_node->ip_faddr[2]); /* Construct an IPv6 log message. */ log_buf->ae_bytesused = snprintf(log_buf->ae_data, MAX_LOG_MSG_LEN, "%c,0x%08x,%zd.%06ld,%x:%x:%x:%x:%x:%x:%x:%x,%u,%x:%x:%x:" "%x:%x:%x:%x:%x,%u,%ld,%ld,%ld,%ld,%ld,%u,%u,%u,%u,%u,%u," "%u,%d,%u,%u,%u,%u,%u,%u\n", direction[pkt_node->direction], pkt_node->hash, pkt_node->tval.tv_sec, pkt_node->tval.tv_usec, UPPER_SHORT(pkt_node->ip_laddr[0]), LOWER_SHORT(pkt_node->ip_laddr[0]), UPPER_SHORT(pkt_node->ip_laddr[1]), LOWER_SHORT(pkt_node->ip_laddr[1]), UPPER_SHORT(pkt_node->ip_laddr[2]), LOWER_SHORT(pkt_node->ip_laddr[2]), UPPER_SHORT(pkt_node->ip_laddr[3]), LOWER_SHORT(pkt_node->ip_laddr[3]), ntohs(pkt_node->tcp_localport), UPPER_SHORT(pkt_node->ip_faddr[0]), LOWER_SHORT(pkt_node->ip_faddr[0]), UPPER_SHORT(pkt_node->ip_faddr[1]), LOWER_SHORT(pkt_node->ip_faddr[1]), UPPER_SHORT(pkt_node->ip_faddr[2]), LOWER_SHORT(pkt_node->ip_faddr[2]), UPPER_SHORT(pkt_node->ip_faddr[3]), LOWER_SHORT(pkt_node->ip_faddr[3]), ntohs(pkt_node->tcp_foreignport), pkt_node->snd_ssthresh, pkt_node->snd_cwnd, pkt_node->snd_bwnd, pkt_node->snd_wnd, pkt_node->rcv_wnd, pkt_node->snd_scale, pkt_node->rcv_scale, pkt_node->conn_state, pkt_node->max_seg_size, pkt_node->smoothed_rtt, pkt_node->sack_enabled, pkt_node->flags, pkt_node->rxt_length, pkt_node->snd_buf_hiwater, pkt_node->snd_buf_cc, pkt_node->rcv_buf_hiwater, pkt_node->rcv_buf_cc, pkt_node->sent_inflight_bytes, pkt_node->t_segqlen); } else { /* IPv4 packet */ pkt_node->ip_laddr[0] = FIRST_OCTET(pkt_node->ip_laddr[3]); pkt_node->ip_laddr[1] = SECOND_OCTET(pkt_node->ip_laddr[3]); pkt_node->ip_laddr[2] = THIRD_OCTET(pkt_node->ip_laddr[3]); pkt_node->ip_laddr[3] = FOURTH_OCTET(pkt_node->ip_laddr[3]); pkt_node->ip_faddr[0] = FIRST_OCTET(pkt_node->ip_faddr[3]); pkt_node->ip_faddr[1] = SECOND_OCTET(pkt_node->ip_faddr[3]); pkt_node->ip_faddr[2] = THIRD_OCTET(pkt_node->ip_faddr[3]); pkt_node->ip_faddr[3] = FOURTH_OCTET(pkt_node->ip_faddr[3]); #endif /* SIFTR_IPV6 */ /* Construct an IPv4 log message. */ log_buf->ae_bytesused = snprintf(log_buf->ae_data, MAX_LOG_MSG_LEN, "%c,0x%08x,%jd.%06ld,%u.%u.%u.%u,%u,%u.%u.%u.%u,%u,%ld,%ld," "%ld,%ld,%ld,%u,%u,%u,%u,%u,%u,%u,%d,%u,%u,%u,%u,%u,%u\n", direction[pkt_node->direction], pkt_node->hash, (intmax_t)pkt_node->tval.tv_sec, pkt_node->tval.tv_usec, pkt_node->ip_laddr[0], pkt_node->ip_laddr[1], pkt_node->ip_laddr[2], pkt_node->ip_laddr[3], ntohs(pkt_node->tcp_localport), pkt_node->ip_faddr[0], pkt_node->ip_faddr[1], pkt_node->ip_faddr[2], pkt_node->ip_faddr[3], ntohs(pkt_node->tcp_foreignport), pkt_node->snd_ssthresh, pkt_node->snd_cwnd, pkt_node->snd_bwnd, pkt_node->snd_wnd, pkt_node->rcv_wnd, pkt_node->snd_scale, pkt_node->rcv_scale, pkt_node->conn_state, pkt_node->max_seg_size, pkt_node->smoothed_rtt, pkt_node->sack_enabled, pkt_node->flags, pkt_node->rxt_length, pkt_node->snd_buf_hiwater, pkt_node->snd_buf_cc, pkt_node->rcv_buf_hiwater, pkt_node->rcv_buf_cc, pkt_node->sent_inflight_bytes, pkt_node->t_segqlen); #ifdef SIFTR_IPV6 } #endif alq_post_flags(siftr_alq, log_buf, 0); } static void siftr_pkt_manager_thread(void *arg) { STAILQ_HEAD(pkthead, pkt_node) tmp_pkt_queue = STAILQ_HEAD_INITIALIZER(tmp_pkt_queue); struct pkt_node *pkt_node, *pkt_node_temp; uint8_t draining; draining = 2; mtx_lock(&siftr_pkt_mgr_mtx); /* draining == 0 when queue has been flushed and it's safe to exit. */ while (draining) { /* * Sleep until we are signalled to wake because thread has * been told to exit or until 1 tick has passed. */ mtx_sleep(&wait_for_pkt, &siftr_pkt_mgr_mtx, PWAIT, "pktwait", 1); /* Gain exclusive access to the pkt_node queue. */ mtx_lock(&siftr_pkt_queue_mtx); /* * Move pkt_queue to tmp_pkt_queue, which leaves * pkt_queue empty and ready to receive more pkt_nodes. */ STAILQ_CONCAT(&tmp_pkt_queue, &pkt_queue); /* * We've finished making changes to the list. Unlock it * so the pfil hooks can continue queuing pkt_nodes. */ mtx_unlock(&siftr_pkt_queue_mtx); /* * We can't hold a mutex whilst calling siftr_process_pkt * because ALQ might sleep waiting for buffer space. */ mtx_unlock(&siftr_pkt_mgr_mtx); /* Flush all pkt_nodes to the log file. */ STAILQ_FOREACH_SAFE(pkt_node, &tmp_pkt_queue, nodes, pkt_node_temp) { siftr_process_pkt(pkt_node); STAILQ_REMOVE_HEAD(&tmp_pkt_queue, nodes); free(pkt_node, M_SIFTR_PKTNODE); } KASSERT(STAILQ_EMPTY(&tmp_pkt_queue), ("SIFTR tmp_pkt_queue not empty after flush")); mtx_lock(&siftr_pkt_mgr_mtx); /* * If siftr_exit_pkt_manager_thread gets set during the window * where we are draining the tmp_pkt_queue above, there might * still be pkts in pkt_queue that need to be drained. * Allow one further iteration to occur after * siftr_exit_pkt_manager_thread has been set to ensure * pkt_queue is completely empty before we kill the thread. * * siftr_exit_pkt_manager_thread is set only after the pfil * hooks have been removed, so only 1 extra iteration * is needed to drain the queue. */ if (siftr_exit_pkt_manager_thread) draining--; } mtx_unlock(&siftr_pkt_mgr_mtx); /* Calls wakeup on this thread's struct thread ptr. */ kthread_exit(); } static uint32_t hash_pkt(struct mbuf *m, uint32_t offset) { uint32_t hash; hash = 0; while (m != NULL && offset > m->m_len) { /* * The IP packet payload does not start in this mbuf, so * need to figure out which mbuf it starts in and what offset * into the mbuf's data region the payload starts at. */ offset -= m->m_len; m = m->m_next; } while (m != NULL) { /* Ensure there is data in the mbuf */ if ((m->m_len - offset) > 0) hash = hash32_buf(m->m_data + offset, m->m_len - offset, hash); m = m->m_next; offset = 0; } return (hash); } /* * Check if a given mbuf has the SIFTR mbuf tag. If it does, log the fact that * it's a reinjected packet and return. If it doesn't, tag the mbuf and return. * Return value >0 means the caller should skip processing this mbuf. */ static inline int siftr_chkreinject(struct mbuf *m, int dir, struct siftr_stats *ss) { if (m_tag_locate(m, PACKET_COOKIE_SIFTR, PACKET_TAG_SIFTR, NULL) != NULL) { if (dir == PFIL_IN) ss->nskip_in_dejavu++; else ss->nskip_out_dejavu++; return (1); } else { struct m_tag *tag = m_tag_alloc(PACKET_COOKIE_SIFTR, PACKET_TAG_SIFTR, 0, M_NOWAIT); if (tag == NULL) { if (dir == PFIL_IN) ss->nskip_in_malloc++; else ss->nskip_out_malloc++; return (1); } m_tag_prepend(m, tag); } return (0); } /* * Look up an inpcb for a packet. Return the inpcb pointer if found, or NULL * otherwise. */ static inline struct inpcb * siftr_findinpcb(int ipver, struct ip *ip, struct mbuf *m, uint16_t sport, uint16_t dport, int dir, struct siftr_stats *ss) { struct inpcb *inp; /* We need the tcbinfo lock. */ INP_INFO_UNLOCK_ASSERT(&V_tcbinfo); - INP_INFO_RLOCK(&V_tcbinfo); if (dir == PFIL_IN) inp = (ipver == INP_IPV4 ? - in_pcblookup_hash(&V_tcbinfo, ip->ip_src, sport, ip->ip_dst, - dport, 0, m->m_pkthdr.rcvif) + in_pcblookup(&V_tcbinfo, ip->ip_src, sport, ip->ip_dst, + dport, INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif) : #ifdef SIFTR_IPV6 - in6_pcblookup_hash(&V_tcbinfo, + in6_pcblookup(&V_tcbinfo, &((struct ip6_hdr *)ip)->ip6_src, sport, - &((struct ip6_hdr *)ip)->ip6_dst, dport, 0, + &((struct ip6_hdr *)ip)->ip6_dst, dport, INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif) #else NULL #endif ); else inp = (ipver == INP_IPV4 ? - in_pcblookup_hash(&V_tcbinfo, ip->ip_dst, dport, ip->ip_src, - sport, 0, m->m_pkthdr.rcvif) + in_pcblookup(&V_tcbinfo, ip->ip_dst, dport, ip->ip_src, + sport, INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif) : #ifdef SIFTR_IPV6 - in6_pcblookup_hash(&V_tcbinfo, + in6_pcblookup(&V_tcbinfo, &((struct ip6_hdr *)ip)->ip6_dst, dport, - &((struct ip6_hdr *)ip)->ip6_src, sport, 0, + &((struct ip6_hdr *)ip)->ip6_src, sport, INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif) #else NULL #endif ); /* If we can't find the inpcb, bail. */ if (inp == NULL) { if (dir == PFIL_IN) ss->nskip_in_inpcb++; else ss->nskip_out_inpcb++; - } else { - /* Acquire the inpcb lock. */ - INP_UNLOCK_ASSERT(inp); - INP_RLOCK(inp); } - INP_INFO_RUNLOCK(&V_tcbinfo); return (inp); } static inline void siftr_siftdata(struct pkt_node *pn, struct inpcb *inp, struct tcpcb *tp, int ipver, int dir, int inp_locally_locked) { #ifdef SIFTR_IPV6 if (ipver == INP_IPV4) { pn->ip_laddr[3] = inp->inp_laddr.s_addr; pn->ip_faddr[3] = inp->inp_faddr.s_addr; #else *((uint32_t *)pn->ip_laddr) = inp->inp_laddr.s_addr; *((uint32_t *)pn->ip_faddr) = inp->inp_faddr.s_addr; #endif #ifdef SIFTR_IPV6 } else { pn->ip_laddr[0] = inp->in6p_laddr.s6_addr32[0]; pn->ip_laddr[1] = inp->in6p_laddr.s6_addr32[1]; pn->ip_laddr[2] = inp->in6p_laddr.s6_addr32[2]; pn->ip_laddr[3] = inp->in6p_laddr.s6_addr32[3]; pn->ip_faddr[0] = inp->in6p_faddr.s6_addr32[0]; pn->ip_faddr[1] = inp->in6p_faddr.s6_addr32[1]; pn->ip_faddr[2] = inp->in6p_faddr.s6_addr32[2]; pn->ip_faddr[3] = inp->in6p_faddr.s6_addr32[3]; } #endif pn->tcp_localport = inp->inp_lport; pn->tcp_foreignport = inp->inp_fport; pn->snd_cwnd = tp->snd_cwnd; pn->snd_wnd = tp->snd_wnd; pn->rcv_wnd = tp->rcv_wnd; pn->snd_bwnd = 0; /* Unused, kept for compat. */ pn->snd_ssthresh = tp->snd_ssthresh; pn->snd_scale = tp->snd_scale; pn->rcv_scale = tp->rcv_scale; pn->conn_state = tp->t_state; pn->max_seg_size = tp->t_maxseg; pn->smoothed_rtt = tp->t_srtt; pn->sack_enabled = (tp->t_flags & TF_SACK_PERMIT) != 0; pn->flags = tp->t_flags; pn->rxt_length = tp->t_rxtcur; pn->snd_buf_hiwater = inp->inp_socket->so_snd.sb_hiwat; pn->snd_buf_cc = inp->inp_socket->so_snd.sb_cc; pn->rcv_buf_hiwater = inp->inp_socket->so_rcv.sb_hiwat; pn->rcv_buf_cc = inp->inp_socket->so_rcv.sb_cc; pn->sent_inflight_bytes = tp->snd_max - tp->snd_una; pn->t_segqlen = tp->t_segqlen; /* We've finished accessing the tcb so release the lock. */ if (inp_locally_locked) INP_RUNLOCK(inp); pn->ipver = ipver; pn->direction = dir; /* * Significantly more accurate than using getmicrotime(), but slower! * Gives true microsecond resolution at the expense of a hit to * maximum pps throughput processing when SIFTR is loaded and enabled. */ microtime(&pn->tval); } /* * pfil hook that is called for each IPv4 packet making its way through the * stack in either direction. * The pfil subsystem holds a non-sleepable mutex somewhere when * calling our hook function, so we can't sleep at all. * It's very important to use the M_NOWAIT flag with all function calls * that support it so that they won't sleep, otherwise you get a panic. */ static int siftr_chkpkt(void *arg, struct mbuf **m, struct ifnet *ifp, int dir, struct inpcb *inp) { struct pkt_node *pn; struct ip *ip; struct tcphdr *th; struct tcpcb *tp; struct siftr_stats *ss; unsigned int ip_hl; int inp_locally_locked; inp_locally_locked = 0; ss = DPCPU_PTR(ss); /* * m_pullup is not required here because ip_{input|output} * already do the heavy lifting for us. */ ip = mtod(*m, struct ip *); /* Only continue processing if the packet is TCP. */ if (ip->ip_p != IPPROTO_TCP) goto ret; /* * If a kernel subsystem reinjects packets into the stack, our pfil * hook will be called multiple times for the same packet. * Make sure we only process unique packets. */ if (siftr_chkreinject(*m, dir, ss)) goto ret; if (dir == PFIL_IN) ss->n_in++; else ss->n_out++; /* * Create a tcphdr struct starting at the correct offset * in the IP packet. ip->ip_hl gives the ip header length * in 4-byte words, so multiply it to get the size in bytes. */ ip_hl = (ip->ip_hl << 2); th = (struct tcphdr *)((caddr_t)ip + ip_hl); /* * If the pfil hooks don't provide a pointer to the * inpcb, we need to find it ourselves and lock it. */ if (!inp) { /* Find the corresponding inpcb for this pkt. */ inp = siftr_findinpcb(INP_IPV4, ip, *m, th->th_sport, th->th_dport, dir, ss); if (inp == NULL) goto ret; else inp_locally_locked = 1; } INP_LOCK_ASSERT(inp); /* Find the TCP control block that corresponds with this packet */ tp = intotcpcb(inp); /* * If we can't find the TCP control block (happens occasionaly for a * packet sent during the shutdown phase of a TCP connection), * or we're in the timewait state, bail */ if (tp == NULL || inp->inp_flags & INP_TIMEWAIT) { if (dir == PFIL_IN) ss->nskip_in_tcpcb++; else ss->nskip_out_tcpcb++; goto inp_unlock; } pn = malloc(sizeof(struct pkt_node), M_SIFTR_PKTNODE, M_NOWAIT|M_ZERO); if (pn == NULL) { if (dir == PFIL_IN) ss->nskip_in_malloc++; else ss->nskip_out_malloc++; goto inp_unlock; } siftr_siftdata(pn, inp, tp, INP_IPV4, dir, inp_locally_locked); if (siftr_generate_hashes) { if ((*m)->m_pkthdr.csum_flags & CSUM_TCP) { /* * For outbound packets, the TCP checksum isn't * calculated yet. This is a problem for our packet * hashing as the receiver will calc a different hash * to ours if we don't include the correct TCP checksum * in the bytes being hashed. To work around this * problem, we manually calc the TCP checksum here in * software. We unset the CSUM_TCP flag so the lower * layers don't recalc it. */ (*m)->m_pkthdr.csum_flags &= ~CSUM_TCP; /* * Calculate the TCP checksum in software and assign * to correct TCP header field, which will follow the * packet mbuf down the stack. The trick here is that * tcp_output() sets th->th_sum to the checksum of the * pseudo header for us already. Because of the nature * of the checksumming algorithm, we can sum over the * entire IP payload (i.e. TCP header and data), which * will include the already calculated pseduo header * checksum, thus giving us the complete TCP checksum. * * To put it in simple terms, if checksum(1,2,3,4)=10, * then checksum(1,2,3,4,5) == checksum(10,5). * This property is what allows us to "cheat" and * checksum only the IP payload which has the TCP * th_sum field populated with the pseudo header's * checksum, and not need to futz around checksumming * pseudo header bytes and TCP header/data in one hit. * Refer to RFC 1071 for more info. * * NB: in_cksum_skip(struct mbuf *m, int len, int skip) * in_cksum_skip 2nd argument is NOT the number of * bytes to read from the mbuf at "skip" bytes offset * from the start of the mbuf (very counter intuitive!). * The number of bytes to read is calculated internally * by the function as len-skip i.e. to sum over the IP * payload (TCP header + data) bytes, it is INCORRECT * to call the function like this: * in_cksum_skip(at, ip->ip_len - offset, offset) * Rather, it should be called like this: * in_cksum_skip(at, ip->ip_len, offset) * which means read "ip->ip_len - offset" bytes from * the mbuf cluster "at" at offset "offset" bytes from * the beginning of the "at" mbuf's data pointer. */ th->th_sum = in_cksum_skip(*m, ip->ip_len, ip_hl); } /* * XXX: Having to calculate the checksum in software and then * hash over all bytes is really inefficient. Would be nice to * find a way to create the hash and checksum in the same pass * over the bytes. */ pn->hash = hash_pkt(*m, ip_hl); } mtx_lock(&siftr_pkt_queue_mtx); STAILQ_INSERT_TAIL(&pkt_queue, pn, nodes); mtx_unlock(&siftr_pkt_queue_mtx); goto ret; inp_unlock: if (inp_locally_locked) INP_RUNLOCK(inp); ret: /* Returning 0 ensures pfil will not discard the pkt */ return (0); } #ifdef SIFTR_IPV6 static int siftr_chkpkt6(void *arg, struct mbuf **m, struct ifnet *ifp, int dir, struct inpcb *inp) { struct pkt_node *pn; struct ip6_hdr *ip6; struct tcphdr *th; struct tcpcb *tp; struct siftr_stats *ss; unsigned int ip6_hl; int inp_locally_locked; inp_locally_locked = 0; ss = DPCPU_PTR(ss); /* * m_pullup is not required here because ip6_{input|output} * already do the heavy lifting for us. */ ip6 = mtod(*m, struct ip6_hdr *); /* * Only continue processing if the packet is TCP * XXX: We should follow the next header fields * as shown on Pg 6 RFC 2460, but right now we'll * only check pkts that have no extension headers. */ if (ip6->ip6_nxt != IPPROTO_TCP) goto ret6; /* * If a kernel subsystem reinjects packets into the stack, our pfil * hook will be called multiple times for the same packet. * Make sure we only process unique packets. */ if (siftr_chkreinject(*m, dir, ss)) goto ret6; if (dir == PFIL_IN) ss->n_in++; else ss->n_out++; ip6_hl = sizeof(struct ip6_hdr); /* * Create a tcphdr struct starting at the correct offset * in the ipv6 packet. ip->ip_hl gives the ip header length * in 4-byte words, so multiply it to get the size in bytes. */ th = (struct tcphdr *)((caddr_t)ip6 + ip6_hl); /* * For inbound packets, the pfil hooks don't provide a pointer to the * inpcb, so we need to find it ourselves and lock it. */ if (!inp) { /* Find the corresponding inpcb for this pkt. */ inp = siftr_findinpcb(INP_IPV6, (struct ip *)ip6, *m, th->th_sport, th->th_dport, dir, ss); if (inp == NULL) goto ret6; else inp_locally_locked = 1; } /* Find the TCP control block that corresponds with this packet. */ tp = intotcpcb(inp); /* * If we can't find the TCP control block (happens occasionaly for a * packet sent during the shutdown phase of a TCP connection), * or we're in the timewait state, bail. */ if (tp == NULL || inp->inp_flags & INP_TIMEWAIT) { if (dir == PFIL_IN) ss->nskip_in_tcpcb++; else ss->nskip_out_tcpcb++; goto inp_unlock6; } pn = malloc(sizeof(struct pkt_node), M_SIFTR_PKTNODE, M_NOWAIT|M_ZERO); if (pn == NULL) { if (dir == PFIL_IN) ss->nskip_in_malloc++; else ss->nskip_out_malloc++; goto inp_unlock6; } siftr_siftdata(pn, inp, tp, INP_IPV6, dir, inp_locally_locked); /* XXX: Figure out how to generate hashes for IPv6 packets. */ mtx_lock(&siftr_pkt_queue_mtx); STAILQ_INSERT_TAIL(&pkt_queue, pn, nodes); mtx_unlock(&siftr_pkt_queue_mtx); goto ret6; inp_unlock6: if (inp_locally_locked) INP_RUNLOCK(inp); ret6: /* Returning 0 ensures pfil will not discard the pkt. */ return (0); } #endif /* #ifdef SIFTR_IPV6 */ static int siftr_pfil(int action) { struct pfil_head *pfh_inet; #ifdef SIFTR_IPV6 struct pfil_head *pfh_inet6; #endif VNET_ITERATOR_DECL(vnet_iter); VNET_LIST_RLOCK(); VNET_FOREACH(vnet_iter) { CURVNET_SET(vnet_iter); pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET); #ifdef SIFTR_IPV6 pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6); #endif if (action == HOOK) { pfil_add_hook(siftr_chkpkt, NULL, PFIL_IN | PFIL_OUT | PFIL_WAITOK, pfh_inet); #ifdef SIFTR_IPV6 pfil_add_hook(siftr_chkpkt6, NULL, PFIL_IN | PFIL_OUT | PFIL_WAITOK, pfh_inet6); #endif } else if (action == UNHOOK) { pfil_remove_hook(siftr_chkpkt, NULL, PFIL_IN | PFIL_OUT | PFIL_WAITOK, pfh_inet); #ifdef SIFTR_IPV6 pfil_remove_hook(siftr_chkpkt6, NULL, PFIL_IN | PFIL_OUT | PFIL_WAITOK, pfh_inet6); #endif } CURVNET_RESTORE(); } VNET_LIST_RUNLOCK(); return (0); } static int siftr_sysctl_logfile_name_handler(SYSCTL_HANDLER_ARGS) { struct alq *new_alq; int error; if (req->newptr == NULL) goto skip; /* If old filename and new filename are different. */ if (strncmp(siftr_logfile, (char *)req->newptr, PATH_MAX)) { error = alq_open(&new_alq, req->newptr, curthread->td_ucred, SIFTR_LOG_FILE_MODE, SIFTR_ALQ_BUFLEN, 0); /* Bail if unable to create new alq. */ if (error) return (1); /* * If disabled, siftr_alq == NULL so we simply close * the alq as we've proved it can be opened. * If enabled, close the existing alq and switch the old * for the new. */ if (siftr_alq == NULL) alq_close(new_alq); else { alq_close(siftr_alq); siftr_alq = new_alq; } } skip: return (sysctl_handle_string(oidp, arg1, arg2, req)); } static int siftr_manage_ops(uint8_t action) { struct siftr_stats totalss; struct timeval tval; struct flow_hash_node *counter, *tmp_counter; struct sbuf *s; int i, key_index, ret, error; uint32_t bytes_to_write, total_skipped_pkts; uint16_t lport, fport; uint8_t *key, ipver; #ifdef SIFTR_IPV6 uint32_t laddr[4]; uint32_t faddr[4]; #else uint8_t laddr[4]; uint8_t faddr[4]; #endif error = 0; total_skipped_pkts = 0; /* Init an autosizing sbuf that initially holds 200 chars. */ if ((s = sbuf_new(NULL, NULL, 200, SBUF_AUTOEXTEND)) == NULL) return (-1); if (action == SIFTR_ENABLE) { /* * Create our alq * XXX: We should abort if alq_open fails! */ alq_open(&siftr_alq, siftr_logfile, curthread->td_ucred, SIFTR_LOG_FILE_MODE, SIFTR_ALQ_BUFLEN, 0); STAILQ_INIT(&pkt_queue); DPCPU_ZERO(ss); siftr_exit_pkt_manager_thread = 0; ret = kthread_add(&siftr_pkt_manager_thread, NULL, NULL, &siftr_pkt_manager_thr, RFNOWAIT, 0, "siftr_pkt_manager_thr"); siftr_pfil(HOOK); microtime(&tval); sbuf_printf(s, "enable_time_secs=%jd\tenable_time_usecs=%06ld\t" "siftrver=%s\thz=%u\ttcp_rtt_scale=%u\tsysname=%s\t" "sysver=%u\tipmode=%u\n", (intmax_t)tval.tv_sec, tval.tv_usec, MODVERSION_STR, hz, TCP_RTT_SCALE, SYS_NAME, __FreeBSD_version, SIFTR_IPMODE); sbuf_finish(s); alq_writen(siftr_alq, sbuf_data(s), sbuf_len(s), ALQ_WAITOK); } else if (action == SIFTR_DISABLE && siftr_pkt_manager_thr != NULL) { /* * Remove the pfil hook functions. All threads currently in * the hook functions are allowed to exit before siftr_pfil() * returns. */ siftr_pfil(UNHOOK); /* This will block until the pkt manager thread unlocks it. */ mtx_lock(&siftr_pkt_mgr_mtx); /* Tell the pkt manager thread that it should exit now. */ siftr_exit_pkt_manager_thread = 1; /* * Wake the pkt_manager thread so it realises that * siftr_exit_pkt_manager_thread == 1 and exits gracefully. * The wakeup won't be delivered until we unlock * siftr_pkt_mgr_mtx so this isn't racy. */ wakeup(&wait_for_pkt); /* Wait for the pkt_manager thread to exit. */ mtx_sleep(siftr_pkt_manager_thr, &siftr_pkt_mgr_mtx, PWAIT, "thrwait", 0); siftr_pkt_manager_thr = NULL; mtx_unlock(&siftr_pkt_mgr_mtx); totalss.n_in = DPCPU_VARSUM(ss, n_in); totalss.n_out = DPCPU_VARSUM(ss, n_out); totalss.nskip_in_malloc = DPCPU_VARSUM(ss, nskip_in_malloc); totalss.nskip_out_malloc = DPCPU_VARSUM(ss, nskip_out_malloc); totalss.nskip_in_mtx = DPCPU_VARSUM(ss, nskip_in_mtx); totalss.nskip_out_mtx = DPCPU_VARSUM(ss, nskip_out_mtx); totalss.nskip_in_tcpcb = DPCPU_VARSUM(ss, nskip_in_tcpcb); totalss.nskip_out_tcpcb = DPCPU_VARSUM(ss, nskip_out_tcpcb); totalss.nskip_in_inpcb = DPCPU_VARSUM(ss, nskip_in_inpcb); totalss.nskip_out_inpcb = DPCPU_VARSUM(ss, nskip_out_inpcb); total_skipped_pkts = totalss.nskip_in_malloc + totalss.nskip_out_malloc + totalss.nskip_in_mtx + totalss.nskip_out_mtx + totalss.nskip_in_tcpcb + totalss.nskip_out_tcpcb + totalss.nskip_in_inpcb + totalss.nskip_out_inpcb; microtime(&tval); sbuf_printf(s, "disable_time_secs=%jd\tdisable_time_usecs=%06ld\t" "num_inbound_tcp_pkts=%ju\tnum_outbound_tcp_pkts=%ju\t" "total_tcp_pkts=%ju\tnum_inbound_skipped_pkts_malloc=%u\t" "num_outbound_skipped_pkts_malloc=%u\t" "num_inbound_skipped_pkts_mtx=%u\t" "num_outbound_skipped_pkts_mtx=%u\t" "num_inbound_skipped_pkts_tcpcb=%u\t" "num_outbound_skipped_pkts_tcpcb=%u\t" "num_inbound_skipped_pkts_inpcb=%u\t" "num_outbound_skipped_pkts_inpcb=%u\t" "total_skipped_tcp_pkts=%u\tflow_list=", (intmax_t)tval.tv_sec, tval.tv_usec, (uintmax_t)totalss.n_in, (uintmax_t)totalss.n_out, (uintmax_t)(totalss.n_in + totalss.n_out), totalss.nskip_in_malloc, totalss.nskip_out_malloc, totalss.nskip_in_mtx, totalss.nskip_out_mtx, totalss.nskip_in_tcpcb, totalss.nskip_out_tcpcb, totalss.nskip_in_inpcb, totalss.nskip_out_inpcb, total_skipped_pkts); /* * Iterate over the flow hash, printing a summary of each * flow seen and freeing any malloc'd memory. * The hash consists of an array of LISTs (man 3 queue). */ for (i = 0; i < siftr_hashmask; i++) { LIST_FOREACH_SAFE(counter, counter_hash + i, nodes, tmp_counter) { key = counter->key; key_index = 1; ipver = key[0]; memcpy(laddr, key + key_index, sizeof(laddr)); key_index += sizeof(laddr); memcpy(&lport, key + key_index, sizeof(lport)); key_index += sizeof(lport); memcpy(faddr, key + key_index, sizeof(faddr)); key_index += sizeof(faddr); memcpy(&fport, key + key_index, sizeof(fport)); #ifdef SIFTR_IPV6 laddr[3] = ntohl(laddr[3]); faddr[3] = ntohl(faddr[3]); if (ipver == INP_IPV6) { laddr[0] = ntohl(laddr[0]); laddr[1] = ntohl(laddr[1]); laddr[2] = ntohl(laddr[2]); faddr[0] = ntohl(faddr[0]); faddr[1] = ntohl(faddr[1]); faddr[2] = ntohl(faddr[2]); sbuf_printf(s, "%x:%x:%x:%x:%x:%x:%x:%x;%u-" "%x:%x:%x:%x:%x:%x:%x:%x;%u,", UPPER_SHORT(laddr[0]), LOWER_SHORT(laddr[0]), UPPER_SHORT(laddr[1]), LOWER_SHORT(laddr[1]), UPPER_SHORT(laddr[2]), LOWER_SHORT(laddr[2]), UPPER_SHORT(laddr[3]), LOWER_SHORT(laddr[3]), ntohs(lport), UPPER_SHORT(faddr[0]), LOWER_SHORT(faddr[0]), UPPER_SHORT(faddr[1]), LOWER_SHORT(faddr[1]), UPPER_SHORT(faddr[2]), LOWER_SHORT(faddr[2]), UPPER_SHORT(faddr[3]), LOWER_SHORT(faddr[3]), ntohs(fport)); } else { laddr[0] = FIRST_OCTET(laddr[3]); laddr[1] = SECOND_OCTET(laddr[3]); laddr[2] = THIRD_OCTET(laddr[3]); laddr[3] = FOURTH_OCTET(laddr[3]); faddr[0] = FIRST_OCTET(faddr[3]); faddr[1] = SECOND_OCTET(faddr[3]); faddr[2] = THIRD_OCTET(faddr[3]); faddr[3] = FOURTH_OCTET(faddr[3]); #endif sbuf_printf(s, "%u.%u.%u.%u;%u-%u.%u.%u.%u;%u,", laddr[0], laddr[1], laddr[2], laddr[3], ntohs(lport), faddr[0], faddr[1], faddr[2], faddr[3], ntohs(fport)); #ifdef SIFTR_IPV6 } #endif free(counter, M_SIFTR_HASHNODE); } LIST_INIT(counter_hash + i); } sbuf_printf(s, "\n"); sbuf_finish(s); i = 0; do { bytes_to_write = min(SIFTR_ALQ_BUFLEN, sbuf_len(s)-i); alq_writen(siftr_alq, sbuf_data(s)+i, bytes_to_write, ALQ_WAITOK); i += bytes_to_write; } while (i < sbuf_len(s)); alq_close(siftr_alq); siftr_alq = NULL; } sbuf_delete(s); /* * XXX: Should be using ret to check if any functions fail * and set error appropriately */ return (error); } static int siftr_sysctl_enabled_handler(SYSCTL_HANDLER_ARGS) { if (req->newptr == NULL) goto skip; /* If the value passed in isn't 0 or 1, return an error. */ if (CAST_PTR_INT(req->newptr) != 0 && CAST_PTR_INT(req->newptr) != 1) return (1); /* If we are changing state (0 to 1 or 1 to 0). */ if (CAST_PTR_INT(req->newptr) != siftr_enabled ) if (siftr_manage_ops(CAST_PTR_INT(req->newptr))) { siftr_manage_ops(SIFTR_DISABLE); return (1); } skip: return (sysctl_handle_int(oidp, arg1, arg2, req)); } static void siftr_shutdown_handler(void *arg) { siftr_manage_ops(SIFTR_DISABLE); } /* * Module is being unloaded or machine is shutting down. Take care of cleanup. */ static int deinit_siftr(void) { /* Cleanup. */ siftr_manage_ops(SIFTR_DISABLE); hashdestroy(counter_hash, M_SIFTR, siftr_hashmask); mtx_destroy(&siftr_pkt_queue_mtx); mtx_destroy(&siftr_pkt_mgr_mtx); return (0); } /* * Module has just been loaded into the kernel. */ static int init_siftr(void) { EVENTHANDLER_REGISTER(shutdown_pre_sync, siftr_shutdown_handler, NULL, SHUTDOWN_PRI_FIRST); /* Initialise our flow counter hash table. */ counter_hash = hashinit(SIFTR_EXPECTED_MAX_TCP_FLOWS, M_SIFTR, &siftr_hashmask); mtx_init(&siftr_pkt_queue_mtx, "siftr_pkt_queue_mtx", NULL, MTX_DEF); mtx_init(&siftr_pkt_mgr_mtx, "siftr_pkt_mgr_mtx", NULL, MTX_DEF); /* Print message to the user's current terminal. */ uprintf("\nStatistical Information For TCP Research (SIFTR) %s\n" " http://caia.swin.edu.au/urp/newtcp\n\n", MODVERSION_STR); return (0); } /* * This is the function that is called to load and unload the module. * When the module is loaded, this function is called once with * "what" == MOD_LOAD * When the module is unloaded, this function is called twice with * "what" = MOD_QUIESCE first, followed by "what" = MOD_UNLOAD second * When the system is shut down e.g. CTRL-ALT-DEL or using the shutdown command, * this function is called once with "what" = MOD_SHUTDOWN * When the system is shut down, the handler isn't called until the very end * of the shutdown sequence i.e. after the disks have been synced. */ static int siftr_load_handler(module_t mod, int what, void *arg) { int ret; switch (what) { case MOD_LOAD: ret = init_siftr(); break; case MOD_QUIESCE: case MOD_SHUTDOWN: ret = deinit_siftr(); break; case MOD_UNLOAD: ret = 0; break; default: ret = EINVAL; break; } return (ret); } static moduledata_t siftr_mod = { .name = "siftr", .evhand = siftr_load_handler, }; /* * Param 1: name of the kernel module * Param 2: moduledata_t struct containing info about the kernel module * and the execution entry point for the module * Param 3: From sysinit_sub_id enumeration in /usr/include/sys/kernel.h * Defines the module initialisation order * Param 4: From sysinit_elem_order enumeration in /usr/include/sys/kernel.h * Defines the initialisation order of this kld relative to others * within the same subsystem as defined by param 3 */ DECLARE_MODULE(siftr, siftr_mod, SI_SUB_SMP, SI_ORDER_ANY); MODULE_DEPEND(siftr, alq, 1, 1, 1); MODULE_VERSION(siftr, MODVERSION); Index: head/sys/netinet/tcp_input.c =================================================================== --- head/sys/netinet/tcp_input.c (revision 222487) +++ head/sys/netinet/tcp_input.c (revision 222488) @@ -1,3664 +1,3624 @@ /*- * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1994, 1995 * The Regents of the University of California. All rights reserved. * Copyright (c) 2007-2008,2010 * Swinburne University of Technology, Melbourne, Australia. * Copyright (c) 2009-2010 Lawrence Stewart * Copyright (c) 2010 The FreeBSD Foundation + * Copyright (c) 2010-2011 Juniper Networks, Inc. * All rights reserved. * * Portions of this software were developed at the Centre for Advanced Internet * Architectures, Swinburne University of Technology, by Lawrence Stewart, * James Healy and David Hayes, made possible in part by a grant from the Cisco * University Research Program Fund at Community Foundation Silicon Valley. * * Portions of this software were developed at the Centre for Advanced * Internet Architectures, Swinburne University of Technology, Melbourne, * Australia by David Hayes under sponsorship from the FreeBSD Foundation. * + * Portions of this software were developed by Robert N. M. Watson under + * contract to Juniper Networks, Inc. + * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)tcp_input.c 8.12 (Berkeley) 5/24/95 */ #include __FBSDID("$FreeBSD$"); #include "opt_ipfw.h" /* for ipfw_fwd */ #include "opt_inet.h" #include "opt_inet6.h" #include "opt_ipsec.h" #include "opt_tcpdebug.h" #include #include #include #include #include #include /* for proc0 declaration */ #include #include #include #include #include #include #include #include /* before tcp_seq.h, for tcp_random18() */ #include #include #include #include #define TCPSTATES /* for logging */ #include #include #include #include #include #include #include /* required for icmp_var.h */ #include /* for ICMP_BANDLIM */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef TCPDEBUG #include #endif /* TCPDEBUG */ #ifdef IPSEC #include #include #endif /*IPSEC*/ #include #include const int tcprexmtthresh = 3; VNET_DEFINE(struct tcpstat, tcpstat); SYSCTL_VNET_STRUCT(_net_inet_tcp, TCPCTL_STATS, stats, CTLFLAG_RW, &VNET_NAME(tcpstat), tcpstat, "TCP statistics (struct tcpstat, netinet/tcp_var.h)"); int tcp_log_in_vain = 0; SYSCTL_INT(_net_inet_tcp, OID_AUTO, log_in_vain, CTLFLAG_RW, &tcp_log_in_vain, 0, "Log all incoming TCP segments to closed ports"); VNET_DEFINE(int, blackhole) = 0; #define V_blackhole VNET(blackhole) SYSCTL_VNET_INT(_net_inet_tcp, OID_AUTO, blackhole, CTLFLAG_RW, &VNET_NAME(blackhole), 0, "Do not send RST on segments to closed ports"); VNET_DEFINE(int, tcp_delack_enabled) = 1; SYSCTL_VNET_INT(_net_inet_tcp, OID_AUTO, delayed_ack, CTLFLAG_RW, &VNET_NAME(tcp_delack_enabled), 0, "Delay ACK to try and piggyback it onto a data packet"); VNET_DEFINE(int, drop_synfin) = 0; #define V_drop_synfin VNET(drop_synfin) SYSCTL_VNET_INT(_net_inet_tcp, OID_AUTO, drop_synfin, CTLFLAG_RW, &VNET_NAME(drop_synfin), 0, "Drop TCP packets with SYN+FIN set"); VNET_DEFINE(int, tcp_do_rfc3042) = 1; #define V_tcp_do_rfc3042 VNET(tcp_do_rfc3042) SYSCTL_VNET_INT(_net_inet_tcp, OID_AUTO, rfc3042, CTLFLAG_RW, &VNET_NAME(tcp_do_rfc3042), 0, "Enable RFC 3042 (Limited Transmit)"); VNET_DEFINE(int, tcp_do_rfc3390) = 1; SYSCTL_VNET_INT(_net_inet_tcp, OID_AUTO, rfc3390, CTLFLAG_RW, &VNET_NAME(tcp_do_rfc3390), 0, "Enable RFC 3390 (Increasing TCP's Initial Congestion Window)"); VNET_DEFINE(int, tcp_do_rfc3465) = 1; SYSCTL_VNET_INT(_net_inet_tcp, OID_AUTO, rfc3465, CTLFLAG_RW, &VNET_NAME(tcp_do_rfc3465), 0, "Enable RFC 3465 (Appropriate Byte Counting)"); VNET_DEFINE(int, tcp_abc_l_var) = 2; SYSCTL_VNET_INT(_net_inet_tcp, OID_AUTO, abc_l_var, CTLFLAG_RW, &VNET_NAME(tcp_abc_l_var), 2, "Cap the max cwnd increment during slow-start to this number of segments"); SYSCTL_NODE(_net_inet_tcp, OID_AUTO, ecn, CTLFLAG_RW, 0, "TCP ECN"); VNET_DEFINE(int, tcp_do_ecn) = 0; SYSCTL_VNET_INT(_net_inet_tcp_ecn, OID_AUTO, enable, CTLFLAG_RW, &VNET_NAME(tcp_do_ecn), 0, "TCP ECN support"); VNET_DEFINE(int, tcp_ecn_maxretries) = 1; SYSCTL_VNET_INT(_net_inet_tcp_ecn, OID_AUTO, maxretries, CTLFLAG_RW, &VNET_NAME(tcp_ecn_maxretries), 0, "Max retries before giving up on ECN"); VNET_DEFINE(int, tcp_insecure_rst) = 0; #define V_tcp_insecure_rst VNET(tcp_insecure_rst) SYSCTL_VNET_INT(_net_inet_tcp, OID_AUTO, insecure_rst, CTLFLAG_RW, &VNET_NAME(tcp_insecure_rst), 0, "Follow the old (insecure) criteria for accepting RST packets"); VNET_DEFINE(int, tcp_do_autorcvbuf) = 1; #define V_tcp_do_autorcvbuf VNET(tcp_do_autorcvbuf) SYSCTL_VNET_INT(_net_inet_tcp, OID_AUTO, recvbuf_auto, CTLFLAG_RW, &VNET_NAME(tcp_do_autorcvbuf), 0, "Enable automatic receive buffer sizing"); VNET_DEFINE(int, tcp_autorcvbuf_inc) = 16*1024; #define V_tcp_autorcvbuf_inc VNET(tcp_autorcvbuf_inc) SYSCTL_VNET_INT(_net_inet_tcp, OID_AUTO, recvbuf_inc, CTLFLAG_RW, &VNET_NAME(tcp_autorcvbuf_inc), 0, "Incrementor step size of automatic receive buffer"); VNET_DEFINE(int, tcp_autorcvbuf_max) = 256*1024; #define V_tcp_autorcvbuf_max VNET(tcp_autorcvbuf_max) SYSCTL_VNET_INT(_net_inet_tcp, OID_AUTO, recvbuf_max, CTLFLAG_RW, &VNET_NAME(tcp_autorcvbuf_max), 0, "Max size of automatic receive buffer"); -int tcp_read_locking = 1; -SYSCTL_INT(_net_inet_tcp, OID_AUTO, read_locking, CTLFLAG_RW, - &tcp_read_locking, 0, "Enable read locking strategy"); - VNET_DEFINE(struct inpcbhead, tcb); #define tcb6 tcb /* for KAME src sync over BSD*'s */ VNET_DEFINE(struct inpcbinfo, tcbinfo); static void tcp_dooptions(struct tcpopt *, u_char *, int, int); static void tcp_do_segment(struct mbuf *, struct tcphdr *, struct socket *, struct tcpcb *, int, int, uint8_t, int); static void tcp_dropwithreset(struct mbuf *, struct tcphdr *, struct tcpcb *, int, int); static void tcp_pulloutofband(struct socket *, struct tcphdr *, struct mbuf *, int); static void tcp_xmit_timer(struct tcpcb *, int); static void tcp_newreno_partial_ack(struct tcpcb *, struct tcphdr *); static void inline tcp_fields_to_host(struct tcphdr *); #ifdef TCP_SIGNATURE static void inline tcp_fields_to_net(struct tcphdr *); static int inline tcp_signature_verify_input(struct mbuf *, int, int, int, struct tcpopt *, struct tcphdr *, u_int); #endif static void inline cc_ack_received(struct tcpcb *tp, struct tcphdr *th, uint16_t type); static void inline cc_conn_init(struct tcpcb *tp); static void inline cc_post_recovery(struct tcpcb *tp, struct tcphdr *th); static void inline hhook_run_tcp_est_in(struct tcpcb *tp, struct tcphdr *th, struct tcpopt *to); /* * Kernel module interface for updating tcpstat. The argument is an index * into tcpstat treated as an array of u_long. While this encodes the * general layout of tcpstat into the caller, it doesn't encode its location, * so that future changes to add, for example, per-CPU stats support won't * cause binary compatibility problems for kernel modules. */ void kmod_tcpstat_inc(int statnum) { (*((u_long *)&V_tcpstat + statnum))++; } /* * Wrapper for the TCP established input helper hook. */ static void inline hhook_run_tcp_est_in(struct tcpcb *tp, struct tcphdr *th, struct tcpopt *to) { struct tcp_hhook_data hhook_data; if (V_tcp_hhh[HHOOK_TCP_EST_IN]->hhh_nhooks > 0) { hhook_data.tp = tp; hhook_data.th = th; hhook_data.to = to; hhook_run_hooks(V_tcp_hhh[HHOOK_TCP_EST_IN], &hhook_data, tp->osd); } } /* * CC wrapper hook functions */ static void inline cc_ack_received(struct tcpcb *tp, struct tcphdr *th, uint16_t type) { INP_WLOCK_ASSERT(tp->t_inpcb); tp->ccv->bytes_this_ack = BYTES_THIS_ACK(tp, th); if (tp->snd_cwnd == min(tp->snd_cwnd, tp->snd_wnd)) tp->ccv->flags |= CCF_CWND_LIMITED; else tp->ccv->flags &= ~CCF_CWND_LIMITED; if (type == CC_ACK) { if (tp->snd_cwnd > tp->snd_ssthresh) { tp->t_bytes_acked += min(tp->ccv->bytes_this_ack, V_tcp_abc_l_var * tp->t_maxseg); if (tp->t_bytes_acked >= tp->snd_cwnd) { tp->t_bytes_acked -= tp->snd_cwnd; tp->ccv->flags |= CCF_ABC_SENTAWND; } } else { tp->ccv->flags &= ~CCF_ABC_SENTAWND; tp->t_bytes_acked = 0; } } if (CC_ALGO(tp)->ack_received != NULL) { /* XXXLAS: Find a way to live without this */ tp->ccv->curack = th->th_ack; CC_ALGO(tp)->ack_received(tp->ccv, type); } } static void inline cc_conn_init(struct tcpcb *tp) { struct hc_metrics_lite metrics; struct inpcb *inp = tp->t_inpcb; int rtt; #ifdef INET6 int isipv6 = ((inp->inp_vflag & INP_IPV6) != 0) ? 1 : 0; #endif INP_WLOCK_ASSERT(tp->t_inpcb); tcp_hc_get(&inp->inp_inc, &metrics); if (tp->t_srtt == 0 && (rtt = metrics.rmx_rtt)) { tp->t_srtt = rtt; tp->t_rttbest = tp->t_srtt + TCP_RTT_SCALE; TCPSTAT_INC(tcps_usedrtt); if (metrics.rmx_rttvar) { tp->t_rttvar = metrics.rmx_rttvar; TCPSTAT_INC(tcps_usedrttvar); } else { /* default variation is +- 1 rtt */ tp->t_rttvar = tp->t_srtt * TCP_RTTVAR_SCALE / TCP_RTT_SCALE; } TCPT_RANGESET(tp->t_rxtcur, ((tp->t_srtt >> 2) + tp->t_rttvar) >> 1, tp->t_rttmin, TCPTV_REXMTMAX); } if (metrics.rmx_ssthresh) { /* * There's some sort of gateway or interface * buffer limit on the path. Use this to set * the slow start threshhold, but set the * threshold to no less than 2*mss. */ tp->snd_ssthresh = max(2 * tp->t_maxseg, metrics.rmx_ssthresh); TCPSTAT_INC(tcps_usedssthresh); } /* * Set the slow-start flight size depending on whether this * is a local network or not. * * Extend this so we cache the cwnd too and retrieve it here. * Make cwnd even bigger than RFC3390 suggests but only if we * have previous experience with the remote host. Be careful * not make cwnd bigger than remote receive window or our own * send socket buffer. Maybe put some additional upper bound * on the retrieved cwnd. Should do incremental updates to * hostcache when cwnd collapses so next connection doesn't * overloads the path again. * * XXXAO: Initializing the CWND from the hostcache is broken * and in its current form not RFC conformant. It is disabled * until fixed or removed entirely. * * RFC3390 says only do this if SYN or SYN/ACK didn't got lost. * We currently check only in syncache_socket for that. */ /* #define TCP_METRICS_CWND */ #ifdef TCP_METRICS_CWND if (metrics.rmx_cwnd) tp->snd_cwnd = max(tp->t_maxseg, min(metrics.rmx_cwnd / 2, min(tp->snd_wnd, so->so_snd.sb_hiwat))); else #endif if (V_tcp_do_rfc3390) tp->snd_cwnd = min(4 * tp->t_maxseg, max(2 * tp->t_maxseg, 4380)); #ifdef INET6 else if (isipv6 && in6_localaddr(&inp->in6p_faddr)) tp->snd_cwnd = tp->t_maxseg * V_ss_fltsz_local; #endif #if defined(INET) && defined(INET6) else if (!isipv6 && in_localaddr(inp->inp_faddr)) tp->snd_cwnd = tp->t_maxseg * V_ss_fltsz_local; #endif #ifdef INET else if (in_localaddr(inp->inp_faddr)) tp->snd_cwnd = tp->t_maxseg * V_ss_fltsz_local; #endif else tp->snd_cwnd = tp->t_maxseg * V_ss_fltsz; if (CC_ALGO(tp)->conn_init != NULL) CC_ALGO(tp)->conn_init(tp->ccv); } void inline cc_cong_signal(struct tcpcb *tp, struct tcphdr *th, uint32_t type) { INP_WLOCK_ASSERT(tp->t_inpcb); switch(type) { case CC_NDUPACK: if (!IN_FASTRECOVERY(tp->t_flags)) { tp->snd_recover = tp->snd_max; if (tp->t_flags & TF_ECN_PERMIT) tp->t_flags |= TF_ECN_SND_CWR; } break; case CC_ECN: if (!IN_CONGRECOVERY(tp->t_flags)) { TCPSTAT_INC(tcps_ecn_rcwnd); tp->snd_recover = tp->snd_max; if (tp->t_flags & TF_ECN_PERMIT) tp->t_flags |= TF_ECN_SND_CWR; } break; case CC_RTO: tp->t_dupacks = 0; tp->t_bytes_acked = 0; EXIT_RECOVERY(tp->t_flags); tp->snd_ssthresh = max(2, min(tp->snd_wnd, tp->snd_cwnd) / 2 / tp->t_maxseg) * tp->t_maxseg; tp->snd_cwnd = tp->t_maxseg; break; case CC_RTO_ERR: TCPSTAT_INC(tcps_sndrexmitbad); /* RTO was unnecessary, so reset everything. */ tp->snd_cwnd = tp->snd_cwnd_prev; tp->snd_ssthresh = tp->snd_ssthresh_prev; tp->snd_recover = tp->snd_recover_prev; if (tp->t_flags & TF_WASFRECOVERY) ENTER_FASTRECOVERY(tp->t_flags); if (tp->t_flags & TF_WASCRECOVERY) ENTER_CONGRECOVERY(tp->t_flags); tp->snd_nxt = tp->snd_max; tp->t_flags &= ~TF_PREVVALID; tp->t_badrxtwin = 0; break; } if (CC_ALGO(tp)->cong_signal != NULL) { if (th != NULL) tp->ccv->curack = th->th_ack; CC_ALGO(tp)->cong_signal(tp->ccv, type); } } static void inline cc_post_recovery(struct tcpcb *tp, struct tcphdr *th) { INP_WLOCK_ASSERT(tp->t_inpcb); /* XXXLAS: KASSERT that we're in recovery? */ if (CC_ALGO(tp)->post_recovery != NULL) { tp->ccv->curack = th->th_ack; CC_ALGO(tp)->post_recovery(tp->ccv); } /* XXXLAS: EXIT_RECOVERY ? */ tp->t_bytes_acked = 0; } static inline void tcp_fields_to_host(struct tcphdr *th) { th->th_seq = ntohl(th->th_seq); th->th_ack = ntohl(th->th_ack); th->th_win = ntohs(th->th_win); th->th_urp = ntohs(th->th_urp); } #ifdef TCP_SIGNATURE static inline void tcp_fields_to_net(struct tcphdr *th) { th->th_seq = htonl(th->th_seq); th->th_ack = htonl(th->th_ack); th->th_win = htons(th->th_win); th->th_urp = htons(th->th_urp); } static inline int tcp_signature_verify_input(struct mbuf *m, int off0, int tlen, int optlen, struct tcpopt *to, struct tcphdr *th, u_int tcpbflag) { int ret; tcp_fields_to_net(th); ret = tcp_signature_verify(m, off0, tlen, optlen, to, th, tcpbflag); tcp_fields_to_host(th); return (ret); } #endif /* Neighbor Discovery, Neighbor Unreachability Detection Upper layer hint. */ #ifdef INET6 #define ND6_HINT(tp) \ do { \ if ((tp) && (tp)->t_inpcb && \ ((tp)->t_inpcb->inp_vflag & INP_IPV6) != 0) \ nd6_nud_hint(NULL, NULL, 0); \ } while (0) #else #define ND6_HINT(tp) #endif /* * Indicate whether this ack should be delayed. We can delay the ack if * - there is no delayed ack timer in progress and * - our last ack wasn't a 0-sized window. We never want to delay * the ack that opens up a 0-sized window and * - delayed acks are enabled or * - this is a half-synchronized T/TCP connection. */ #define DELAY_ACK(tp) \ ((!tcp_timer_active(tp, TT_DELACK) && \ (tp->t_flags & TF_RXWIN0SENT) == 0) && \ (V_tcp_delack_enabled || (tp->t_flags & TF_NEEDSYN))) /* * TCP input handling is split into multiple parts: * tcp6_input is a thin wrapper around tcp_input for the extended * ip6_protox[] call format in ip6_input * tcp_input handles primary segment validation, inpcb lookup and * SYN processing on listen sockets * tcp_do_segment processes the ACK and text of the segment for * establishing, established and closing connections */ #ifdef INET6 int tcp6_input(struct mbuf **mp, int *offp, int proto) { struct mbuf *m = *mp; struct in6_ifaddr *ia6; IP6_EXTHDR_CHECK(m, *offp, sizeof(struct tcphdr), IPPROTO_DONE); /* * draft-itojun-ipv6-tcp-to-anycast * better place to put this in? */ ia6 = ip6_getdstifaddr(m); if (ia6 && (ia6->ia6_flags & IN6_IFF_ANYCAST)) { struct ip6_hdr *ip6; ifa_free(&ia6->ia_ifa); ip6 = mtod(m, struct ip6_hdr *); icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_ADDR, (caddr_t)&ip6->ip6_dst - (caddr_t)ip6); return IPPROTO_DONE; } tcp_input(m, *offp); return IPPROTO_DONE; } #endif /* INET6 */ void tcp_input(struct mbuf *m, int off0) { struct tcphdr *th = NULL; struct ip *ip = NULL; #ifdef INET struct ipovly *ipov; #endif struct inpcb *inp = NULL; struct tcpcb *tp = NULL; struct socket *so = NULL; u_char *optp = NULL; int optlen = 0; #ifdef INET int len; #endif int tlen = 0, off; int drop_hdrlen; int thflags; int rstreason = 0; /* For badport_bandlim accounting purposes */ #ifdef TCP_SIGNATURE uint8_t sig_checked = 0; #endif uint8_t iptos = 0; #ifdef INET #ifdef IPFIREWALL_FORWARD struct m_tag *fwd_tag; #endif #endif /* INET */ #ifdef INET6 struct ip6_hdr *ip6 = NULL; int isipv6; #else const void *ip6 = NULL; #if (defined(INET) && defined(IPFIREWALL_FORWARD)) || defined(TCPDEBUG) const int isipv6 = 0; #endif #endif /* INET6 */ struct tcpopt to; /* options in this segment */ char *s = NULL; /* address and port logging */ int ti_locked; #define TI_UNLOCKED 1 -#define TI_RLOCKED 2 -#define TI_WLOCKED 3 +#define TI_WLOCKED 2 #ifdef TCPDEBUG /* * The size of tcp_saveipgen must be the size of the max ip header, * now IPv6. */ u_char tcp_saveipgen[IP6_HDR_LEN]; struct tcphdr tcp_savetcp; short ostate = 0; #endif #ifdef INET6 isipv6 = (mtod(m, struct ip *)->ip_v == 6) ? 1 : 0; #endif to.to_flags = 0; TCPSTAT_INC(tcps_rcvtotal); #ifdef INET6 if (isipv6) { /* IP6_EXTHDR_CHECK() is already done at tcp6_input(). */ ip6 = mtod(m, struct ip6_hdr *); tlen = sizeof(*ip6) + ntohs(ip6->ip6_plen) - off0; if (in6_cksum(m, IPPROTO_TCP, off0, tlen)) { TCPSTAT_INC(tcps_rcvbadsum); goto drop; } th = (struct tcphdr *)((caddr_t)ip6 + off0); /* * Be proactive about unspecified IPv6 address in source. * As we use all-zero to indicate unbounded/unconnected pcb, * unspecified IPv6 address can be used to confuse us. * * Note that packets with unspecified IPv6 destination is * already dropped in ip6_input. */ if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) { /* XXX stat */ goto drop; } } #endif #if defined(INET) && defined(INET6) else #endif #ifdef INET { /* * Get IP and TCP header together in first mbuf. * Note: IP leaves IP header in first mbuf. */ if (off0 > sizeof (struct ip)) { ip_stripoptions(m, (struct mbuf *)0); off0 = sizeof(struct ip); } if (m->m_len < sizeof (struct tcpiphdr)) { if ((m = m_pullup(m, sizeof (struct tcpiphdr))) == NULL) { TCPSTAT_INC(tcps_rcvshort); return; } } ip = mtod(m, struct ip *); ipov = (struct ipovly *)ip; th = (struct tcphdr *)((caddr_t)ip + off0); tlen = ip->ip_len; if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) { if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) th->th_sum = m->m_pkthdr.csum_data; else th->th_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr, htonl(m->m_pkthdr.csum_data + ip->ip_len + IPPROTO_TCP)); th->th_sum ^= 0xffff; #ifdef TCPDEBUG ipov->ih_len = (u_short)tlen; ipov->ih_len = htons(ipov->ih_len); #endif } else { /* * Checksum extended TCP header and data. */ len = sizeof (struct ip) + tlen; bzero(ipov->ih_x1, sizeof(ipov->ih_x1)); ipov->ih_len = (u_short)tlen; ipov->ih_len = htons(ipov->ih_len); th->th_sum = in_cksum(m, len); } if (th->th_sum) { TCPSTAT_INC(tcps_rcvbadsum); goto drop; } /* Re-initialization for later version check */ ip->ip_v = IPVERSION; } #endif /* INET */ #ifdef INET6 if (isipv6) iptos = (ntohl(ip6->ip6_flow) >> 20) & 0xff; #endif #if defined(INET) && defined(INET6) else #endif #ifdef INET iptos = ip->ip_tos; #endif /* * Check that TCP offset makes sense, * pull out TCP options and adjust length. XXX */ off = th->th_off << 2; if (off < sizeof (struct tcphdr) || off > tlen) { TCPSTAT_INC(tcps_rcvbadoff); goto drop; } tlen -= off; /* tlen is used instead of ti->ti_len */ if (off > sizeof (struct tcphdr)) { #ifdef INET6 if (isipv6) { IP6_EXTHDR_CHECK(m, off0, off, ); ip6 = mtod(m, struct ip6_hdr *); th = (struct tcphdr *)((caddr_t)ip6 + off0); } #endif #if defined(INET) && defined(INET6) else #endif #ifdef INET { if (m->m_len < sizeof(struct ip) + off) { if ((m = m_pullup(m, sizeof (struct ip) + off)) == NULL) { TCPSTAT_INC(tcps_rcvshort); return; } ip = mtod(m, struct ip *); ipov = (struct ipovly *)ip; th = (struct tcphdr *)((caddr_t)ip + off0); } } #endif optlen = off - sizeof (struct tcphdr); optp = (u_char *)(th + 1); } thflags = th->th_flags; /* * Convert TCP protocol specific fields to host format. */ tcp_fields_to_host(th); /* * Delay dropping TCP, IP headers, IPv6 ext headers, and TCP options. */ drop_hdrlen = off0 + off; /* - * Locate pcb for segment, which requires a lock on tcbinfo. - * Optimisticaly acquire a global read lock rather than a write lock - * unless header flags necessarily imply a state change. There are - * two cases where we might discover later we need a write lock - * despite the flags: ACKs moving a connection out of the syncache, - * and ACKs for a connection in TIMEWAIT. + * Locate pcb for segment; if we're likely to add or remove a + * connection then first acquire pcbinfo lock. There are two cases + * where we might discover later we need a write lock despite the + * flags: ACKs moving a connection out of the syncache, and ACKs for + * a connection in TIMEWAIT. */ - if ((thflags & (TH_SYN | TH_FIN | TH_RST)) != 0 || - tcp_read_locking == 0) { + if ((thflags & (TH_SYN | TH_FIN | TH_RST)) != 0) { INP_INFO_WLOCK(&V_tcbinfo); ti_locked = TI_WLOCKED; - } else { - INP_INFO_RLOCK(&V_tcbinfo); - ti_locked = TI_RLOCKED; - } + } else + ti_locked = TI_UNLOCKED; findpcb: #ifdef INVARIANTS - if (ti_locked == TI_RLOCKED) - INP_INFO_RLOCK_ASSERT(&V_tcbinfo); - else if (ti_locked == TI_WLOCKED) + if (ti_locked == TI_WLOCKED) { INP_INFO_WLOCK_ASSERT(&V_tcbinfo); - else - panic("%s: findpcb ti_locked %d\n", __func__, ti_locked); + } else { + INP_INFO_UNLOCK_ASSERT(&V_tcbinfo); + } #endif #ifdef INET #ifdef IPFIREWALL_FORWARD /* * Grab info from PACKET_TAG_IPFORWARD tag prepended to the chain. */ fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL); if (fwd_tag != NULL && isipv6 == 0) { /* IPv6 support is not yet */ struct sockaddr_in *next_hop; next_hop = (struct sockaddr_in *)(fwd_tag+1); /* * Transparently forwarded. Pretend to be the destination. * already got one like this? */ - inp = in_pcblookup_hash(&V_tcbinfo, - ip->ip_src, th->th_sport, - ip->ip_dst, th->th_dport, - 0, m->m_pkthdr.rcvif); + inp = in_pcblookup(&V_tcbinfo, ip->ip_src, th->th_sport, + ip->ip_dst, th->th_dport, INPLOOKUP_WLOCKPCB, + m->m_pkthdr.rcvif); if (!inp) { - /* It's new. Try to find the ambushing socket. */ - inp = in_pcblookup_hash(&V_tcbinfo, - ip->ip_src, th->th_sport, - next_hop->sin_addr, - next_hop->sin_port ? - ntohs(next_hop->sin_port) : - th->th_dport, - INPLOOKUP_WILDCARD, - m->m_pkthdr.rcvif); + /* + * It's new. Try to find the ambushing socket. + */ + inp = in_pcblookup(&V_tcbinfo, ip->ip_src, + th->th_sport, next_hop->sin_addr, + next_hop->sin_port ? ntohs(next_hop->sin_port) : + th->th_dport, INPLOOKUP_WILDCARD | + INPLOOKUP_WLOCKPCB, m->m_pkthdr.rcvif); } /* Remove the tag from the packet. We don't need it anymore. */ m_tag_delete(m, fwd_tag); } else #endif /* IPFIREWALL_FORWARD */ #endif /* INET */ { #ifdef INET6 if (isipv6) - inp = in6_pcblookup_hash(&V_tcbinfo, - &ip6->ip6_src, th->th_sport, - &ip6->ip6_dst, th->th_dport, - INPLOOKUP_WILDCARD, - m->m_pkthdr.rcvif); + inp = in6_pcblookup(&V_tcbinfo, &ip6->ip6_src, + th->th_sport, &ip6->ip6_dst, th->th_dport, + INPLOOKUP_WILDCARD | INPLOOKUP_WLOCKPCB, + m->m_pkthdr.rcvif); #endif #if defined(INET) && defined(INET6) else #endif #ifdef INET - inp = in_pcblookup_hash(&V_tcbinfo, - ip->ip_src, th->th_sport, - ip->ip_dst, th->th_dport, - INPLOOKUP_WILDCARD, - m->m_pkthdr.rcvif); + inp = in_pcblookup(&V_tcbinfo, ip->ip_src, + th->th_sport, ip->ip_dst, th->th_dport, + INPLOOKUP_WILDCARD | INPLOOKUP_WLOCKPCB, + m->m_pkthdr.rcvif); #endif } /* * If the INPCB does not exist then all data in the incoming * segment is discarded and an appropriate RST is sent back. * XXX MRT Send RST using which routing table? */ if (inp == NULL) { /* * Log communication attempts to ports that are not * in use. */ if ((tcp_log_in_vain == 1 && (thflags & TH_SYN)) || tcp_log_in_vain == 2) { if ((s = tcp_log_vain(NULL, th, (void *)ip, ip6))) log(LOG_INFO, "%s; %s: Connection attempt " "to closed port\n", s, __func__); } /* * When blackholing do not respond with a RST but * completely ignore the segment and drop it. */ if ((V_blackhole == 1 && (thflags & TH_SYN)) || V_blackhole == 2) goto dropunlock; rstreason = BANDLIM_RST_CLOSEDPORT; goto dropwithreset; } - INP_WLOCK(inp); + INP_WLOCK_ASSERT(inp); if (!(inp->inp_flags & INP_HW_FLOWID) && (m->m_flags & M_FLOWID) && ((inp->inp_socket == NULL) || !(inp->inp_socket->so_options & SO_ACCEPTCONN))) { inp->inp_flags |= INP_HW_FLOWID; inp->inp_flags &= ~INP_SW_FLOWID; inp->inp_flowid = m->m_pkthdr.flowid; } #ifdef IPSEC #ifdef INET6 if (isipv6 && ipsec6_in_reject(m, inp)) { V_ipsec6stat.in_polvio++; goto dropunlock; } else #endif /* INET6 */ if (ipsec4_in_reject(m, inp) != 0) { V_ipsec4stat.in_polvio++; goto dropunlock; } #endif /* IPSEC */ /* * Check the minimum TTL for socket. */ if (inp->inp_ip_minttl != 0) { #ifdef INET6 if (isipv6 && inp->inp_ip_minttl > ip6->ip6_hlim) goto dropunlock; else #endif if (inp->inp_ip_minttl > ip->ip_ttl) goto dropunlock; } /* * A previous connection in TIMEWAIT state is supposed to catch stray * or duplicate segments arriving late. If this segment was a * legitimate new connection attempt the old INPCB gets removed and * we can try again to find a listening socket. * - * At this point, due to earlier optimism, we may hold a read lock on - * the inpcbinfo, rather than a write lock. If so, we need to - * upgrade, or if that fails, acquire a reference on the inpcb, drop - * all locks, acquire a global write lock, and then re-acquire the - * inpcb lock. We may at that point discover that another thread has - * tried to free the inpcb, in which case we need to loop back and - * try to find a new inpcb to deliver to. + * At this point, due to earlier optimism, we may hold only an inpcb + * lock, and not the inpcbinfo write lock. If so, we need to try to + * acquire it, or if that fails, acquire a reference on the inpcb, + * drop all locks, acquire a global write lock, and then re-acquire + * the inpcb lock. We may at that point discover that another thread + * has tried to free the inpcb, in which case we need to loop back + * and try to find a new inpcb to deliver to. + * + * XXXRW: It may be time to rethink timewait locking. */ relocked: if (inp->inp_flags & INP_TIMEWAIT) { - KASSERT(ti_locked == TI_RLOCKED || ti_locked == TI_WLOCKED, - ("%s: INP_TIMEWAIT ti_locked %d", __func__, ti_locked)); - - if (ti_locked == TI_RLOCKED) { - if (INP_INFO_TRY_UPGRADE(&V_tcbinfo) == 0) { + if (ti_locked == TI_UNLOCKED) { + if (INP_INFO_TRY_WLOCK(&V_tcbinfo) == 0) { in_pcbref(inp); INP_WUNLOCK(inp); - INP_INFO_RUNLOCK(&V_tcbinfo); INP_INFO_WLOCK(&V_tcbinfo); ti_locked = TI_WLOCKED; INP_WLOCK(inp); - if (in_pcbrele(inp)) { + if (in_pcbrele_wlocked(inp)) { inp = NULL; goto findpcb; } } else ti_locked = TI_WLOCKED; } INP_INFO_WLOCK_ASSERT(&V_tcbinfo); #ifdef TCP_SIGNATURE tcp_dooptions(&to, optp, optlen, (thflags & TH_SYN) ? TO_SYN : 0); if (sig_checked == 0) { tp = intotcpcb(inp); if (tp == NULL || tp->t_state == TCPS_CLOSED) { rstreason = BANDLIM_RST_CLOSEDPORT; goto dropwithreset; } if (!tcp_signature_verify_input(m, off0, tlen, optlen, &to, th, tp->t_flags)) goto dropunlock; sig_checked = 1; } #else if (thflags & TH_SYN) tcp_dooptions(&to, optp, optlen, TO_SYN); #endif /* * NB: tcp_twcheck unlocks the INP and frees the mbuf. */ if (tcp_twcheck(inp, &to, th, m, tlen)) goto findpcb; INP_INFO_WUNLOCK(&V_tcbinfo); return; } /* * The TCPCB may no longer exist if the connection is winding * down or it is in the CLOSED state. Either way we drop the * segment and send an appropriate response. */ tp = intotcpcb(inp); if (tp == NULL || tp->t_state == TCPS_CLOSED) { rstreason = BANDLIM_RST_CLOSEDPORT; goto dropwithreset; } /* * We've identified a valid inpcb, but it could be that we need an - * inpcbinfo write lock and have only a read lock. In this case, - * attempt to upgrade/relock using the same strategy as the TIMEWAIT - * case above. If we relock, we have to jump back to 'relocked' as - * the connection might now be in TIMEWAIT. + * inpcbinfo write lock but don't hold it. In this case, attempt to + * acquire using the same strategy as the TIMEWAIT case above. If we + * relock, we have to jump back to 'relocked' as the connection might + * now be in TIMEWAIT. */ - if (tp->t_state != TCPS_ESTABLISHED || - (thflags & (TH_SYN | TH_FIN | TH_RST)) != 0 || - tcp_read_locking == 0) { - KASSERT(ti_locked == TI_RLOCKED || ti_locked == TI_WLOCKED, - ("%s: upgrade check ti_locked %d", __func__, ti_locked)); - - if (ti_locked == TI_RLOCKED) { - if (INP_INFO_TRY_UPGRADE(&V_tcbinfo) == 0) { +#ifdef INVARIANTS + if ((thflags & (TH_SYN | TH_FIN | TH_RST)) != 0) + INP_INFO_WLOCK_ASSERT(&V_tcbinfo); +#endif + if (tp->t_state != TCPS_ESTABLISHED) { + if (ti_locked == TI_UNLOCKED) { + if (INP_INFO_TRY_WLOCK(&V_tcbinfo) == 0) { in_pcbref(inp); INP_WUNLOCK(inp); - INP_INFO_RUNLOCK(&V_tcbinfo); INP_INFO_WLOCK(&V_tcbinfo); ti_locked = TI_WLOCKED; INP_WLOCK(inp); - if (in_pcbrele(inp)) { + if (in_pcbrele_wlocked(inp)) { inp = NULL; goto findpcb; } goto relocked; } else ti_locked = TI_WLOCKED; } INP_INFO_WLOCK_ASSERT(&V_tcbinfo); } #ifdef MAC INP_WLOCK_ASSERT(inp); if (mac_inpcb_check_deliver(inp, m)) goto dropunlock; #endif so = inp->inp_socket; KASSERT(so != NULL, ("%s: so == NULL", __func__)); #ifdef TCPDEBUG if (so->so_options & SO_DEBUG) { ostate = tp->t_state; if (isipv6) { #ifdef INET6 bcopy((char *)ip6, (char *)tcp_saveipgen, sizeof(*ip6)); #endif } else bcopy((char *)ip, (char *)tcp_saveipgen, sizeof(*ip)); tcp_savetcp = *th; } #endif /* TCPDEBUG */ /* * When the socket is accepting connections (the INPCB is in LISTEN * state) we look into the SYN cache if this is a new connection - * attempt or the completion of a previous one. + * attempt or the completion of a previous one. Because listen + * sockets are never in TCPS_ESTABLISHED, the V_tcbinfo lock will be + * held in this case. */ if (so->so_options & SO_ACCEPTCONN) { struct in_conninfo inc; KASSERT(tp->t_state == TCPS_LISTEN, ("%s: so accepting but " "tp not listening", __func__)); + INP_INFO_WLOCK_ASSERT(&V_tcbinfo); bzero(&inc, sizeof(inc)); #ifdef INET6 if (isipv6) { inc.inc_flags |= INC_ISIPV6; inc.inc6_faddr = ip6->ip6_src; inc.inc6_laddr = ip6->ip6_dst; } else #endif { inc.inc_faddr = ip->ip_src; inc.inc_laddr = ip->ip_dst; } inc.inc_fport = th->th_sport; inc.inc_lport = th->th_dport; inc.inc_fibnum = so->so_fibnum; /* * Check for an existing connection attempt in syncache if * the flag is only ACK. A successful lookup creates a new * socket appended to the listen queue in SYN_RECEIVED state. */ if ((thflags & (TH_RST|TH_ACK|TH_SYN)) == TH_ACK) { /* * Parse the TCP options here because * syncookies need access to the reflected * timestamp. */ tcp_dooptions(&to, optp, optlen, 0); /* * NB: syncache_expand() doesn't unlock * inp and tcpinfo locks. */ if (!syncache_expand(&inc, &to, th, &so, m)) { /* * No syncache entry or ACK was not * for our SYN/ACK. Send a RST. * NB: syncache did its own logging * of the failure cause. */ rstreason = BANDLIM_RST_OPENPORT; goto dropwithreset; } if (so == NULL) { /* * We completed the 3-way handshake * but could not allocate a socket * either due to memory shortage, * listen queue length limits or * global socket limits. Send RST * or wait and have the remote end * retransmit the ACK for another * try. */ if ((s = tcp_log_addrs(&inc, th, NULL, NULL))) log(LOG_DEBUG, "%s; %s: Listen socket: " "Socket allocation failed due to " "limits or memory shortage, %s\n", s, __func__, V_tcp_sc_rst_sock_fail ? "sending RST" : "try again"); if (V_tcp_sc_rst_sock_fail) { rstreason = BANDLIM_UNLIMITED; goto dropwithreset; } else goto dropunlock; } /* * Socket is created in state SYN_RECEIVED. * Unlock the listen socket, lock the newly * created socket and update the tp variable. */ INP_WUNLOCK(inp); /* listen socket */ inp = sotoinpcb(so); INP_WLOCK(inp); /* new connection */ tp = intotcpcb(inp); KASSERT(tp->t_state == TCPS_SYN_RECEIVED, ("%s: ", __func__)); #ifdef TCP_SIGNATURE if (sig_checked == 0) { tcp_dooptions(&to, optp, optlen, (thflags & TH_SYN) ? TO_SYN : 0); if (!tcp_signature_verify_input(m, off0, tlen, optlen, &to, th, tp->t_flags)) { /* * In SYN_SENT state if it receives an * RST, it is allowed for further * processing. */ if ((thflags & TH_RST) == 0 || (tp->t_state == TCPS_SYN_SENT) == 0) goto dropunlock; } sig_checked = 1; } #endif /* * Process the segment and the data it * contains. tcp_do_segment() consumes * the mbuf chain and unlocks the inpcb. */ tcp_do_segment(m, th, so, tp, drop_hdrlen, tlen, iptos, ti_locked); INP_INFO_UNLOCK_ASSERT(&V_tcbinfo); return; } /* * Segment flag validation for new connection attempts: * * Our (SYN|ACK) response was rejected. * Check with syncache and remove entry to prevent * retransmits. * * NB: syncache_chkrst does its own logging of failure * causes. */ if (thflags & TH_RST) { syncache_chkrst(&inc, th); goto dropunlock; } /* * We can't do anything without SYN. */ if ((thflags & TH_SYN) == 0) { if ((s = tcp_log_addrs(&inc, th, NULL, NULL))) log(LOG_DEBUG, "%s; %s: Listen socket: " "SYN is missing, segment ignored\n", s, __func__); TCPSTAT_INC(tcps_badsyn); goto dropunlock; } /* * (SYN|ACK) is bogus on a listen socket. */ if (thflags & TH_ACK) { if ((s = tcp_log_addrs(&inc, th, NULL, NULL))) log(LOG_DEBUG, "%s; %s: Listen socket: " "SYN|ACK invalid, segment rejected\n", s, __func__); syncache_badack(&inc); /* XXX: Not needed! */ TCPSTAT_INC(tcps_badsyn); rstreason = BANDLIM_RST_OPENPORT; goto dropwithreset; } /* * If the drop_synfin option is enabled, drop all * segments with both the SYN and FIN bits set. * This prevents e.g. nmap from identifying the * TCP/IP stack. * XXX: Poor reasoning. nmap has other methods * and is constantly refining its stack detection * strategies. * XXX: This is a violation of the TCP specification * and was used by RFC1644. */ if ((thflags & TH_FIN) && V_drop_synfin) { if ((s = tcp_log_addrs(&inc, th, NULL, NULL))) log(LOG_DEBUG, "%s; %s: Listen socket: " "SYN|FIN segment ignored (based on " "sysctl setting)\n", s, __func__); TCPSTAT_INC(tcps_badsyn); goto dropunlock; } /* * Segment's flags are (SYN) or (SYN|FIN). * * TH_PUSH, TH_URG, TH_ECE, TH_CWR are ignored * as they do not affect the state of the TCP FSM. * The data pointed to by TH_URG and th_urp is ignored. */ KASSERT((thflags & (TH_RST|TH_ACK)) == 0, ("%s: Listen socket: TH_RST or TH_ACK set", __func__)); KASSERT(thflags & (TH_SYN), ("%s: Listen socket: TH_SYN not set", __func__)); #ifdef INET6 /* * If deprecated address is forbidden, * we do not accept SYN to deprecated interface * address to prevent any new inbound connection from * getting established. * When we do not accept SYN, we send a TCP RST, * with deprecated source address (instead of dropping * it). We compromise it as it is much better for peer * to send a RST, and RST will be the final packet * for the exchange. * * If we do not forbid deprecated addresses, we accept * the SYN packet. RFC2462 does not suggest dropping * SYN in this case. * If we decipher RFC2462 5.5.4, it says like this: * 1. use of deprecated addr with existing * communication is okay - "SHOULD continue to be * used" * 2. use of it with new communication: * (2a) "SHOULD NOT be used if alternate address * with sufficient scope is available" * (2b) nothing mentioned otherwise. * Here we fall into (2b) case as we have no choice in * our source address selection - we must obey the peer. * * The wording in RFC2462 is confusing, and there are * multiple description text for deprecated address * handling - worse, they are not exactly the same. * I believe 5.5.4 is the best one, so we follow 5.5.4. */ if (isipv6 && !V_ip6_use_deprecated) { struct in6_ifaddr *ia6; ia6 = ip6_getdstifaddr(m); if (ia6 != NULL && (ia6->ia6_flags & IN6_IFF_DEPRECATED)) { ifa_free(&ia6->ia_ifa); if ((s = tcp_log_addrs(&inc, th, NULL, NULL))) log(LOG_DEBUG, "%s; %s: Listen socket: " "Connection attempt to deprecated " "IPv6 address rejected\n", s, __func__); rstreason = BANDLIM_RST_OPENPORT; goto dropwithreset; } ifa_free(&ia6->ia_ifa); } #endif /* INET6 */ /* * Basic sanity checks on incoming SYN requests: * Don't respond if the destination is a link layer * broadcast according to RFC1122 4.2.3.10, p. 104. * If it is from this socket it must be forged. * Don't respond if the source or destination is a * global or subnet broad- or multicast address. * Note that it is quite possible to receive unicast * link-layer packets with a broadcast IP address. Use * in_broadcast() to find them. */ if (m->m_flags & (M_BCAST|M_MCAST)) { if ((s = tcp_log_addrs(&inc, th, NULL, NULL))) log(LOG_DEBUG, "%s; %s: Listen socket: " "Connection attempt from broad- or multicast " "link layer address ignored\n", s, __func__); goto dropunlock; } #ifdef INET6 if (isipv6) { if (th->th_dport == th->th_sport && IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &ip6->ip6_src)) { if ((s = tcp_log_addrs(&inc, th, NULL, NULL))) log(LOG_DEBUG, "%s; %s: Listen socket: " "Connection attempt to/from self " "ignored\n", s, __func__); goto dropunlock; } if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) { if ((s = tcp_log_addrs(&inc, th, NULL, NULL))) log(LOG_DEBUG, "%s; %s: Listen socket: " "Connection attempt from/to multicast " "address ignored\n", s, __func__); goto dropunlock; } } #endif #if defined(INET) && defined(INET6) else #endif #ifdef INET { if (th->th_dport == th->th_sport && ip->ip_dst.s_addr == ip->ip_src.s_addr) { if ((s = tcp_log_addrs(&inc, th, NULL, NULL))) log(LOG_DEBUG, "%s; %s: Listen socket: " "Connection attempt from/to self " "ignored\n", s, __func__); goto dropunlock; } if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) || IN_MULTICAST(ntohl(ip->ip_src.s_addr)) || ip->ip_src.s_addr == htonl(INADDR_BROADCAST) || in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) { if ((s = tcp_log_addrs(&inc, th, NULL, NULL))) log(LOG_DEBUG, "%s; %s: Listen socket: " "Connection attempt from/to broad- " "or multicast address ignored\n", s, __func__); goto dropunlock; } } #endif /* * SYN appears to be valid. Create compressed TCP state * for syncache. */ #ifdef TCPDEBUG if (so->so_options & SO_DEBUG) tcp_trace(TA_INPUT, ostate, tp, (void *)tcp_saveipgen, &tcp_savetcp, 0); #endif tcp_dooptions(&to, optp, optlen, TO_SYN); syncache_add(&inc, &to, th, inp, &so, m); /* * Entry added to syncache and mbuf consumed. * Everything already unlocked by syncache_add(). */ INP_INFO_UNLOCK_ASSERT(&V_tcbinfo); return; } #ifdef TCP_SIGNATURE if (sig_checked == 0) { tcp_dooptions(&to, optp, optlen, (thflags & TH_SYN) ? TO_SYN : 0); if (!tcp_signature_verify_input(m, off0, tlen, optlen, &to, th, tp->t_flags)) { /* * In SYN_SENT state if it receives an RST, it is * allowed for further processing. */ if ((thflags & TH_RST) == 0 || (tp->t_state == TCPS_SYN_SENT) == 0) goto dropunlock; } sig_checked = 1; } #endif /* * Segment belongs to a connection in SYN_SENT, ESTABLISHED or later * state. tcp_do_segment() always consumes the mbuf chain, unlocks * the inpcb, and unlocks pcbinfo. */ tcp_do_segment(m, th, so, tp, drop_hdrlen, tlen, iptos, ti_locked); INP_INFO_UNLOCK_ASSERT(&V_tcbinfo); return; dropwithreset: - if (ti_locked == TI_RLOCKED) - INP_INFO_RUNLOCK(&V_tcbinfo); - else if (ti_locked == TI_WLOCKED) + if (ti_locked == TI_WLOCKED) { INP_INFO_WUNLOCK(&V_tcbinfo); - else - panic("%s: dropwithreset ti_locked %d", __func__, ti_locked); - ti_locked = TI_UNLOCKED; + ti_locked = TI_UNLOCKED; + } +#ifdef INVARIANTS + else { + KASSERT(ti_locked == TI_UNLOCKED, ("%s: dropwithreset " + "ti_locked: %d", __func__, ti_locked)); + INP_INFO_UNLOCK_ASSERT(&V_tcbinfo); + } +#endif if (inp != NULL) { tcp_dropwithreset(m, th, tp, tlen, rstreason); INP_WUNLOCK(inp); } else tcp_dropwithreset(m, th, NULL, tlen, rstreason); m = NULL; /* mbuf chain got consumed. */ goto drop; dropunlock: - if (ti_locked == TI_RLOCKED) - INP_INFO_RUNLOCK(&V_tcbinfo); - else if (ti_locked == TI_WLOCKED) + if (ti_locked == TI_WLOCKED) { INP_INFO_WUNLOCK(&V_tcbinfo); - else - panic("%s: dropunlock ti_locked %d", __func__, ti_locked); - ti_locked = TI_UNLOCKED; + ti_locked = TI_UNLOCKED; + } +#ifdef INVARIANTS + else { + KASSERT(ti_locked == TI_UNLOCKED, ("%s: dropunlock " + "ti_locked: %d", __func__, ti_locked)); + INP_INFO_UNLOCK_ASSERT(&V_tcbinfo); + } +#endif if (inp != NULL) INP_WUNLOCK(inp); drop: INP_INFO_UNLOCK_ASSERT(&V_tcbinfo); if (s != NULL) free(s, M_TCPLOG); if (m != NULL) m_freem(m); } static void tcp_do_segment(struct mbuf *m, struct tcphdr *th, struct socket *so, struct tcpcb *tp, int drop_hdrlen, int tlen, uint8_t iptos, int ti_locked) { int thflags, acked, ourfinisacked, needoutput = 0; int rstreason, todrop, win; u_long tiwin; struct tcpopt to; #ifdef TCPDEBUG /* * The size of tcp_saveipgen must be the size of the max ip header, * now IPv6. */ u_char tcp_saveipgen[IP6_HDR_LEN]; struct tcphdr tcp_savetcp; short ostate = 0; #endif thflags = th->th_flags; tp->sackhint.last_sack_ack = 0; /* * If this is either a state-changing packet or current state isn't * established, we require a write lock on tcbinfo. Otherwise, we * allow either a read lock or a write lock, as we may have acquired * a write lock due to a race. * * Require a global write lock for SYN/FIN/RST segments or * non-established connections; otherwise accept either a read or * write lock, as we may have conservatively acquired a write lock in * certain cases in tcp_input() (is this still true?). Currently we * will never enter with no lock, so we try to drop it quickly in the * common pure ack/pure data cases. */ if ((thflags & (TH_SYN | TH_FIN | TH_RST)) != 0 || tp->t_state != TCPS_ESTABLISHED) { KASSERT(ti_locked == TI_WLOCKED, ("%s ti_locked %d for " "SYN/FIN/RST/!EST", __func__, ti_locked)); INP_INFO_WLOCK_ASSERT(&V_tcbinfo); } else { #ifdef INVARIANTS - if (ti_locked == TI_RLOCKED) - INP_INFO_RLOCK_ASSERT(&V_tcbinfo); - else if (ti_locked == TI_WLOCKED) + if (ti_locked == TI_WLOCKED) INP_INFO_WLOCK_ASSERT(&V_tcbinfo); - else - panic("%s: ti_locked %d for EST", __func__, - ti_locked); + else { + KASSERT(ti_locked == TI_UNLOCKED, ("%s: EST " + "ti_locked: %d", __func__, ti_locked)); + INP_INFO_UNLOCK_ASSERT(&V_tcbinfo); + } #endif } INP_WLOCK_ASSERT(tp->t_inpcb); KASSERT(tp->t_state > TCPS_LISTEN, ("%s: TCPS_LISTEN", __func__)); KASSERT(tp->t_state != TCPS_TIME_WAIT, ("%s: TCPS_TIME_WAIT", __func__)); /* * Segment received on connection. * Reset idle time and keep-alive timer. * XXX: This should be done after segment * validation to ignore broken/spoofed segs. */ tp->t_rcvtime = ticks; if (TCPS_HAVEESTABLISHED(tp->t_state)) tcp_timer_activate(tp, TT_KEEP, tcp_keepidle); /* * Unscale the window into a 32-bit value. * For the SYN_SENT state the scale is zero. */ tiwin = th->th_win << tp->snd_scale; /* * TCP ECN processing. */ if (tp->t_flags & TF_ECN_PERMIT) { if (thflags & TH_CWR) tp->t_flags &= ~TF_ECN_SND_ECE; switch (iptos & IPTOS_ECN_MASK) { case IPTOS_ECN_CE: tp->t_flags |= TF_ECN_SND_ECE; TCPSTAT_INC(tcps_ecn_ce); break; case IPTOS_ECN_ECT0: TCPSTAT_INC(tcps_ecn_ect0); break; case IPTOS_ECN_ECT1: TCPSTAT_INC(tcps_ecn_ect1); break; } /* Congestion experienced. */ if (thflags & TH_ECE) { cc_cong_signal(tp, th, CC_ECN); } } /* * Parse options on any incoming segment. */ tcp_dooptions(&to, (u_char *)(th + 1), (th->th_off << 2) - sizeof(struct tcphdr), (thflags & TH_SYN) ? TO_SYN : 0); /* * If echoed timestamp is later than the current time, * fall back to non RFC1323 RTT calculation. Normalize * timestamp if syncookies were used when this connection * was established. */ if ((to.to_flags & TOF_TS) && (to.to_tsecr != 0)) { to.to_tsecr -= tp->ts_offset; if (TSTMP_GT(to.to_tsecr, ticks)) to.to_tsecr = 0; } /* * Process options only when we get SYN/ACK back. The SYN case * for incoming connections is handled in tcp_syncache. * According to RFC1323 the window field in a SYN (i.e., a * or ) segment itself is never scaled. * XXX this is traditional behavior, may need to be cleaned up. */ if (tp->t_state == TCPS_SYN_SENT && (thflags & TH_SYN)) { if ((to.to_flags & TOF_SCALE) && (tp->t_flags & TF_REQ_SCALE)) { tp->t_flags |= TF_RCVD_SCALE; tp->snd_scale = to.to_wscale; } /* * Initial send window. It will be updated with * the next incoming segment to the scaled value. */ tp->snd_wnd = th->th_win; if (to.to_flags & TOF_TS) { tp->t_flags |= TF_RCVD_TSTMP; tp->ts_recent = to.to_tsval; tp->ts_recent_age = ticks; } if (to.to_flags & TOF_MSS) tcp_mss(tp, to.to_mss); if ((tp->t_flags & TF_SACK_PERMIT) && (to.to_flags & TOF_SACKPERM) == 0) tp->t_flags &= ~TF_SACK_PERMIT; } /* * Header prediction: check for the two common cases * of a uni-directional data xfer. If the packet has * no control flags, is in-sequence, the window didn't * change and we're not retransmitting, it's a * candidate. If the length is zero and the ack moved * forward, we're the sender side of the xfer. Just * free the data acked & wake any higher level process * that was blocked waiting for space. If the length * is non-zero and the ack didn't move, we're the * receiver side. If we're getting packets in-order * (the reassembly queue is empty), add the data to * the socket buffer and note that we need a delayed ack. * Make sure that the hidden state-flags are also off. * Since we check for TCPS_ESTABLISHED first, it can only * be TH_NEEDSYN. */ if (tp->t_state == TCPS_ESTABLISHED && th->th_seq == tp->rcv_nxt && (thflags & (TH_SYN|TH_FIN|TH_RST|TH_URG|TH_ACK)) == TH_ACK && tp->snd_nxt == tp->snd_max && tiwin && tiwin == tp->snd_wnd && ((tp->t_flags & (TF_NEEDSYN|TF_NEEDFIN)) == 0) && LIST_EMPTY(&tp->t_segq) && ((to.to_flags & TOF_TS) == 0 || TSTMP_GEQ(to.to_tsval, tp->ts_recent)) ) { /* * If last ACK falls within this segment's sequence numbers, * record the timestamp. * NOTE that the test is modified according to the latest * proposal of the tcplw@cray.com list (Braden 1993/04/26). */ if ((to.to_flags & TOF_TS) != 0 && SEQ_LEQ(th->th_seq, tp->last_ack_sent)) { tp->ts_recent_age = ticks; tp->ts_recent = to.to_tsval; } if (tlen == 0) { if (SEQ_GT(th->th_ack, tp->snd_una) && SEQ_LEQ(th->th_ack, tp->snd_max) && !IN_RECOVERY(tp->t_flags) && (to.to_flags & TOF_SACK) == 0 && TAILQ_EMPTY(&tp->snd_holes)) { /* * This is a pure ack for outstanding data. */ - if (ti_locked == TI_RLOCKED) - INP_INFO_RUNLOCK(&V_tcbinfo); - else if (ti_locked == TI_WLOCKED) + if (ti_locked == TI_WLOCKED) INP_INFO_WUNLOCK(&V_tcbinfo); - else - panic("%s: ti_locked %d on pure ACK", - __func__, ti_locked); ti_locked = TI_UNLOCKED; TCPSTAT_INC(tcps_predack); /* * "bad retransmit" recovery. */ if (tp->t_rxtshift == 1 && tp->t_flags & TF_PREVVALID && (int)(ticks - tp->t_badrxtwin) < 0) { cc_cong_signal(tp, th, CC_RTO_ERR); } /* * Recalculate the transmit timer / rtt. * * Some boxes send broken timestamp replies * during the SYN+ACK phase, ignore * timestamps of 0 or we could calculate a * huge RTT and blow up the retransmit timer. */ if ((to.to_flags & TOF_TS) != 0 && to.to_tsecr) { if (!tp->t_rttlow || tp->t_rttlow > ticks - to.to_tsecr) tp->t_rttlow = ticks - to.to_tsecr; tcp_xmit_timer(tp, ticks - to.to_tsecr + 1); } else if (tp->t_rtttime && SEQ_GT(th->th_ack, tp->t_rtseq)) { if (!tp->t_rttlow || tp->t_rttlow > ticks - tp->t_rtttime) tp->t_rttlow = ticks - tp->t_rtttime; tcp_xmit_timer(tp, ticks - tp->t_rtttime); } acked = BYTES_THIS_ACK(tp, th); /* Run HHOOK_TCP_ESTABLISHED_IN helper hooks. */ hhook_run_tcp_est_in(tp, th, &to); TCPSTAT_INC(tcps_rcvackpack); TCPSTAT_ADD(tcps_rcvackbyte, acked); sbdrop(&so->so_snd, acked); if (SEQ_GT(tp->snd_una, tp->snd_recover) && SEQ_LEQ(th->th_ack, tp->snd_recover)) tp->snd_recover = th->th_ack - 1; /* * Let the congestion control algorithm update * congestion control related information. This * typically means increasing the congestion * window. */ cc_ack_received(tp, th, CC_ACK); tp->snd_una = th->th_ack; /* * Pull snd_wl2 up to prevent seq wrap relative * to th_ack. */ tp->snd_wl2 = th->th_ack; tp->t_dupacks = 0; m_freem(m); ND6_HINT(tp); /* Some progress has been made. */ /* * If all outstanding data are acked, stop * retransmit timer, otherwise restart timer * using current (possibly backed-off) value. * If process is waiting for space, * wakeup/selwakeup/signal. If data * are ready to send, let tcp_output * decide between more output or persist. */ #ifdef TCPDEBUG if (so->so_options & SO_DEBUG) tcp_trace(TA_INPUT, ostate, tp, (void *)tcp_saveipgen, &tcp_savetcp, 0); #endif if (tp->snd_una == tp->snd_max) tcp_timer_activate(tp, TT_REXMT, 0); else if (!tcp_timer_active(tp, TT_PERSIST)) tcp_timer_activate(tp, TT_REXMT, tp->t_rxtcur); sowwakeup(so); if (so->so_snd.sb_cc) (void) tcp_output(tp); goto check_delack; } } else if (th->th_ack == tp->snd_una && tlen <= sbspace(&so->so_rcv)) { int newsize = 0; /* automatic sockbuf scaling */ /* * This is a pure, in-sequence data packet with * nothing on the reassembly queue and we have enough * buffer space to take it. */ - if (ti_locked == TI_RLOCKED) - INP_INFO_RUNLOCK(&V_tcbinfo); - else if (ti_locked == TI_WLOCKED) + if (ti_locked == TI_WLOCKED) INP_INFO_WUNLOCK(&V_tcbinfo); - else - panic("%s: ti_locked %d on pure data " - "segment", __func__, ti_locked); ti_locked = TI_UNLOCKED; /* Clean receiver SACK report if present */ if ((tp->t_flags & TF_SACK_PERMIT) && tp->rcv_numsacks) tcp_clean_sackreport(tp); TCPSTAT_INC(tcps_preddat); tp->rcv_nxt += tlen; /* * Pull snd_wl1 up to prevent seq wrap relative to * th_seq. */ tp->snd_wl1 = th->th_seq; /* * Pull rcv_up up to prevent seq wrap relative to * rcv_nxt. */ tp->rcv_up = tp->rcv_nxt; TCPSTAT_INC(tcps_rcvpack); TCPSTAT_ADD(tcps_rcvbyte, tlen); ND6_HINT(tp); /* Some progress has been made */ #ifdef TCPDEBUG if (so->so_options & SO_DEBUG) tcp_trace(TA_INPUT, ostate, tp, (void *)tcp_saveipgen, &tcp_savetcp, 0); #endif /* * Automatic sizing of receive socket buffer. Often the send * buffer size is not optimally adjusted to the actual network * conditions at hand (delay bandwidth product). Setting the * buffer size too small limits throughput on links with high * bandwidth and high delay (eg. trans-continental/oceanic links). * * On the receive side the socket buffer memory is only rarely * used to any significant extent. This allows us to be much * more aggressive in scaling the receive socket buffer. For * the case that the buffer space is actually used to a large * extent and we run out of kernel memory we can simply drop * the new segments; TCP on the sender will just retransmit it * later. Setting the buffer size too big may only consume too * much kernel memory if the application doesn't read() from * the socket or packet loss or reordering makes use of the * reassembly queue. * * The criteria to step up the receive buffer one notch are: * 1. the number of bytes received during the time it takes * one timestamp to be reflected back to us (the RTT); * 2. received bytes per RTT is within seven eighth of the * current socket buffer size; * 3. receive buffer size has not hit maximal automatic size; * * This algorithm does one step per RTT at most and only if * we receive a bulk stream w/o packet losses or reorderings. * Shrinking the buffer during idle times is not necessary as * it doesn't consume any memory when idle. * * TODO: Only step up if the application is actually serving * the buffer to better manage the socket buffer resources. */ if (V_tcp_do_autorcvbuf && to.to_tsecr && (so->so_rcv.sb_flags & SB_AUTOSIZE)) { if (TSTMP_GT(to.to_tsecr, tp->rfbuf_ts) && to.to_tsecr - tp->rfbuf_ts < hz) { if (tp->rfbuf_cnt > (so->so_rcv.sb_hiwat / 8 * 7) && so->so_rcv.sb_hiwat < V_tcp_autorcvbuf_max) { newsize = min(so->so_rcv.sb_hiwat + V_tcp_autorcvbuf_inc, V_tcp_autorcvbuf_max); } /* Start over with next RTT. */ tp->rfbuf_ts = 0; tp->rfbuf_cnt = 0; } else tp->rfbuf_cnt += tlen; /* add up */ } /* Add data to socket buffer. */ SOCKBUF_LOCK(&so->so_rcv); if (so->so_rcv.sb_state & SBS_CANTRCVMORE) { m_freem(m); } else { /* * Set new socket buffer size. * Give up when limit is reached. */ if (newsize) if (!sbreserve_locked(&so->so_rcv, newsize, so, NULL)) so->so_rcv.sb_flags &= ~SB_AUTOSIZE; m_adj(m, drop_hdrlen); /* delayed header drop */ sbappendstream_locked(&so->so_rcv, m); } /* NB: sorwakeup_locked() does an implicit unlock. */ sorwakeup_locked(so); if (DELAY_ACK(tp)) { tp->t_flags |= TF_DELACK; } else { tp->t_flags |= TF_ACKNOW; tcp_output(tp); } goto check_delack; } } /* * Calculate amount of space in receive window, * and then do TCP input processing. * Receive window is amount of space in rcv queue, * but not less than advertised window. */ win = sbspace(&so->so_rcv); if (win < 0) win = 0; KASSERT(SEQ_GEQ(tp->rcv_adv, tp->rcv_nxt), ("tcp_input negative window: tp %p rcv_nxt %u rcv_adv %u", tp, tp->rcv_nxt, tp->rcv_adv)); tp->rcv_wnd = imax(win, (int)(tp->rcv_adv - tp->rcv_nxt)); /* Reset receive buffer auto scaling when not in bulk receive mode. */ tp->rfbuf_ts = 0; tp->rfbuf_cnt = 0; switch (tp->t_state) { /* * If the state is SYN_RECEIVED: * if seg contains an ACK, but not for our SYN/ACK, send a RST. */ case TCPS_SYN_RECEIVED: if ((thflags & TH_ACK) && (SEQ_LEQ(th->th_ack, tp->snd_una) || SEQ_GT(th->th_ack, tp->snd_max))) { rstreason = BANDLIM_RST_OPENPORT; goto dropwithreset; } break; /* * If the state is SYN_SENT: * if seg contains an ACK, but not for our SYN, drop the input. * if seg contains a RST, then drop the connection. * if seg does not contain SYN, then drop it. * Otherwise this is an acceptable SYN segment * initialize tp->rcv_nxt and tp->irs * if seg contains ack then advance tp->snd_una * if seg contains an ECE and ECN support is enabled, the stream * is ECN capable. * if SYN has been acked change to ESTABLISHED else SYN_RCVD state * arrange for segment to be acked (eventually) * continue processing rest of data/controls, beginning with URG */ case TCPS_SYN_SENT: if ((thflags & TH_ACK) && (SEQ_LEQ(th->th_ack, tp->iss) || SEQ_GT(th->th_ack, tp->snd_max))) { rstreason = BANDLIM_UNLIMITED; goto dropwithreset; } if ((thflags & (TH_ACK|TH_RST)) == (TH_ACK|TH_RST)) tp = tcp_drop(tp, ECONNREFUSED); if (thflags & TH_RST) goto drop; if (!(thflags & TH_SYN)) goto drop; tp->irs = th->th_seq; tcp_rcvseqinit(tp); if (thflags & TH_ACK) { TCPSTAT_INC(tcps_connects); soisconnected(so); #ifdef MAC mac_socketpeer_set_from_mbuf(m, so); #endif /* Do window scaling on this connection? */ if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) == (TF_RCVD_SCALE|TF_REQ_SCALE)) { tp->rcv_scale = tp->request_r_scale; } tp->rcv_adv += imin(tp->rcv_wnd, TCP_MAXWIN << tp->rcv_scale); tp->snd_una++; /* SYN is acked */ /* * If there's data, delay ACK; if there's also a FIN * ACKNOW will be turned on later. */ if (DELAY_ACK(tp) && tlen != 0) tcp_timer_activate(tp, TT_DELACK, tcp_delacktime); else tp->t_flags |= TF_ACKNOW; if ((thflags & TH_ECE) && V_tcp_do_ecn) { tp->t_flags |= TF_ECN_PERMIT; TCPSTAT_INC(tcps_ecn_shs); } /* * Received in SYN_SENT[*] state. * Transitions: * SYN_SENT --> ESTABLISHED * SYN_SENT* --> FIN_WAIT_1 */ tp->t_starttime = ticks; if (tp->t_flags & TF_NEEDFIN) { tp->t_state = TCPS_FIN_WAIT_1; tp->t_flags &= ~TF_NEEDFIN; thflags &= ~TH_SYN; } else { tp->t_state = TCPS_ESTABLISHED; cc_conn_init(tp); tcp_timer_activate(tp, TT_KEEP, tcp_keepidle); } } else { /* * Received initial SYN in SYN-SENT[*] state => * simultaneous open. If segment contains CC option * and there is a cached CC, apply TAO test. * If it succeeds, connection is * half-synchronized. * Otherwise, do 3-way handshake: * SYN-SENT -> SYN-RECEIVED * SYN-SENT* -> SYN-RECEIVED* * If there was no CC option, clear cached CC value. */ tp->t_flags |= (TF_ACKNOW | TF_NEEDSYN); tcp_timer_activate(tp, TT_REXMT, 0); tp->t_state = TCPS_SYN_RECEIVED; } KASSERT(ti_locked == TI_WLOCKED, ("%s: trimthenstep6: " "ti_locked %d", __func__, ti_locked)); INP_INFO_WLOCK_ASSERT(&V_tcbinfo); INP_WLOCK_ASSERT(tp->t_inpcb); /* * Advance th->th_seq to correspond to first data byte. * If data, trim to stay within window, * dropping FIN if necessary. */ th->th_seq++; if (tlen > tp->rcv_wnd) { todrop = tlen - tp->rcv_wnd; m_adj(m, -todrop); tlen = tp->rcv_wnd; thflags &= ~TH_FIN; TCPSTAT_INC(tcps_rcvpackafterwin); TCPSTAT_ADD(tcps_rcvbyteafterwin, todrop); } tp->snd_wl1 = th->th_seq - 1; tp->rcv_up = th->th_seq; /* * Client side of transaction: already sent SYN and data. * If the remote host used T/TCP to validate the SYN, * our data will be ACK'd; if so, enter normal data segment * processing in the middle of step 5, ack processing. * Otherwise, goto step 6. */ if (thflags & TH_ACK) goto process_ACK; goto step6; /* * If the state is LAST_ACK or CLOSING or TIME_WAIT: * do normal processing. * * NB: Leftover from RFC1644 T/TCP. Cases to be reused later. */ case TCPS_LAST_ACK: case TCPS_CLOSING: break; /* continue normal processing */ } /* * States other than LISTEN or SYN_SENT. * First check the RST flag and sequence number since reset segments * are exempt from the timestamp and connection count tests. This * fixes a bug introduced by the Stevens, vol. 2, p. 960 bugfix * below which allowed reset segments in half the sequence space * to fall though and be processed (which gives forged reset * segments with a random sequence number a 50 percent chance of * killing a connection). * Then check timestamp, if present. * Then check the connection count, if present. * Then check that at least some bytes of segment are within * receive window. If segment begins before rcv_nxt, * drop leading data (and SYN); if nothing left, just ack. * * * If the RST bit is set, check the sequence number to see * if this is a valid reset segment. * RFC 793 page 37: * In all states except SYN-SENT, all reset (RST) segments * are validated by checking their SEQ-fields. A reset is * valid if its sequence number is in the window. * Note: this does not take into account delayed ACKs, so * we should test against last_ack_sent instead of rcv_nxt. * The sequence number in the reset segment is normally an * echo of our outgoing acknowlegement numbers, but some hosts * send a reset with the sequence number at the rightmost edge * of our receive window, and we have to handle this case. * Note 2: Paul Watson's paper "Slipping in the Window" has shown * that brute force RST attacks are possible. To combat this, * we use a much stricter check while in the ESTABLISHED state, * only accepting RSTs where the sequence number is equal to * last_ack_sent. In all other states (the states in which a * RST is more likely), the more permissive check is used. * If we have multiple segments in flight, the initial reset * segment sequence numbers will be to the left of last_ack_sent, * but they will eventually catch up. * In any case, it never made sense to trim reset segments to * fit the receive window since RFC 1122 says: * 4.2.2.12 RST Segment: RFC-793 Section 3.4 * * A TCP SHOULD allow a received RST segment to include data. * * DISCUSSION * It has been suggested that a RST segment could contain * ASCII text that encoded and explained the cause of the * RST. No standard has yet been established for such * data. * * If the reset segment passes the sequence number test examine * the state: * SYN_RECEIVED STATE: * If passive open, return to LISTEN state. * If active open, inform user that connection was refused. * ESTABLISHED, FIN_WAIT_1, FIN_WAIT_2, CLOSE_WAIT STATES: * Inform user that connection was reset, and close tcb. * CLOSING, LAST_ACK STATES: * Close the tcb. * TIME_WAIT STATE: * Drop the segment - see Stevens, vol. 2, p. 964 and * RFC 1337. */ if (thflags & TH_RST) { if (SEQ_GEQ(th->th_seq, tp->last_ack_sent - 1) && SEQ_LEQ(th->th_seq, tp->last_ack_sent + tp->rcv_wnd)) { switch (tp->t_state) { case TCPS_SYN_RECEIVED: so->so_error = ECONNREFUSED; goto close; case TCPS_ESTABLISHED: if (V_tcp_insecure_rst == 0 && !(SEQ_GEQ(th->th_seq, tp->rcv_nxt - 1) && SEQ_LEQ(th->th_seq, tp->rcv_nxt + 1)) && !(SEQ_GEQ(th->th_seq, tp->last_ack_sent - 1) && SEQ_LEQ(th->th_seq, tp->last_ack_sent + 1))) { TCPSTAT_INC(tcps_badrst); goto drop; } /* FALLTHROUGH */ case TCPS_FIN_WAIT_1: case TCPS_FIN_WAIT_2: case TCPS_CLOSE_WAIT: so->so_error = ECONNRESET; close: KASSERT(ti_locked == TI_WLOCKED, ("tcp_do_segment: TH_RST 1 ti_locked %d", ti_locked)); INP_INFO_WLOCK_ASSERT(&V_tcbinfo); tp->t_state = TCPS_CLOSED; TCPSTAT_INC(tcps_drops); tp = tcp_close(tp); break; case TCPS_CLOSING: case TCPS_LAST_ACK: KASSERT(ti_locked == TI_WLOCKED, ("tcp_do_segment: TH_RST 2 ti_locked %d", ti_locked)); INP_INFO_WLOCK_ASSERT(&V_tcbinfo); tp = tcp_close(tp); break; } } goto drop; } /* * RFC 1323 PAWS: If we have a timestamp reply on this segment * and it's less than ts_recent, drop it. */ if ((to.to_flags & TOF_TS) != 0 && tp->ts_recent && TSTMP_LT(to.to_tsval, tp->ts_recent)) { /* Check to see if ts_recent is over 24 days old. */ if (ticks - tp->ts_recent_age > TCP_PAWS_IDLE) { /* * Invalidate ts_recent. If this segment updates * ts_recent, the age will be reset later and ts_recent * will get a valid value. If it does not, setting * ts_recent to zero will at least satisfy the * requirement that zero be placed in the timestamp * echo reply when ts_recent isn't valid. The * age isn't reset until we get a valid ts_recent * because we don't want out-of-order segments to be * dropped when ts_recent is old. */ tp->ts_recent = 0; } else { TCPSTAT_INC(tcps_rcvduppack); TCPSTAT_ADD(tcps_rcvdupbyte, tlen); TCPSTAT_INC(tcps_pawsdrop); if (tlen) goto dropafterack; goto drop; } } /* * In the SYN-RECEIVED state, validate that the packet belongs to * this connection before trimming the data to fit the receive * window. Check the sequence number versus IRS since we know * the sequence numbers haven't wrapped. This is a partial fix * for the "LAND" DoS attack. */ if (tp->t_state == TCPS_SYN_RECEIVED && SEQ_LT(th->th_seq, tp->irs)) { rstreason = BANDLIM_RST_OPENPORT; goto dropwithreset; } todrop = tp->rcv_nxt - th->th_seq; if (todrop > 0) { /* * If this is a duplicate SYN for our current connection, * advance over it and pretend and it's not a SYN. */ if (thflags & TH_SYN && th->th_seq == tp->irs) { thflags &= ~TH_SYN; th->th_seq++; if (th->th_urp > 1) th->th_urp--; else thflags &= ~TH_URG; todrop--; } /* * Following if statement from Stevens, vol. 2, p. 960. */ if (todrop > tlen || (todrop == tlen && (thflags & TH_FIN) == 0)) { /* * Any valid FIN must be to the left of the window. * At this point the FIN must be a duplicate or out * of sequence; drop it. */ thflags &= ~TH_FIN; /* * Send an ACK to resynchronize and drop any data. * But keep on processing for RST or ACK. */ tp->t_flags |= TF_ACKNOW; todrop = tlen; TCPSTAT_INC(tcps_rcvduppack); TCPSTAT_ADD(tcps_rcvdupbyte, todrop); } else { TCPSTAT_INC(tcps_rcvpartduppack); TCPSTAT_ADD(tcps_rcvpartdupbyte, todrop); } drop_hdrlen += todrop; /* drop from the top afterwards */ th->th_seq += todrop; tlen -= todrop; if (th->th_urp > todrop) th->th_urp -= todrop; else { thflags &= ~TH_URG; th->th_urp = 0; } } /* * If new data are received on a connection after the * user processes are gone, then RST the other end. */ if ((so->so_state & SS_NOFDREF) && tp->t_state > TCPS_CLOSE_WAIT && tlen) { char *s; KASSERT(ti_locked == TI_WLOCKED, ("%s: SS_NOFDEREF && " "CLOSE_WAIT && tlen ti_locked %d", __func__, ti_locked)); INP_INFO_WLOCK_ASSERT(&V_tcbinfo); if ((s = tcp_log_addrs(&tp->t_inpcb->inp_inc, th, NULL, NULL))) { log(LOG_DEBUG, "%s; %s: %s: Received %d bytes of data after socket " "was closed, sending RST and removing tcpcb\n", s, __func__, tcpstates[tp->t_state], tlen); free(s, M_TCPLOG); } tp = tcp_close(tp); TCPSTAT_INC(tcps_rcvafterclose); rstreason = BANDLIM_UNLIMITED; goto dropwithreset; } /* * If segment ends after window, drop trailing data * (and PUSH and FIN); if nothing left, just ACK. */ todrop = (th->th_seq + tlen) - (tp->rcv_nxt + tp->rcv_wnd); if (todrop > 0) { TCPSTAT_INC(tcps_rcvpackafterwin); if (todrop >= tlen) { TCPSTAT_ADD(tcps_rcvbyteafterwin, tlen); /* * If window is closed can only take segments at * window edge, and have to drop data and PUSH from * incoming segments. Continue processing, but * remember to ack. Otherwise, drop segment * and ack. */ if (tp->rcv_wnd == 0 && th->th_seq == tp->rcv_nxt) { tp->t_flags |= TF_ACKNOW; TCPSTAT_INC(tcps_rcvwinprobe); } else goto dropafterack; } else TCPSTAT_ADD(tcps_rcvbyteafterwin, todrop); m_adj(m, -todrop); tlen -= todrop; thflags &= ~(TH_PUSH|TH_FIN); } /* * If last ACK falls within this segment's sequence numbers, * record its timestamp. * NOTE: * 1) That the test incorporates suggestions from the latest * proposal of the tcplw@cray.com list (Braden 1993/04/26). * 2) That updating only on newer timestamps interferes with * our earlier PAWS tests, so this check should be solely * predicated on the sequence space of this segment. * 3) That we modify the segment boundary check to be * Last.ACK.Sent <= SEG.SEQ + SEG.Len * instead of RFC1323's * Last.ACK.Sent < SEG.SEQ + SEG.Len, * This modified check allows us to overcome RFC1323's * limitations as described in Stevens TCP/IP Illustrated * Vol. 2 p.869. In such cases, we can still calculate the * RTT correctly when RCV.NXT == Last.ACK.Sent. */ if ((to.to_flags & TOF_TS) != 0 && SEQ_LEQ(th->th_seq, tp->last_ack_sent) && SEQ_LEQ(tp->last_ack_sent, th->th_seq + tlen + ((thflags & (TH_SYN|TH_FIN)) != 0))) { tp->ts_recent_age = ticks; tp->ts_recent = to.to_tsval; } /* * If a SYN is in the window, then this is an * error and we send an RST and drop the connection. */ if (thflags & TH_SYN) { KASSERT(ti_locked == TI_WLOCKED, ("tcp_do_segment: TH_SYN ti_locked %d", ti_locked)); INP_INFO_WLOCK_ASSERT(&V_tcbinfo); tp = tcp_drop(tp, ECONNRESET); rstreason = BANDLIM_UNLIMITED; goto drop; } /* * If the ACK bit is off: if in SYN-RECEIVED state or SENDSYN * flag is on (half-synchronized state), then queue data for * later processing; else drop segment and return. */ if ((thflags & TH_ACK) == 0) { if (tp->t_state == TCPS_SYN_RECEIVED || (tp->t_flags & TF_NEEDSYN)) goto step6; else if (tp->t_flags & TF_ACKNOW) goto dropafterack; else goto drop; } /* * Ack processing. */ switch (tp->t_state) { /* * In SYN_RECEIVED state, the ack ACKs our SYN, so enter * ESTABLISHED state and continue processing. * The ACK was checked above. */ case TCPS_SYN_RECEIVED: TCPSTAT_INC(tcps_connects); soisconnected(so); /* Do window scaling? */ if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) == (TF_RCVD_SCALE|TF_REQ_SCALE)) { tp->rcv_scale = tp->request_r_scale; tp->snd_wnd = tiwin; } /* * Make transitions: * SYN-RECEIVED -> ESTABLISHED * SYN-RECEIVED* -> FIN-WAIT-1 */ tp->t_starttime = ticks; if (tp->t_flags & TF_NEEDFIN) { tp->t_state = TCPS_FIN_WAIT_1; tp->t_flags &= ~TF_NEEDFIN; } else { tp->t_state = TCPS_ESTABLISHED; cc_conn_init(tp); tcp_timer_activate(tp, TT_KEEP, tcp_keepidle); } /* * If segment contains data or ACK, will call tcp_reass() * later; if not, do so now to pass queued data to user. */ if (tlen == 0 && (thflags & TH_FIN) == 0) (void) tcp_reass(tp, (struct tcphdr *)0, 0, (struct mbuf *)0); tp->snd_wl1 = th->th_seq - 1; /* FALLTHROUGH */ /* * In ESTABLISHED state: drop duplicate ACKs; ACK out of range * ACKs. If the ack is in the range * tp->snd_una < th->th_ack <= tp->snd_max * then advance tp->snd_una to th->th_ack and drop * data from the retransmission queue. If this ACK reflects * more up to date window information we update our window information. */ case TCPS_ESTABLISHED: case TCPS_FIN_WAIT_1: case TCPS_FIN_WAIT_2: case TCPS_CLOSE_WAIT: case TCPS_CLOSING: case TCPS_LAST_ACK: if (SEQ_GT(th->th_ack, tp->snd_max)) { TCPSTAT_INC(tcps_rcvacktoomuch); goto dropafterack; } if ((tp->t_flags & TF_SACK_PERMIT) && ((to.to_flags & TOF_SACK) || !TAILQ_EMPTY(&tp->snd_holes))) tcp_sack_doack(tp, &to, th->th_ack); /* Run HHOOK_TCP_ESTABLISHED_IN helper hooks. */ hhook_run_tcp_est_in(tp, th, &to); if (SEQ_LEQ(th->th_ack, tp->snd_una)) { if (tlen == 0 && tiwin == tp->snd_wnd) { TCPSTAT_INC(tcps_rcvdupack); /* * If we have outstanding data (other than * a window probe), this is a completely * duplicate ack (ie, window info didn't * change), the ack is the biggest we've * seen and we've seen exactly our rexmt * threshhold of them, assume a packet * has been dropped and retransmit it. * Kludge snd_nxt & the congestion * window so we send only this one * packet. * * We know we're losing at the current * window size so do congestion avoidance * (set ssthresh to half the current window * and pull our congestion window back to * the new ssthresh). * * Dup acks mean that packets have left the * network (they're now cached at the receiver) * so bump cwnd by the amount in the receiver * to keep a constant cwnd packets in the * network. * * When using TCP ECN, notify the peer that * we reduced the cwnd. */ if (!tcp_timer_active(tp, TT_REXMT) || th->th_ack != tp->snd_una) tp->t_dupacks = 0; else if (++tp->t_dupacks > tcprexmtthresh || IN_FASTRECOVERY(tp->t_flags)) { cc_ack_received(tp, th, CC_DUPACK); if ((tp->t_flags & TF_SACK_PERMIT) && IN_FASTRECOVERY(tp->t_flags)) { int awnd; /* * Compute the amount of data in flight first. * We can inject new data into the pipe iff * we have less than 1/2 the original window's * worth of data in flight. */ awnd = (tp->snd_nxt - tp->snd_fack) + tp->sackhint.sack_bytes_rexmit; if (awnd < tp->snd_ssthresh) { tp->snd_cwnd += tp->t_maxseg; if (tp->snd_cwnd > tp->snd_ssthresh) tp->snd_cwnd = tp->snd_ssthresh; } } else tp->snd_cwnd += tp->t_maxseg; (void) tcp_output(tp); goto drop; } else if (tp->t_dupacks == tcprexmtthresh) { tcp_seq onxt = tp->snd_nxt; /* * If we're doing sack, check to * see if we're already in sack * recovery. If we're not doing sack, * check to see if we're in newreno * recovery. */ if (tp->t_flags & TF_SACK_PERMIT) { if (IN_FASTRECOVERY(tp->t_flags)) { tp->t_dupacks = 0; break; } } else { if (SEQ_LEQ(th->th_ack, tp->snd_recover)) { tp->t_dupacks = 0; break; } } /* Congestion signal before ack. */ cc_cong_signal(tp, th, CC_NDUPACK); cc_ack_received(tp, th, CC_DUPACK); tcp_timer_activate(tp, TT_REXMT, 0); tp->t_rtttime = 0; if (tp->t_flags & TF_SACK_PERMIT) { TCPSTAT_INC( tcps_sack_recovery_episode); tp->sack_newdata = tp->snd_nxt; tp->snd_cwnd = tp->t_maxseg; (void) tcp_output(tp); goto drop; } tp->snd_nxt = th->th_ack; tp->snd_cwnd = tp->t_maxseg; (void) tcp_output(tp); KASSERT(tp->snd_limited <= 2, ("%s: tp->snd_limited too big", __func__)); tp->snd_cwnd = tp->snd_ssthresh + tp->t_maxseg * (tp->t_dupacks - tp->snd_limited); if (SEQ_GT(onxt, tp->snd_nxt)) tp->snd_nxt = onxt; goto drop; } else if (V_tcp_do_rfc3042) { cc_ack_received(tp, th, CC_DUPACK); u_long oldcwnd = tp->snd_cwnd; tcp_seq oldsndmax = tp->snd_max; u_int sent; KASSERT(tp->t_dupacks == 1 || tp->t_dupacks == 2, ("%s: dupacks not 1 or 2", __func__)); if (tp->t_dupacks == 1) tp->snd_limited = 0; tp->snd_cwnd = (tp->snd_nxt - tp->snd_una) + (tp->t_dupacks - tp->snd_limited) * tp->t_maxseg; (void) tcp_output(tp); sent = tp->snd_max - oldsndmax; if (sent > tp->t_maxseg) { KASSERT((tp->t_dupacks == 2 && tp->snd_limited == 0) || (sent == tp->t_maxseg + 1 && tp->t_flags & TF_SENTFIN), ("%s: sent too much", __func__)); tp->snd_limited = 2; } else if (sent > 0) ++tp->snd_limited; tp->snd_cwnd = oldcwnd; goto drop; } } else tp->t_dupacks = 0; break; } KASSERT(SEQ_GT(th->th_ack, tp->snd_una), ("%s: th_ack <= snd_una", __func__)); /* * If the congestion window was inflated to account * for the other side's cached packets, retract it. */ if (IN_FASTRECOVERY(tp->t_flags)) { if (SEQ_LT(th->th_ack, tp->snd_recover)) { if (tp->t_flags & TF_SACK_PERMIT) tcp_sack_partialack(tp, th); else tcp_newreno_partial_ack(tp, th); } else cc_post_recovery(tp, th); } tp->t_dupacks = 0; /* * If we reach this point, ACK is not a duplicate, * i.e., it ACKs something we sent. */ if (tp->t_flags & TF_NEEDSYN) { /* * T/TCP: Connection was half-synchronized, and our * SYN has been ACK'd (so connection is now fully * synchronized). Go to non-starred state, * increment snd_una for ACK of SYN, and check if * we can do window scaling. */ tp->t_flags &= ~TF_NEEDSYN; tp->snd_una++; /* Do window scaling? */ if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) == (TF_RCVD_SCALE|TF_REQ_SCALE)) { tp->rcv_scale = tp->request_r_scale; /* Send window already scaled. */ } } process_ACK: - INP_INFO_LOCK_ASSERT(&V_tcbinfo); - KASSERT(ti_locked == TI_RLOCKED || ti_locked == TI_WLOCKED, - ("tcp_input: process_ACK ti_locked %d", ti_locked)); INP_WLOCK_ASSERT(tp->t_inpcb); acked = BYTES_THIS_ACK(tp, th); TCPSTAT_INC(tcps_rcvackpack); TCPSTAT_ADD(tcps_rcvackbyte, acked); /* * If we just performed our first retransmit, and the ACK * arrives within our recovery window, then it was a mistake * to do the retransmit in the first place. Recover our * original cwnd and ssthresh, and proceed to transmit where * we left off. */ if (tp->t_rxtshift == 1 && tp->t_flags & TF_PREVVALID && (int)(ticks - tp->t_badrxtwin) < 0) cc_cong_signal(tp, th, CC_RTO_ERR); /* * If we have a timestamp reply, update smoothed * round trip time. If no timestamp is present but * transmit timer is running and timed sequence * number was acked, update smoothed round trip time. * Since we now have an rtt measurement, cancel the * timer backoff (cf., Phil Karn's retransmit alg.). * Recompute the initial retransmit timer. * * Some boxes send broken timestamp replies * during the SYN+ACK phase, ignore * timestamps of 0 or we could calculate a * huge RTT and blow up the retransmit timer. */ if ((to.to_flags & TOF_TS) != 0 && to.to_tsecr) { if (!tp->t_rttlow || tp->t_rttlow > ticks - to.to_tsecr) tp->t_rttlow = ticks - to.to_tsecr; tcp_xmit_timer(tp, ticks - to.to_tsecr + 1); } else if (tp->t_rtttime && SEQ_GT(th->th_ack, tp->t_rtseq)) { if (!tp->t_rttlow || tp->t_rttlow > ticks - tp->t_rtttime) tp->t_rttlow = ticks - tp->t_rtttime; tcp_xmit_timer(tp, ticks - tp->t_rtttime); } /* * If all outstanding data is acked, stop retransmit * timer and remember to restart (more output or persist). * If there is more data to be acked, restart retransmit * timer, using current (possibly backed-off) value. */ if (th->th_ack == tp->snd_max) { tcp_timer_activate(tp, TT_REXMT, 0); needoutput = 1; } else if (!tcp_timer_active(tp, TT_PERSIST)) tcp_timer_activate(tp, TT_REXMT, tp->t_rxtcur); /* * If no data (only SYN) was ACK'd, * skip rest of ACK processing. */ if (acked == 0) goto step6; /* * Let the congestion control algorithm update congestion * control related information. This typically means increasing * the congestion window. */ cc_ack_received(tp, th, CC_ACK); SOCKBUF_LOCK(&so->so_snd); if (acked > so->so_snd.sb_cc) { tp->snd_wnd -= so->so_snd.sb_cc; sbdrop_locked(&so->so_snd, (int)so->so_snd.sb_cc); ourfinisacked = 1; } else { sbdrop_locked(&so->so_snd, acked); tp->snd_wnd -= acked; ourfinisacked = 0; } /* NB: sowwakeup_locked() does an implicit unlock. */ sowwakeup_locked(so); /* Detect una wraparound. */ if (!IN_RECOVERY(tp->t_flags) && SEQ_GT(tp->snd_una, tp->snd_recover) && SEQ_LEQ(th->th_ack, tp->snd_recover)) tp->snd_recover = th->th_ack - 1; /* XXXLAS: Can this be moved up into cc_post_recovery? */ if (IN_RECOVERY(tp->t_flags) && SEQ_GEQ(th->th_ack, tp->snd_recover)) { EXIT_RECOVERY(tp->t_flags); } tp->snd_una = th->th_ack; if (tp->t_flags & TF_SACK_PERMIT) { if (SEQ_GT(tp->snd_una, tp->snd_recover)) tp->snd_recover = tp->snd_una; } if (SEQ_LT(tp->snd_nxt, tp->snd_una)) tp->snd_nxt = tp->snd_una; switch (tp->t_state) { /* * In FIN_WAIT_1 STATE in addition to the processing * for the ESTABLISHED state if our FIN is now acknowledged * then enter FIN_WAIT_2. */ case TCPS_FIN_WAIT_1: if (ourfinisacked) { /* * If we can't receive any more * data, then closing user can proceed. * Starting the timer is contrary to the * specification, but if we don't get a FIN * we'll hang forever. * * XXXjl: * we should release the tp also, and use a * compressed state. */ if (so->so_rcv.sb_state & SBS_CANTRCVMORE) { int timeout; soisdisconnected(so); timeout = (tcp_fast_finwait2_recycle) ? tcp_finwait2_timeout : tcp_maxidle; tcp_timer_activate(tp, TT_2MSL, timeout); } tp->t_state = TCPS_FIN_WAIT_2; } break; /* * In CLOSING STATE in addition to the processing for * the ESTABLISHED state if the ACK acknowledges our FIN * then enter the TIME-WAIT state, otherwise ignore * the segment. */ case TCPS_CLOSING: if (ourfinisacked) { INP_INFO_WLOCK_ASSERT(&V_tcbinfo); tcp_twstart(tp); INP_INFO_WUNLOCK(&V_tcbinfo); m_freem(m); return; } break; /* * In LAST_ACK, we may still be waiting for data to drain * and/or to be acked, as well as for the ack of our FIN. * If our FIN is now acknowledged, delete the TCB, * enter the closed state and return. */ case TCPS_LAST_ACK: if (ourfinisacked) { INP_INFO_WLOCK_ASSERT(&V_tcbinfo); tp = tcp_close(tp); goto drop; } break; } } step6: - INP_INFO_LOCK_ASSERT(&V_tcbinfo); - KASSERT(ti_locked == TI_RLOCKED || ti_locked == TI_WLOCKED, - ("tcp_do_segment: step6 ti_locked %d", ti_locked)); INP_WLOCK_ASSERT(tp->t_inpcb); /* * Update window information. * Don't look at window if no ACK: TAC's send garbage on first SYN. */ if ((thflags & TH_ACK) && (SEQ_LT(tp->snd_wl1, th->th_seq) || (tp->snd_wl1 == th->th_seq && (SEQ_LT(tp->snd_wl2, th->th_ack) || (tp->snd_wl2 == th->th_ack && tiwin > tp->snd_wnd))))) { /* keep track of pure window updates */ if (tlen == 0 && tp->snd_wl2 == th->th_ack && tiwin > tp->snd_wnd) TCPSTAT_INC(tcps_rcvwinupd); tp->snd_wnd = tiwin; tp->snd_wl1 = th->th_seq; tp->snd_wl2 = th->th_ack; if (tp->snd_wnd > tp->max_sndwnd) tp->max_sndwnd = tp->snd_wnd; needoutput = 1; } /* * Process segments with URG. */ if ((thflags & TH_URG) && th->th_urp && TCPS_HAVERCVDFIN(tp->t_state) == 0) { /* * This is a kludge, but if we receive and accept * random urgent pointers, we'll crash in * soreceive. It's hard to imagine someone * actually wanting to send this much urgent data. */ SOCKBUF_LOCK(&so->so_rcv); if (th->th_urp + so->so_rcv.sb_cc > sb_max) { th->th_urp = 0; /* XXX */ thflags &= ~TH_URG; /* XXX */ SOCKBUF_UNLOCK(&so->so_rcv); /* XXX */ goto dodata; /* XXX */ } /* * If this segment advances the known urgent pointer, * then mark the data stream. This should not happen * in CLOSE_WAIT, CLOSING, LAST_ACK or TIME_WAIT STATES since * a FIN has been received from the remote side. * In these states we ignore the URG. * * According to RFC961 (Assigned Protocols), * the urgent pointer points to the last octet * of urgent data. We continue, however, * to consider it to indicate the first octet * of data past the urgent section as the original * spec states (in one of two places). */ if (SEQ_GT(th->th_seq+th->th_urp, tp->rcv_up)) { tp->rcv_up = th->th_seq + th->th_urp; so->so_oobmark = so->so_rcv.sb_cc + (tp->rcv_up - tp->rcv_nxt) - 1; if (so->so_oobmark == 0) so->so_rcv.sb_state |= SBS_RCVATMARK; sohasoutofband(so); tp->t_oobflags &= ~(TCPOOB_HAVEDATA | TCPOOB_HADDATA); } SOCKBUF_UNLOCK(&so->so_rcv); /* * Remove out of band data so doesn't get presented to user. * This can happen independent of advancing the URG pointer, * but if two URG's are pending at once, some out-of-band * data may creep in... ick. */ if (th->th_urp <= (u_long)tlen && !(so->so_options & SO_OOBINLINE)) { /* hdr drop is delayed */ tcp_pulloutofband(so, th, m, drop_hdrlen); } } else { /* * If no out of band data is expected, * pull receive urgent pointer along * with the receive window. */ if (SEQ_GT(tp->rcv_nxt, tp->rcv_up)) tp->rcv_up = tp->rcv_nxt; } dodata: /* XXX */ - INP_INFO_LOCK_ASSERT(&V_tcbinfo); - KASSERT(ti_locked == TI_RLOCKED || ti_locked == TI_WLOCKED, - ("tcp_do_segment: dodata ti_locked %d", ti_locked)); INP_WLOCK_ASSERT(tp->t_inpcb); /* * Process the segment text, merging it into the TCP sequencing queue, * and arranging for acknowledgment of receipt if necessary. * This process logically involves adjusting tp->rcv_wnd as data * is presented to the user (this happens in tcp_usrreq.c, * case PRU_RCVD). If a FIN has already been received on this * connection then we just ignore the text. */ if ((tlen || (thflags & TH_FIN)) && TCPS_HAVERCVDFIN(tp->t_state) == 0) { tcp_seq save_start = th->th_seq; m_adj(m, drop_hdrlen); /* delayed header drop */ /* * Insert segment which includes th into TCP reassembly queue * with control block tp. Set thflags to whether reassembly now * includes a segment with FIN. This handles the common case * inline (segment is the next to be received on an established * connection, and the queue is empty), avoiding linkage into * and removal from the queue and repetition of various * conversions. * Set DELACK for segments received in order, but ack * immediately when segments are out of order (so * fast retransmit can work). */ if (th->th_seq == tp->rcv_nxt && LIST_EMPTY(&tp->t_segq) && TCPS_HAVEESTABLISHED(tp->t_state)) { if (DELAY_ACK(tp)) tp->t_flags |= TF_DELACK; else tp->t_flags |= TF_ACKNOW; tp->rcv_nxt += tlen; thflags = th->th_flags & TH_FIN; TCPSTAT_INC(tcps_rcvpack); TCPSTAT_ADD(tcps_rcvbyte, tlen); ND6_HINT(tp); SOCKBUF_LOCK(&so->so_rcv); if (so->so_rcv.sb_state & SBS_CANTRCVMORE) m_freem(m); else sbappendstream_locked(&so->so_rcv, m); /* NB: sorwakeup_locked() does an implicit unlock. */ sorwakeup_locked(so); } else { /* * XXX: Due to the header drop above "th" is * theoretically invalid by now. Fortunately * m_adj() doesn't actually frees any mbufs * when trimming from the head. */ thflags = tcp_reass(tp, th, &tlen, m); tp->t_flags |= TF_ACKNOW; } if (tlen > 0 && (tp->t_flags & TF_SACK_PERMIT)) tcp_update_sack_list(tp, save_start, save_start + tlen); #if 0 /* * Note the amount of data that peer has sent into * our window, in order to estimate the sender's * buffer size. * XXX: Unused. */ if (SEQ_GT(tp->rcv_adv, tp->rcv_nxt)) len = so->so_rcv.sb_hiwat - (tp->rcv_adv - tp->rcv_nxt); else len = so->so_rcv.sb_hiwat; #endif } else { m_freem(m); thflags &= ~TH_FIN; } /* * If FIN is received ACK the FIN and let the user know * that the connection is closing. */ if (thflags & TH_FIN) { if (TCPS_HAVERCVDFIN(tp->t_state) == 0) { socantrcvmore(so); /* * If connection is half-synchronized * (ie NEEDSYN flag on) then delay ACK, * so it may be piggybacked when SYN is sent. * Otherwise, since we received a FIN then no * more input can be expected, send ACK now. */ if (tp->t_flags & TF_NEEDSYN) tp->t_flags |= TF_DELACK; else tp->t_flags |= TF_ACKNOW; tp->rcv_nxt++; } switch (tp->t_state) { /* * In SYN_RECEIVED and ESTABLISHED STATES * enter the CLOSE_WAIT state. */ case TCPS_SYN_RECEIVED: tp->t_starttime = ticks; /* FALLTHROUGH */ case TCPS_ESTABLISHED: tp->t_state = TCPS_CLOSE_WAIT; break; /* * If still in FIN_WAIT_1 STATE FIN has not been acked so * enter the CLOSING state. */ case TCPS_FIN_WAIT_1: tp->t_state = TCPS_CLOSING; break; /* * In FIN_WAIT_2 state enter the TIME_WAIT state, * starting the time-wait timer, turning off the other * standard timers. */ case TCPS_FIN_WAIT_2: INP_INFO_WLOCK_ASSERT(&V_tcbinfo); KASSERT(ti_locked == TI_WLOCKED, ("%s: dodata " "TCP_FIN_WAIT_2 ti_locked: %d", __func__, ti_locked)); tcp_twstart(tp); INP_INFO_WUNLOCK(&V_tcbinfo); return; } } - if (ti_locked == TI_RLOCKED) - INP_INFO_RUNLOCK(&V_tcbinfo); - else if (ti_locked == TI_WLOCKED) + if (ti_locked == TI_WLOCKED) INP_INFO_WUNLOCK(&V_tcbinfo); - else - panic("%s: dodata epilogue ti_locked %d", __func__, - ti_locked); ti_locked = TI_UNLOCKED; #ifdef TCPDEBUG if (so->so_options & SO_DEBUG) tcp_trace(TA_INPUT, ostate, tp, (void *)tcp_saveipgen, &tcp_savetcp, 0); #endif /* * Return any desired output. */ if (needoutput || (tp->t_flags & TF_ACKNOW)) (void) tcp_output(tp); check_delack: KASSERT(ti_locked == TI_UNLOCKED, ("%s: check_delack ti_locked %d", __func__, ti_locked)); INP_INFO_UNLOCK_ASSERT(&V_tcbinfo); INP_WLOCK_ASSERT(tp->t_inpcb); if (tp->t_flags & TF_DELACK) { tp->t_flags &= ~TF_DELACK; tcp_timer_activate(tp, TT_DELACK, tcp_delacktime); } INP_WUNLOCK(tp->t_inpcb); return; dropafterack: - KASSERT(ti_locked == TI_RLOCKED || ti_locked == TI_WLOCKED, - ("tcp_do_segment: dropafterack ti_locked %d", ti_locked)); - /* * Generate an ACK dropping incoming segment if it occupies * sequence space, where the ACK reflects our state. * * We can now skip the test for the RST flag since all * paths to this code happen after packets containing * RST have been dropped. * * In the SYN-RECEIVED state, don't send an ACK unless the * segment we received passes the SYN-RECEIVED ACK test. * If it fails send a RST. This breaks the loop in the * "LAND" DoS attack, and also prevents an ACK storm * between two listening ports that have been sent forged * SYN segments, each with the source address of the other. */ if (tp->t_state == TCPS_SYN_RECEIVED && (thflags & TH_ACK) && (SEQ_GT(tp->snd_una, th->th_ack) || SEQ_GT(th->th_ack, tp->snd_max)) ) { rstreason = BANDLIM_RST_OPENPORT; goto dropwithreset; } #ifdef TCPDEBUG if (so->so_options & SO_DEBUG) tcp_trace(TA_DROP, ostate, tp, (void *)tcp_saveipgen, &tcp_savetcp, 0); #endif - if (ti_locked == TI_RLOCKED) - INP_INFO_RUNLOCK(&V_tcbinfo); - else if (ti_locked == TI_WLOCKED) + if (ti_locked == TI_WLOCKED) INP_INFO_WUNLOCK(&V_tcbinfo); - else - panic("%s: dropafterack epilogue ti_locked %d", __func__, - ti_locked); ti_locked = TI_UNLOCKED; tp->t_flags |= TF_ACKNOW; (void) tcp_output(tp); INP_WUNLOCK(tp->t_inpcb); m_freem(m); return; dropwithreset: - if (ti_locked == TI_RLOCKED) - INP_INFO_RUNLOCK(&V_tcbinfo); - else if (ti_locked == TI_WLOCKED) + if (ti_locked == TI_WLOCKED) INP_INFO_WUNLOCK(&V_tcbinfo); - else - panic("%s: dropwithreset ti_locked %d", __func__, ti_locked); ti_locked = TI_UNLOCKED; if (tp != NULL) { tcp_dropwithreset(m, th, tp, tlen, rstreason); INP_WUNLOCK(tp->t_inpcb); } else tcp_dropwithreset(m, th, NULL, tlen, rstreason); return; drop: - if (ti_locked == TI_RLOCKED) - INP_INFO_RUNLOCK(&V_tcbinfo); - else if (ti_locked == TI_WLOCKED) + if (ti_locked == TI_WLOCKED) { INP_INFO_WUNLOCK(&V_tcbinfo); + ti_locked = TI_UNLOCKED; + } #ifdef INVARIANTS else INP_INFO_UNLOCK_ASSERT(&V_tcbinfo); #endif - ti_locked = TI_UNLOCKED; /* * Drop space held by incoming segment and return. */ #ifdef TCPDEBUG if (tp == NULL || (tp->t_inpcb->inp_socket->so_options & SO_DEBUG)) tcp_trace(TA_DROP, ostate, tp, (void *)tcp_saveipgen, &tcp_savetcp, 0); #endif if (tp != NULL) INP_WUNLOCK(tp->t_inpcb); m_freem(m); } /* * Issue RST and make ACK acceptable to originator of segment. * The mbuf must still include the original packet header. * tp may be NULL. */ static void tcp_dropwithreset(struct mbuf *m, struct tcphdr *th, struct tcpcb *tp, int tlen, int rstreason) { #ifdef INET struct ip *ip; #endif #ifdef INET6 struct ip6_hdr *ip6; #endif if (tp != NULL) { INP_WLOCK_ASSERT(tp->t_inpcb); } /* Don't bother if destination was broadcast/multicast. */ if ((th->th_flags & TH_RST) || m->m_flags & (M_BCAST|M_MCAST)) goto drop; #ifdef INET6 if (mtod(m, struct ip *)->ip_v == 6) { ip6 = mtod(m, struct ip6_hdr *); if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) goto drop; /* IPv6 anycast check is done at tcp6_input() */ } #endif #if defined(INET) && defined(INET6) else #endif #ifdef INET { ip = mtod(m, struct ip *); if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) || IN_MULTICAST(ntohl(ip->ip_src.s_addr)) || ip->ip_src.s_addr == htonl(INADDR_BROADCAST) || in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) goto drop; } #endif /* Perform bandwidth limiting. */ if (badport_bandlim(rstreason) < 0) goto drop; /* tcp_respond consumes the mbuf chain. */ if (th->th_flags & TH_ACK) { tcp_respond(tp, mtod(m, void *), th, m, (tcp_seq)0, th->th_ack, TH_RST); } else { if (th->th_flags & TH_SYN) tlen++; tcp_respond(tp, mtod(m, void *), th, m, th->th_seq+tlen, (tcp_seq)0, TH_RST|TH_ACK); } return; drop: m_freem(m); } /* * Parse TCP options and place in tcpopt. */ static void tcp_dooptions(struct tcpopt *to, u_char *cp, int cnt, int flags) { int opt, optlen; to->to_flags = 0; for (; cnt > 0; cnt -= optlen, cp += optlen) { opt = cp[0]; if (opt == TCPOPT_EOL) break; if (opt == TCPOPT_NOP) optlen = 1; else { if (cnt < 2) break; optlen = cp[1]; if (optlen < 2 || optlen > cnt) break; } switch (opt) { case TCPOPT_MAXSEG: if (optlen != TCPOLEN_MAXSEG) continue; if (!(flags & TO_SYN)) continue; to->to_flags |= TOF_MSS; bcopy((char *)cp + 2, (char *)&to->to_mss, sizeof(to->to_mss)); to->to_mss = ntohs(to->to_mss); break; case TCPOPT_WINDOW: if (optlen != TCPOLEN_WINDOW) continue; if (!(flags & TO_SYN)) continue; to->to_flags |= TOF_SCALE; to->to_wscale = min(cp[2], TCP_MAX_WINSHIFT); break; case TCPOPT_TIMESTAMP: if (optlen != TCPOLEN_TIMESTAMP) continue; to->to_flags |= TOF_TS; bcopy((char *)cp + 2, (char *)&to->to_tsval, sizeof(to->to_tsval)); to->to_tsval = ntohl(to->to_tsval); bcopy((char *)cp + 6, (char *)&to->to_tsecr, sizeof(to->to_tsecr)); to->to_tsecr = ntohl(to->to_tsecr); break; #ifdef TCP_SIGNATURE /* * XXX In order to reply to a host which has set the * TCP_SIGNATURE option in its initial SYN, we have to * record the fact that the option was observed here * for the syncache code to perform the correct response. */ case TCPOPT_SIGNATURE: if (optlen != TCPOLEN_SIGNATURE) continue; to->to_flags |= TOF_SIGNATURE; to->to_signature = cp + 2; break; #endif case TCPOPT_SACK_PERMITTED: if (optlen != TCPOLEN_SACK_PERMITTED) continue; if (!(flags & TO_SYN)) continue; if (!V_tcp_do_sack) continue; to->to_flags |= TOF_SACKPERM; break; case TCPOPT_SACK: if (optlen <= 2 || (optlen - 2) % TCPOLEN_SACK != 0) continue; if (flags & TO_SYN) continue; to->to_flags |= TOF_SACK; to->to_nsacks = (optlen - 2) / TCPOLEN_SACK; to->to_sacks = cp + 2; TCPSTAT_INC(tcps_sack_rcv_blocks); break; default: continue; } } } /* * Pull out of band byte out of a segment so * it doesn't appear in the user's data queue. * It is still reflected in the segment length for * sequencing purposes. */ static void tcp_pulloutofband(struct socket *so, struct tcphdr *th, struct mbuf *m, int off) { int cnt = off + th->th_urp - 1; while (cnt >= 0) { if (m->m_len > cnt) { char *cp = mtod(m, caddr_t) + cnt; struct tcpcb *tp = sototcpcb(so); INP_WLOCK_ASSERT(tp->t_inpcb); tp->t_iobc = *cp; tp->t_oobflags |= TCPOOB_HAVEDATA; bcopy(cp+1, cp, (unsigned)(m->m_len - cnt - 1)); m->m_len--; if (m->m_flags & M_PKTHDR) m->m_pkthdr.len--; return; } cnt -= m->m_len; m = m->m_next; if (m == NULL) break; } panic("tcp_pulloutofband"); } /* * Collect new round-trip time estimate * and update averages and current timeout. */ static void tcp_xmit_timer(struct tcpcb *tp, int rtt) { int delta; INP_WLOCK_ASSERT(tp->t_inpcb); TCPSTAT_INC(tcps_rttupdated); tp->t_rttupdated++; if (tp->t_srtt != 0) { /* * srtt is stored as fixed point with 5 bits after the * binary point (i.e., scaled by 8). The following magic * is equivalent to the smoothing algorithm in rfc793 with * an alpha of .875 (srtt = rtt/8 + srtt*7/8 in fixed * point). Adjust rtt to origin 0. */ delta = ((rtt - 1) << TCP_DELTA_SHIFT) - (tp->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT)); if ((tp->t_srtt += delta) <= 0) tp->t_srtt = 1; /* * We accumulate a smoothed rtt variance (actually, a * smoothed mean difference), then set the retransmit * timer to smoothed rtt + 4 times the smoothed variance. * rttvar is stored as fixed point with 4 bits after the * binary point (scaled by 16). The following is * equivalent to rfc793 smoothing with an alpha of .75 * (rttvar = rttvar*3/4 + |delta| / 4). This replaces * rfc793's wired-in beta. */ if (delta < 0) delta = -delta; delta -= tp->t_rttvar >> (TCP_RTTVAR_SHIFT - TCP_DELTA_SHIFT); if ((tp->t_rttvar += delta) <= 0) tp->t_rttvar = 1; if (tp->t_rttbest > tp->t_srtt + tp->t_rttvar) tp->t_rttbest = tp->t_srtt + tp->t_rttvar; } else { /* * No rtt measurement yet - use the unsmoothed rtt. * Set the variance to half the rtt (so our first * retransmit happens at 3*rtt). */ tp->t_srtt = rtt << TCP_RTT_SHIFT; tp->t_rttvar = rtt << (TCP_RTTVAR_SHIFT - 1); tp->t_rttbest = tp->t_srtt + tp->t_rttvar; } tp->t_rtttime = 0; tp->t_rxtshift = 0; /* * the retransmit should happen at rtt + 4 * rttvar. * Because of the way we do the smoothing, srtt and rttvar * will each average +1/2 tick of bias. When we compute * the retransmit timer, we want 1/2 tick of rounding and * 1 extra tick because of +-1/2 tick uncertainty in the * firing of the timer. The bias will give us exactly the * 1.5 tick we need. But, because the bias is * statistical, we have to test that we don't drop below * the minimum feasible timer (which is 2 ticks). */ TCPT_RANGESET(tp->t_rxtcur, TCP_REXMTVAL(tp), max(tp->t_rttmin, rtt + 2), TCPTV_REXMTMAX); /* * We received an ack for a packet that wasn't retransmitted; * it is probably safe to discard any error indications we've * received recently. This isn't quite right, but close enough * for now (a route might have failed after we sent a segment, * and the return path might not be symmetrical). */ tp->t_softerror = 0; } /* * Determine a reasonable value for maxseg size. * If the route is known, check route for mtu. * If none, use an mss that can be handled on the outgoing * interface without forcing IP to fragment; if bigger than * an mbuf cluster (MCLBYTES), round down to nearest multiple of MCLBYTES * to utilize large mbufs. If no route is found, route has no mtu, * or the destination isn't local, use a default, hopefully conservative * size (usually 512 or the default IP max size, but no more than the mtu * of the interface), as we can't discover anything about intervening * gateways or networks. We also initialize the congestion/slow start * window to be a single segment if the destination isn't local. * While looking at the routing entry, we also initialize other path-dependent * parameters from pre-set or cached values in the routing entry. * * Also take into account the space needed for options that we * send regularly. Make maxseg shorter by that amount to assure * that we can send maxseg amount of data even when the options * are present. Store the upper limit of the length of options plus * data in maxopd. * * In case of T/TCP, we call this routine during implicit connection * setup as well (offer = -1), to initialize maxseg from the cached * MSS of our peer. * * NOTE that this routine is only called when we process an incoming * segment. Outgoing SYN/ACK MSS settings are handled in tcp_mssopt(). */ void tcp_mss_update(struct tcpcb *tp, int offer, struct hc_metrics_lite *metricptr, int *mtuflags) { int mss = 0; u_long maxmtu = 0; struct inpcb *inp = tp->t_inpcb; struct hc_metrics_lite metrics; int origoffer = offer; #ifdef INET6 int isipv6 = ((inp->inp_vflag & INP_IPV6) != 0) ? 1 : 0; size_t min_protoh = isipv6 ? sizeof (struct ip6_hdr) + sizeof (struct tcphdr) : sizeof (struct tcpiphdr); #else const size_t min_protoh = sizeof(struct tcpiphdr); #endif INP_WLOCK_ASSERT(tp->t_inpcb); /* Initialize. */ #ifdef INET6 if (isipv6) { maxmtu = tcp_maxmtu6(&inp->inp_inc, mtuflags); tp->t_maxopd = tp->t_maxseg = V_tcp_v6mssdflt; } #endif #if defined(INET) && defined(INET6) else #endif #ifdef INET { maxmtu = tcp_maxmtu(&inp->inp_inc, mtuflags); tp->t_maxopd = tp->t_maxseg = V_tcp_mssdflt; } #endif /* * No route to sender, stay with default mss and return. */ if (maxmtu == 0) { /* * In case we return early we need to initialize metrics * to a defined state as tcp_hc_get() would do for us * if there was no cache hit. */ if (metricptr != NULL) bzero(metricptr, sizeof(struct hc_metrics_lite)); return; } /* What have we got? */ switch (offer) { case 0: /* * Offer == 0 means that there was no MSS on the SYN * segment, in this case we use tcp_mssdflt as * already assigned to t_maxopd above. */ offer = tp->t_maxopd; break; case -1: /* * Offer == -1 means that we didn't receive SYN yet. */ /* FALLTHROUGH */ default: /* * Prevent DoS attack with too small MSS. Round up * to at least minmss. */ offer = max(offer, V_tcp_minmss); } /* * rmx information is now retrieved from tcp_hostcache. */ tcp_hc_get(&inp->inp_inc, &metrics); if (metricptr != NULL) bcopy(&metrics, metricptr, sizeof(struct hc_metrics_lite)); /* * If there's a discovered mtu int tcp hostcache, use it * else, use the link mtu. */ if (metrics.rmx_mtu) mss = min(metrics.rmx_mtu, maxmtu) - min_protoh; else { #ifdef INET6 if (isipv6) { mss = maxmtu - min_protoh; if (!V_path_mtu_discovery && !in6_localaddr(&inp->in6p_faddr)) mss = min(mss, V_tcp_v6mssdflt); } #endif #if defined(INET) && defined(INET6) else #endif #ifdef INET { mss = maxmtu - min_protoh; if (!V_path_mtu_discovery && !in_localaddr(inp->inp_faddr)) mss = min(mss, V_tcp_mssdflt); } #endif /* * XXX - The above conditional (mss = maxmtu - min_protoh) * probably violates the TCP spec. * The problem is that, since we don't know the * other end's MSS, we are supposed to use a conservative * default. But, if we do that, then MTU discovery will * never actually take place, because the conservative * default is much less than the MTUs typically seen * on the Internet today. For the moment, we'll sweep * this under the carpet. * * The conservative default might not actually be a problem * if the only case this occurs is when sending an initial * SYN with options and data to a host we've never talked * to before. Then, they will reply with an MSS value which * will get recorded and the new parameters should get * recomputed. For Further Study. */ } mss = min(mss, offer); /* * Sanity check: make sure that maxopd will be large * enough to allow some data on segments even if the * all the option space is used (40bytes). Otherwise * funny things may happen in tcp_output. */ mss = max(mss, 64); /* * maxopd stores the maximum length of data AND options * in a segment; maxseg is the amount of data in a normal * segment. We need to store this value (maxopd) apart * from maxseg, because now every segment carries options * and thus we normally have somewhat less data in segments. */ tp->t_maxopd = mss; /* * origoffer==-1 indicates that no segments were received yet. * In this case we just guess. */ if ((tp->t_flags & (TF_REQ_TSTMP|TF_NOOPT)) == TF_REQ_TSTMP && (origoffer == -1 || (tp->t_flags & TF_RCVD_TSTMP) == TF_RCVD_TSTMP)) mss -= TCPOLEN_TSTAMP_APPA; #if (MCLBYTES & (MCLBYTES - 1)) == 0 if (mss > MCLBYTES) mss &= ~(MCLBYTES-1); #else if (mss > MCLBYTES) mss = mss / MCLBYTES * MCLBYTES; #endif tp->t_maxseg = mss; } void tcp_mss(struct tcpcb *tp, int offer) { int mss; u_long bufsize; struct inpcb *inp; struct socket *so; struct hc_metrics_lite metrics; int mtuflags = 0; KASSERT(tp != NULL, ("%s: tp == NULL", __func__)); tcp_mss_update(tp, offer, &metrics, &mtuflags); mss = tp->t_maxseg; inp = tp->t_inpcb; /* * If there's a pipesize, change the socket buffer to that size, * don't change if sb_hiwat is different than default (then it * has been changed on purpose with setsockopt). * Make the socket buffers an integral number of mss units; * if the mss is larger than the socket buffer, decrease the mss. */ so = inp->inp_socket; SOCKBUF_LOCK(&so->so_snd); if ((so->so_snd.sb_hiwat == tcp_sendspace) && metrics.rmx_sendpipe) bufsize = metrics.rmx_sendpipe; else bufsize = so->so_snd.sb_hiwat; if (bufsize < mss) mss = bufsize; else { bufsize = roundup(bufsize, mss); if (bufsize > sb_max) bufsize = sb_max; if (bufsize > so->so_snd.sb_hiwat) (void)sbreserve_locked(&so->so_snd, bufsize, so, NULL); } SOCKBUF_UNLOCK(&so->so_snd); tp->t_maxseg = mss; SOCKBUF_LOCK(&so->so_rcv); if ((so->so_rcv.sb_hiwat == tcp_recvspace) && metrics.rmx_recvpipe) bufsize = metrics.rmx_recvpipe; else bufsize = so->so_rcv.sb_hiwat; if (bufsize > mss) { bufsize = roundup(bufsize, mss); if (bufsize > sb_max) bufsize = sb_max; if (bufsize > so->so_rcv.sb_hiwat) (void)sbreserve_locked(&so->so_rcv, bufsize, so, NULL); } SOCKBUF_UNLOCK(&so->so_rcv); /* Check the interface for TSO capabilities. */ if (mtuflags & CSUM_TSO) tp->t_flags |= TF_TSO; } /* * Determine the MSS option to send on an outgoing SYN. */ int tcp_mssopt(struct in_conninfo *inc) { int mss = 0; u_long maxmtu = 0; u_long thcmtu = 0; size_t min_protoh; KASSERT(inc != NULL, ("tcp_mssopt with NULL in_conninfo pointer")); #ifdef INET6 if (inc->inc_flags & INC_ISIPV6) { mss = V_tcp_v6mssdflt; maxmtu = tcp_maxmtu6(inc, NULL); thcmtu = tcp_hc_getmtu(inc); /* IPv4 and IPv6 */ min_protoh = sizeof(struct ip6_hdr) + sizeof(struct tcphdr); } #endif #if defined(INET) && defined(INET6) else #endif #ifdef INET { mss = V_tcp_mssdflt; maxmtu = tcp_maxmtu(inc, NULL); thcmtu = tcp_hc_getmtu(inc); /* IPv4 and IPv6 */ min_protoh = sizeof(struct tcpiphdr); } #endif if (maxmtu && thcmtu) mss = min(maxmtu, thcmtu) - min_protoh; else if (maxmtu || thcmtu) mss = max(maxmtu, thcmtu) - min_protoh; return (mss); } /* * On a partial ack arrives, force the retransmission of the * next unacknowledged segment. Do not clear tp->t_dupacks. * By setting snd_nxt to ti_ack, this forces retransmission timer to * be started again. */ static void tcp_newreno_partial_ack(struct tcpcb *tp, struct tcphdr *th) { tcp_seq onxt = tp->snd_nxt; u_long ocwnd = tp->snd_cwnd; INP_WLOCK_ASSERT(tp->t_inpcb); tcp_timer_activate(tp, TT_REXMT, 0); tp->t_rtttime = 0; tp->snd_nxt = th->th_ack; /* * Set snd_cwnd to one segment beyond acknowledged offset. * (tp->snd_una has not yet been updated when this function is called.) */ tp->snd_cwnd = tp->t_maxseg + BYTES_THIS_ACK(tp, th); tp->t_flags |= TF_ACKNOW; (void) tcp_output(tp); tp->snd_cwnd = ocwnd; if (SEQ_GT(onxt, tp->snd_nxt)) tp->snd_nxt = onxt; /* * Partial window deflation. Relies on fact that tp->snd_una * not updated yet. */ if (tp->snd_cwnd > BYTES_THIS_ACK(tp, th)) tp->snd_cwnd -= BYTES_THIS_ACK(tp, th); else tp->snd_cwnd = 0; tp->snd_cwnd += tp->t_maxseg; } Index: head/sys/netinet/tcp_subr.c =================================================================== --- head/sys/netinet/tcp_subr.c (revision 222487) +++ head/sys/netinet/tcp_subr.c (revision 222488) @@ -1,2315 +1,2303 @@ /*- * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)tcp_subr.c 8.2 (Berkeley) 5/24/95 */ #include __FBSDID("$FreeBSD$"); #include "opt_compat.h" #include "opt_inet.h" #include "opt_inet6.h" #include "opt_ipsec.h" #include "opt_tcpdebug.h" #include #include #include #include #include #include #include #include #include #include #ifdef INET6 #include #endif #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef INET6 #include #include #include #include #include #endif #include #include #include #include #include #include #ifdef INET6 #include #endif #include #ifdef TCPDEBUG #include #endif #ifdef INET6 #include #endif #ifdef IPSEC #include #include #ifdef INET6 #include #endif #include #include #endif /*IPSEC*/ #include #include #include VNET_DEFINE(int, tcp_mssdflt) = TCP_MSS; #ifdef INET6 VNET_DEFINE(int, tcp_v6mssdflt) = TCP6_MSS; #endif static int sysctl_net_inet_tcp_mss_check(SYSCTL_HANDLER_ARGS) { int error, new; new = V_tcp_mssdflt; error = sysctl_handle_int(oidp, &new, 0, req); if (error == 0 && req->newptr) { if (new < TCP_MINMSS) error = EINVAL; else V_tcp_mssdflt = new; } return (error); } SYSCTL_VNET_PROC(_net_inet_tcp, TCPCTL_MSSDFLT, mssdflt, CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(tcp_mssdflt), 0, &sysctl_net_inet_tcp_mss_check, "I", "Default TCP Maximum Segment Size"); #ifdef INET6 static int sysctl_net_inet_tcp_mss_v6_check(SYSCTL_HANDLER_ARGS) { int error, new; new = V_tcp_v6mssdflt; error = sysctl_handle_int(oidp, &new, 0, req); if (error == 0 && req->newptr) { if (new < TCP_MINMSS) error = EINVAL; else V_tcp_v6mssdflt = new; } return (error); } SYSCTL_VNET_PROC(_net_inet_tcp, TCPCTL_V6MSSDFLT, v6mssdflt, CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(tcp_v6mssdflt), 0, &sysctl_net_inet_tcp_mss_v6_check, "I", "Default TCP Maximum Segment Size for IPv6"); #endif /* INET6 */ /* * Minimum MSS we accept and use. This prevents DoS attacks where * we are forced to a ridiculous low MSS like 20 and send hundreds * of packets instead of one. The effect scales with the available * bandwidth and quickly saturates the CPU and network interface * with packet generation and sending. Set to zero to disable MINMSS * checking. This setting prevents us from sending too small packets. */ VNET_DEFINE(int, tcp_minmss) = TCP_MINMSS; SYSCTL_VNET_INT(_net_inet_tcp, OID_AUTO, minmss, CTLFLAG_RW, &VNET_NAME(tcp_minmss), 0, "Minmum TCP Maximum Segment Size"); VNET_DEFINE(int, tcp_do_rfc1323) = 1; SYSCTL_VNET_INT(_net_inet_tcp, TCPCTL_DO_RFC1323, rfc1323, CTLFLAG_RW, &VNET_NAME(tcp_do_rfc1323), 0, "Enable rfc1323 (high performance TCP) extensions"); static int tcp_log_debug = 0; SYSCTL_INT(_net_inet_tcp, OID_AUTO, log_debug, CTLFLAG_RW, &tcp_log_debug, 0, "Log errors caused by incoming TCP segments"); static int tcp_tcbhashsize = 0; SYSCTL_INT(_net_inet_tcp, OID_AUTO, tcbhashsize, CTLFLAG_RDTUN, &tcp_tcbhashsize, 0, "Size of TCP control-block hashtable"); static int do_tcpdrain = 1; SYSCTL_INT(_net_inet_tcp, OID_AUTO, do_tcpdrain, CTLFLAG_RW, &do_tcpdrain, 0, "Enable tcp_drain routine for extra help when low on mbufs"); SYSCTL_VNET_UINT(_net_inet_tcp, OID_AUTO, pcbcount, CTLFLAG_RD, &VNET_NAME(tcbinfo.ipi_count), 0, "Number of active PCBs"); static VNET_DEFINE(int, icmp_may_rst) = 1; #define V_icmp_may_rst VNET(icmp_may_rst) SYSCTL_VNET_INT(_net_inet_tcp, OID_AUTO, icmp_may_rst, CTLFLAG_RW, &VNET_NAME(icmp_may_rst), 0, "Certain ICMP unreachable messages may abort connections in SYN_SENT"); static VNET_DEFINE(int, tcp_isn_reseed_interval) = 0; #define V_tcp_isn_reseed_interval VNET(tcp_isn_reseed_interval) SYSCTL_VNET_INT(_net_inet_tcp, OID_AUTO, isn_reseed_interval, CTLFLAG_RW, &VNET_NAME(tcp_isn_reseed_interval), 0, "Seconds between reseeding of ISN secret"); #ifdef TCP_SORECEIVE_STREAM static int tcp_soreceive_stream = 0; SYSCTL_INT(_net_inet_tcp, OID_AUTO, soreceive_stream, CTLFLAG_RDTUN, &tcp_soreceive_stream, 0, "Using soreceive_stream for TCP sockets"); #endif #ifdef TCP_SIGNATURE static int tcp_sig_checksigs = 1; SYSCTL_INT(_net_inet_tcp, OID_AUTO, signature_verify_input, CTLFLAG_RW, &tcp_sig_checksigs, 0, "Verify RFC2385 digests on inbound traffic"); #endif VNET_DEFINE(uma_zone_t, sack_hole_zone); #define V_sack_hole_zone VNET(sack_hole_zone) VNET_DEFINE(struct hhook_head *, tcp_hhh[HHOOK_TCP_LAST+1]); static struct inpcb *tcp_notify(struct inpcb *, int); static char * tcp_log_addr(struct in_conninfo *inc, struct tcphdr *th, void *ip4hdr, const void *ip6hdr); /* * Target size of TCP PCB hash tables. Must be a power of two. * * Note that this can be overridden by the kernel environment * variable net.inet.tcp.tcbhashsize */ #ifndef TCBHASHSIZE #define TCBHASHSIZE 512 #endif /* * XXX * Callouts should be moved into struct tcp directly. They are currently * separate because the tcpcb structure is exported to userland for sysctl * parsing purposes, which do not know about callouts. */ struct tcpcb_mem { struct tcpcb tcb; struct tcp_timer tt; struct cc_var ccv; struct osd osd; }; static VNET_DEFINE(uma_zone_t, tcpcb_zone); #define V_tcpcb_zone VNET(tcpcb_zone) MALLOC_DEFINE(M_TCPLOG, "tcplog", "TCP address and flags print buffers"); static struct mtx isn_mtx; #define ISN_LOCK_INIT() mtx_init(&isn_mtx, "isn_mtx", NULL, MTX_DEF) #define ISN_LOCK() mtx_lock(&isn_mtx) #define ISN_UNLOCK() mtx_unlock(&isn_mtx) /* * TCP initialization. */ static void tcp_zone_change(void *tag) { uma_zone_set_max(V_tcbinfo.ipi_zone, maxsockets); uma_zone_set_max(V_tcpcb_zone, maxsockets); tcp_tw_zone_change(); } static int tcp_inpcb_init(void *mem, int size, int flags) { struct inpcb *inp = mem; INP_LOCK_INIT(inp, "inp", "tcpinp"); return (0); } void tcp_init(void) { int hashsize; if (hhook_head_register(HHOOK_TYPE_TCP, HHOOK_TCP_EST_IN, &V_tcp_hhh[HHOOK_TCP_EST_IN], HHOOK_NOWAIT|HHOOK_HEADISINVNET) != 0) printf("%s: WARNING: unable to register helper hook\n", __func__); if (hhook_head_register(HHOOK_TYPE_TCP, HHOOK_TCP_EST_OUT, &V_tcp_hhh[HHOOK_TCP_EST_OUT], HHOOK_NOWAIT|HHOOK_HEADISINVNET) != 0) printf("%s: WARNING: unable to register helper hook\n", __func__); hashsize = TCBHASHSIZE; TUNABLE_INT_FETCH("net.inet.tcp.tcbhashsize", &hashsize); if (!powerof2(hashsize)) { printf("WARNING: TCB hash size not a power of 2\n"); hashsize = 512; /* safe default */ } in_pcbinfo_init(&V_tcbinfo, "tcp", &V_tcb, hashsize, hashsize, "tcp_inpcb", tcp_inpcb_init, NULL, UMA_ZONE_NOFREE); /* * These have to be type stable for the benefit of the timers. */ V_tcpcb_zone = uma_zcreate("tcpcb", sizeof(struct tcpcb_mem), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); uma_zone_set_max(V_tcpcb_zone, maxsockets); tcp_tw_init(); syncache_init(); tcp_hc_init(); tcp_reass_init(); TUNABLE_INT_FETCH("net.inet.tcp.sack.enable", &V_tcp_do_sack); V_sack_hole_zone = uma_zcreate("sackhole", sizeof(struct sackhole), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); /* Skip initialization of globals for non-default instances. */ if (!IS_DEFAULT_VNET(curvnet)) return; /* XXX virtualize those bellow? */ tcp_delacktime = TCPTV_DELACK; tcp_keepinit = TCPTV_KEEP_INIT; tcp_keepidle = TCPTV_KEEP_IDLE; tcp_keepintvl = TCPTV_KEEPINTVL; tcp_maxpersistidle = TCPTV_KEEP_IDLE; tcp_msl = TCPTV_MSL; tcp_rexmit_min = TCPTV_MIN; if (tcp_rexmit_min < 1) tcp_rexmit_min = 1; tcp_rexmit_slop = TCPTV_CPU_VAR; tcp_finwait2_timeout = TCPTV_FINWAIT2_TIMEOUT; tcp_tcbhashsize = hashsize; #ifdef TCP_SORECEIVE_STREAM TUNABLE_INT_FETCH("net.inet.tcp.soreceive_stream", &tcp_soreceive_stream); if (tcp_soreceive_stream) { tcp_usrreqs.pru_soreceive = soreceive_stream; tcp6_usrreqs.pru_soreceive = soreceive_stream; } #endif #ifdef INET6 #define TCP_MINPROTOHDR (sizeof(struct ip6_hdr) + sizeof(struct tcphdr)) #else /* INET6 */ #define TCP_MINPROTOHDR (sizeof(struct tcpiphdr)) #endif /* INET6 */ if (max_protohdr < TCP_MINPROTOHDR) max_protohdr = TCP_MINPROTOHDR; if (max_linkhdr + TCP_MINPROTOHDR > MHLEN) panic("tcp_init"); #undef TCP_MINPROTOHDR ISN_LOCK_INIT(); EVENTHANDLER_REGISTER(shutdown_pre_sync, tcp_fini, NULL, SHUTDOWN_PRI_DEFAULT); EVENTHANDLER_REGISTER(maxsockets_change, tcp_zone_change, NULL, EVENTHANDLER_PRI_ANY); } #ifdef VIMAGE void tcp_destroy(void) { tcp_reass_destroy(); tcp_hc_destroy(); syncache_destroy(); tcp_tw_destroy(); in_pcbinfo_destroy(&V_tcbinfo); uma_zdestroy(V_sack_hole_zone); uma_zdestroy(V_tcpcb_zone); } #endif void tcp_fini(void *xtp) { } /* * Fill in the IP and TCP headers for an outgoing packet, given the tcpcb. * tcp_template used to store this data in mbufs, but we now recopy it out * of the tcpcb each time to conserve mbufs. */ void tcpip_fillheaders(struct inpcb *inp, void *ip_ptr, void *tcp_ptr) { struct tcphdr *th = (struct tcphdr *)tcp_ptr; INP_WLOCK_ASSERT(inp); #ifdef INET6 if ((inp->inp_vflag & INP_IPV6) != 0) { struct ip6_hdr *ip6; ip6 = (struct ip6_hdr *)ip_ptr; ip6->ip6_flow = (ip6->ip6_flow & ~IPV6_FLOWINFO_MASK) | (inp->inp_flow & IPV6_FLOWINFO_MASK); ip6->ip6_vfc = (ip6->ip6_vfc & ~IPV6_VERSION_MASK) | (IPV6_VERSION & IPV6_VERSION_MASK); ip6->ip6_nxt = IPPROTO_TCP; ip6->ip6_plen = htons(sizeof(struct tcphdr)); ip6->ip6_src = inp->in6p_laddr; ip6->ip6_dst = inp->in6p_faddr; } #endif /* INET6 */ #if defined(INET6) && defined(INET) else #endif #ifdef INET { struct ip *ip; ip = (struct ip *)ip_ptr; ip->ip_v = IPVERSION; ip->ip_hl = 5; ip->ip_tos = inp->inp_ip_tos; ip->ip_len = 0; ip->ip_id = 0; ip->ip_off = 0; ip->ip_ttl = inp->inp_ip_ttl; ip->ip_sum = 0; ip->ip_p = IPPROTO_TCP; ip->ip_src = inp->inp_laddr; ip->ip_dst = inp->inp_faddr; } #endif /* INET */ th->th_sport = inp->inp_lport; th->th_dport = inp->inp_fport; th->th_seq = 0; th->th_ack = 0; th->th_x2 = 0; th->th_off = 5; th->th_flags = 0; th->th_win = 0; th->th_urp = 0; th->th_sum = 0; /* in_pseudo() is called later for ipv4 */ } /* * Create template to be used to send tcp packets on a connection. * Allocates an mbuf and fills in a skeletal tcp/ip header. The only * use for this function is in keepalives, which use tcp_respond. */ struct tcptemp * tcpip_maketemplate(struct inpcb *inp) { struct tcptemp *t; t = malloc(sizeof(*t), M_TEMP, M_NOWAIT); if (t == NULL) return (NULL); tcpip_fillheaders(inp, (void *)&t->tt_ipgen, (void *)&t->tt_t); return (t); } /* * Send a single message to the TCP at address specified by * the given TCP/IP header. If m == NULL, then we make a copy * of the tcpiphdr at ti and send directly to the addressed host. * This is used to force keep alive messages out using the TCP * template for a connection. If flags are given then we send * a message back to the TCP which originated the * segment ti, * and discard the mbuf containing it and any other attached mbufs. * * In any case the ack and sequence number of the transmitted * segment are as specified by the parameters. * * NOTE: If m != NULL, then ti must point to *inside* the mbuf. */ void tcp_respond(struct tcpcb *tp, void *ipgen, struct tcphdr *th, struct mbuf *m, tcp_seq ack, tcp_seq seq, int flags) { int tlen; int win = 0; struct ip *ip; struct tcphdr *nth; #ifdef INET6 struct ip6_hdr *ip6; int isipv6; #endif /* INET6 */ int ipflags = 0; struct inpcb *inp; KASSERT(tp != NULL || m != NULL, ("tcp_respond: tp and m both NULL")); #ifdef INET6 isipv6 = ((struct ip *)ipgen)->ip_v == (IPV6_VERSION >> 4); ip6 = ipgen; #endif /* INET6 */ ip = ipgen; if (tp != NULL) { inp = tp->t_inpcb; KASSERT(inp != NULL, ("tcp control block w/o inpcb")); INP_WLOCK_ASSERT(inp); } else inp = NULL; if (tp != NULL) { if (!(flags & TH_RST)) { win = sbspace(&inp->inp_socket->so_rcv); if (win > (long)TCP_MAXWIN << tp->rcv_scale) win = (long)TCP_MAXWIN << tp->rcv_scale; } } if (m == NULL) { m = m_gethdr(M_DONTWAIT, MT_DATA); if (m == NULL) return; tlen = 0; m->m_data += max_linkhdr; #ifdef INET6 if (isipv6) { bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(struct ip6_hdr)); ip6 = mtod(m, struct ip6_hdr *); nth = (struct tcphdr *)(ip6 + 1); } else #endif /* INET6 */ { bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip)); ip = mtod(m, struct ip *); nth = (struct tcphdr *)(ip + 1); } bcopy((caddr_t)th, (caddr_t)nth, sizeof(struct tcphdr)); flags = TH_ACK; } else { /* * reuse the mbuf. * XXX MRT We inherrit the FIB, which is lucky. */ m_freem(m->m_next); m->m_next = NULL; m->m_data = (caddr_t)ipgen; /* m_len is set later */ tlen = 0; #define xchg(a,b,type) { type t; t=a; a=b; b=t; } #ifdef INET6 if (isipv6) { xchg(ip6->ip6_dst, ip6->ip6_src, struct in6_addr); nth = (struct tcphdr *)(ip6 + 1); } else #endif /* INET6 */ { xchg(ip->ip_dst.s_addr, ip->ip_src.s_addr, uint32_t); nth = (struct tcphdr *)(ip + 1); } if (th != nth) { /* * this is usually a case when an extension header * exists between the IPv6 header and the * TCP header. */ nth->th_sport = th->th_sport; nth->th_dport = th->th_dport; } xchg(nth->th_dport, nth->th_sport, uint16_t); #undef xchg } #ifdef INET6 if (isipv6) { ip6->ip6_flow = 0; ip6->ip6_vfc = IPV6_VERSION; ip6->ip6_nxt = IPPROTO_TCP; ip6->ip6_plen = htons((u_short)(sizeof (struct tcphdr) + tlen)); tlen += sizeof (struct ip6_hdr) + sizeof (struct tcphdr); } #endif #if defined(INET) && defined(INET6) else #endif #ifdef INET { tlen += sizeof (struct tcpiphdr); ip->ip_len = tlen; ip->ip_ttl = V_ip_defttl; if (V_path_mtu_discovery) ip->ip_off |= IP_DF; } #endif m->m_len = tlen; m->m_pkthdr.len = tlen; m->m_pkthdr.rcvif = NULL; #ifdef MAC if (inp != NULL) { /* * Packet is associated with a socket, so allow the * label of the response to reflect the socket label. */ INP_WLOCK_ASSERT(inp); mac_inpcb_create_mbuf(inp, m); } else { /* * Packet is not associated with a socket, so possibly * update the label in place. */ mac_netinet_tcp_reply(m); } #endif nth->th_seq = htonl(seq); nth->th_ack = htonl(ack); nth->th_x2 = 0; nth->th_off = sizeof (struct tcphdr) >> 2; nth->th_flags = flags; if (tp != NULL) nth->th_win = htons((u_short) (win >> tp->rcv_scale)); else nth->th_win = htons((u_short)win); nth->th_urp = 0; #ifdef INET6 if (isipv6) { nth->th_sum = 0; nth->th_sum = in6_cksum(m, IPPROTO_TCP, sizeof(struct ip6_hdr), tlen - sizeof(struct ip6_hdr)); ip6->ip6_hlim = in6_selecthlim(tp != NULL ? tp->t_inpcb : NULL, NULL); } #endif /* INET6 */ #if defined(INET6) && defined(INET) else #endif #ifdef INET { nth->th_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr, htons((u_short)(tlen - sizeof(struct ip) + ip->ip_p))); m->m_pkthdr.csum_flags = CSUM_TCP; m->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum); } #endif /* INET */ #ifdef TCPDEBUG if (tp == NULL || (inp->inp_socket->so_options & SO_DEBUG)) tcp_trace(TA_OUTPUT, 0, tp, mtod(m, void *), th, 0); #endif #ifdef INET6 if (isipv6) (void) ip6_output(m, NULL, NULL, ipflags, NULL, NULL, inp); #endif /* INET6 */ #if defined(INET) && defined(INET6) else #endif #ifdef INET (void) ip_output(m, NULL, NULL, ipflags, NULL, inp); #endif } /* * Create a new TCP control block, making an * empty reassembly queue and hooking it to the argument * protocol control block. The `inp' parameter must have * come from the zone allocator set up in tcp_init(). */ struct tcpcb * tcp_newtcpcb(struct inpcb *inp) { struct tcpcb_mem *tm; struct tcpcb *tp; #ifdef INET6 int isipv6 = (inp->inp_vflag & INP_IPV6) != 0; #endif /* INET6 */ tm = uma_zalloc(V_tcpcb_zone, M_NOWAIT | M_ZERO); if (tm == NULL) return (NULL); tp = &tm->tcb; /* Initialise cc_var struct for this tcpcb. */ tp->ccv = &tm->ccv; tp->ccv->type = IPPROTO_TCP; tp->ccv->ccvc.tcp = tp; /* * Use the current system default CC algorithm. */ CC_LIST_RLOCK(); KASSERT(!STAILQ_EMPTY(&cc_list), ("cc_list is empty!")); CC_ALGO(tp) = CC_DEFAULT(); CC_LIST_RUNLOCK(); if (CC_ALGO(tp)->cb_init != NULL) if (CC_ALGO(tp)->cb_init(tp->ccv) > 0) { uma_zfree(V_tcpcb_zone, tm); return (NULL); } tp->osd = &tm->osd; if (khelp_init_osd(HELPER_CLASS_TCP, tp->osd)) { uma_zfree(V_tcpcb_zone, tm); return (NULL); } #ifdef VIMAGE tp->t_vnet = inp->inp_vnet; #endif tp->t_timers = &tm->tt; /* LIST_INIT(&tp->t_segq); */ /* XXX covered by M_ZERO */ tp->t_maxseg = tp->t_maxopd = #ifdef INET6 isipv6 ? V_tcp_v6mssdflt : #endif /* INET6 */ V_tcp_mssdflt; /* Set up our timeouts. */ callout_init(&tp->t_timers->tt_rexmt, CALLOUT_MPSAFE); callout_init(&tp->t_timers->tt_persist, CALLOUT_MPSAFE); callout_init(&tp->t_timers->tt_keep, CALLOUT_MPSAFE); callout_init(&tp->t_timers->tt_2msl, CALLOUT_MPSAFE); callout_init(&tp->t_timers->tt_delack, CALLOUT_MPSAFE); if (V_tcp_do_rfc1323) tp->t_flags = (TF_REQ_SCALE|TF_REQ_TSTMP); if (V_tcp_do_sack) tp->t_flags |= TF_SACK_PERMIT; TAILQ_INIT(&tp->snd_holes); tp->t_inpcb = inp; /* XXX */ /* * Init srtt to TCPTV_SRTTBASE (0), so we can tell that we have no * rtt estimate. Set rttvar so that srtt + 4 * rttvar gives * reasonable initial retransmit time. */ tp->t_srtt = TCPTV_SRTTBASE; tp->t_rttvar = ((TCPTV_RTOBASE - TCPTV_SRTTBASE) << TCP_RTTVAR_SHIFT) / 4; tp->t_rttmin = tcp_rexmit_min; tp->t_rxtcur = TCPTV_RTOBASE; tp->snd_cwnd = TCP_MAXWIN << TCP_MAX_WINSHIFT; tp->snd_ssthresh = TCP_MAXWIN << TCP_MAX_WINSHIFT; tp->t_rcvtime = ticks; /* * IPv4 TTL initialization is necessary for an IPv6 socket as well, * because the socket may be bound to an IPv6 wildcard address, * which may match an IPv4-mapped IPv6 address. */ inp->inp_ip_ttl = V_ip_defttl; inp->inp_ppcb = tp; return (tp); /* XXX */ } /* * Switch the congestion control algorithm back to NewReno for any active * control blocks using an algorithm which is about to go away. * This ensures the CC framework can allow the unload to proceed without leaving * any dangling pointers which would trigger a panic. * Returning non-zero would inform the CC framework that something went wrong * and it would be unsafe to allow the unload to proceed. However, there is no * way for this to occur with this implementation so we always return zero. */ int tcp_ccalgounload(struct cc_algo *unload_algo) { struct cc_algo *tmpalgo; struct inpcb *inp; struct tcpcb *tp; VNET_ITERATOR_DECL(vnet_iter); /* * Check all active control blocks across all network stacks and change * any that are using "unload_algo" back to NewReno. If "unload_algo" * requires cleanup code to be run, call it. */ VNET_LIST_RLOCK(); VNET_FOREACH(vnet_iter) { CURVNET_SET(vnet_iter); INP_INFO_RLOCK(&V_tcbinfo); /* * New connections already part way through being initialised * with the CC algo we're removing will not race with this code * because the INP_INFO_WLOCK is held during initialisation. We * therefore don't enter the loop below until the connection * list has stabilised. */ LIST_FOREACH(inp, &V_tcb, inp_list) { INP_WLOCK(inp); /* Important to skip tcptw structs. */ if (!(inp->inp_flags & INP_TIMEWAIT) && (tp = intotcpcb(inp)) != NULL) { /* * By holding INP_WLOCK here, we are assured * that the connection is not currently * executing inside the CC module's functions * i.e. it is safe to make the switch back to * NewReno. */ if (CC_ALGO(tp) == unload_algo) { tmpalgo = CC_ALGO(tp); /* NewReno does not require any init. */ CC_ALGO(tp) = &newreno_cc_algo; if (tmpalgo->cb_destroy != NULL) tmpalgo->cb_destroy(tp->ccv); } } INP_WUNLOCK(inp); } INP_INFO_RUNLOCK(&V_tcbinfo); CURVNET_RESTORE(); } VNET_LIST_RUNLOCK(); return (0); } /* * Drop a TCP connection, reporting * the specified error. If connection is synchronized, * then send a RST to peer. */ struct tcpcb * tcp_drop(struct tcpcb *tp, int errno) { struct socket *so = tp->t_inpcb->inp_socket; INP_INFO_WLOCK_ASSERT(&V_tcbinfo); INP_WLOCK_ASSERT(tp->t_inpcb); if (TCPS_HAVERCVDSYN(tp->t_state)) { tp->t_state = TCPS_CLOSED; (void) tcp_output_reset(tp); TCPSTAT_INC(tcps_drops); } else TCPSTAT_INC(tcps_conndrops); if (errno == ETIMEDOUT && tp->t_softerror) errno = tp->t_softerror; so->so_error = errno; return (tcp_close(tp)); } void tcp_discardcb(struct tcpcb *tp) { struct inpcb *inp = tp->t_inpcb; struct socket *so = inp->inp_socket; #ifdef INET6 int isipv6 = (inp->inp_vflag & INP_IPV6) != 0; #endif /* INET6 */ INP_WLOCK_ASSERT(inp); /* * Make sure that all of our timers are stopped before we delete the * PCB. * * XXXRW: Really, we would like to use callout_drain() here in order * to avoid races experienced in tcp_timer.c where a timer is already * executing at this point. However, we can't, both because we're * running in a context where we can't sleep, and also because we * hold locks required by the timers. What we instead need to do is * test to see if callout_drain() is required, and if so, defer some * portion of the remainder of tcp_discardcb() to an asynchronous * context that can callout_drain() and then continue. Some care * will be required to ensure that no further processing takes place * on the tcpcb, even though it hasn't been freed (a flag?). */ callout_stop(&tp->t_timers->tt_rexmt); callout_stop(&tp->t_timers->tt_persist); callout_stop(&tp->t_timers->tt_keep); callout_stop(&tp->t_timers->tt_2msl); callout_stop(&tp->t_timers->tt_delack); /* * If we got enough samples through the srtt filter, * save the rtt and rttvar in the routing entry. * 'Enough' is arbitrarily defined as 4 rtt samples. * 4 samples is enough for the srtt filter to converge * to within enough % of the correct value; fewer samples * and we could save a bogus rtt. The danger is not high * as tcp quickly recovers from everything. * XXX: Works very well but needs some more statistics! */ if (tp->t_rttupdated >= 4) { struct hc_metrics_lite metrics; u_long ssthresh; bzero(&metrics, sizeof(metrics)); /* * Update the ssthresh always when the conditions below * are satisfied. This gives us better new start value * for the congestion avoidance for new connections. * ssthresh is only set if packet loss occured on a session. * * XXXRW: 'so' may be NULL here, and/or socket buffer may be * being torn down. Ideally this code would not use 'so'. */ ssthresh = tp->snd_ssthresh; if (ssthresh != 0 && ssthresh < so->so_snd.sb_hiwat / 2) { /* * convert the limit from user data bytes to * packets then to packet data bytes. */ ssthresh = (ssthresh + tp->t_maxseg / 2) / tp->t_maxseg; if (ssthresh < 2) ssthresh = 2; ssthresh *= (u_long)(tp->t_maxseg + #ifdef INET6 (isipv6 ? sizeof (struct ip6_hdr) + sizeof (struct tcphdr) : #endif sizeof (struct tcpiphdr) #ifdef INET6 ) #endif ); } else ssthresh = 0; metrics.rmx_ssthresh = ssthresh; metrics.rmx_rtt = tp->t_srtt; metrics.rmx_rttvar = tp->t_rttvar; metrics.rmx_cwnd = tp->snd_cwnd; metrics.rmx_sendpipe = 0; metrics.rmx_recvpipe = 0; tcp_hc_update(&inp->inp_inc, &metrics); } /* free the reassembly queue, if any */ tcp_reass_flush(tp); /* Disconnect offload device, if any. */ tcp_offload_detach(tp); tcp_free_sackholes(tp); /* Allow the CC algorithm to clean up after itself. */ if (CC_ALGO(tp)->cb_destroy != NULL) CC_ALGO(tp)->cb_destroy(tp->ccv); khelp_destroy_osd(tp->osd); CC_ALGO(tp) = NULL; inp->inp_ppcb = NULL; tp->t_inpcb = NULL; uma_zfree(V_tcpcb_zone, tp); } /* * Attempt to close a TCP control block, marking it as dropped, and freeing * the socket if we hold the only reference. */ struct tcpcb * tcp_close(struct tcpcb *tp) { struct inpcb *inp = tp->t_inpcb; struct socket *so; INP_INFO_WLOCK_ASSERT(&V_tcbinfo); INP_WLOCK_ASSERT(inp); /* Notify any offload devices of listener close */ if (tp->t_state == TCPS_LISTEN) tcp_offload_listen_close(tp); in_pcbdrop(inp); TCPSTAT_INC(tcps_closed); KASSERT(inp->inp_socket != NULL, ("tcp_close: inp_socket NULL")); so = inp->inp_socket; soisdisconnected(so); if (inp->inp_flags & INP_SOCKREF) { KASSERT(so->so_state & SS_PROTOREF, ("tcp_close: !SS_PROTOREF")); inp->inp_flags &= ~INP_SOCKREF; INP_WUNLOCK(inp); ACCEPT_LOCK(); SOCK_LOCK(so); so->so_state &= ~SS_PROTOREF; sofree(so); return (NULL); } return (tp); } void tcp_drain(void) { VNET_ITERATOR_DECL(vnet_iter); if (!do_tcpdrain) return; VNET_LIST_RLOCK_NOSLEEP(); VNET_FOREACH(vnet_iter) { CURVNET_SET(vnet_iter); struct inpcb *inpb; struct tcpcb *tcpb; /* * Walk the tcpbs, if existing, and flush the reassembly queue, * if there is one... * XXX: The "Net/3" implementation doesn't imply that the TCP * reassembly queue should be flushed, but in a situation * where we're really low on mbufs, this is potentially * usefull. */ INP_INFO_RLOCK(&V_tcbinfo); LIST_FOREACH(inpb, V_tcbinfo.ipi_listhead, inp_list) { if (inpb->inp_flags & INP_TIMEWAIT) continue; INP_WLOCK(inpb); if ((tcpb = intotcpcb(inpb)) != NULL) { tcp_reass_flush(tcpb); tcp_clean_sackreport(tcpb); } INP_WUNLOCK(inpb); } INP_INFO_RUNLOCK(&V_tcbinfo); CURVNET_RESTORE(); } VNET_LIST_RUNLOCK_NOSLEEP(); } /* * Notify a tcp user of an asynchronous error; * store error as soft error, but wake up user * (for now, won't do anything until can select for soft error). * * Do not wake up user since there currently is no mechanism for * reporting soft errors (yet - a kqueue filter may be added). */ static struct inpcb * tcp_notify(struct inpcb *inp, int error) { struct tcpcb *tp; INP_INFO_WLOCK_ASSERT(&V_tcbinfo); INP_WLOCK_ASSERT(inp); if ((inp->inp_flags & INP_TIMEWAIT) || (inp->inp_flags & INP_DROPPED)) return (inp); tp = intotcpcb(inp); KASSERT(tp != NULL, ("tcp_notify: tp == NULL")); /* * Ignore some errors if we are hooked up. * If connection hasn't completed, has retransmitted several times, * and receives a second error, give up now. This is better * than waiting a long time to establish a connection that * can never complete. */ if (tp->t_state == TCPS_ESTABLISHED && (error == EHOSTUNREACH || error == ENETUNREACH || error == EHOSTDOWN)) { return (inp); } else if (tp->t_state < TCPS_ESTABLISHED && tp->t_rxtshift > 3 && tp->t_softerror) { tp = tcp_drop(tp, error); if (tp != NULL) return (inp); else return (NULL); } else { tp->t_softerror = error; return (inp); } #if 0 wakeup( &so->so_timeo); sorwakeup(so); sowwakeup(so); #endif } static int tcp_pcblist(SYSCTL_HANDLER_ARGS) { int error, i, m, n, pcb_count; struct inpcb *inp, **inp_list; inp_gen_t gencnt; struct xinpgen xig; /* * The process of preparing the TCB list is too time-consuming and * resource-intensive to repeat twice on every request. */ if (req->oldptr == NULL) { n = V_tcbinfo.ipi_count + syncache_pcbcount(); n += imax(n / 8, 10); req->oldidx = 2 * (sizeof xig) + n * sizeof(struct xtcpcb); return (0); } if (req->newptr != NULL) return (EPERM); /* * OK, now we're committed to doing something. */ INP_INFO_RLOCK(&V_tcbinfo); gencnt = V_tcbinfo.ipi_gencnt; n = V_tcbinfo.ipi_count; INP_INFO_RUNLOCK(&V_tcbinfo); m = syncache_pcbcount(); error = sysctl_wire_old_buffer(req, 2 * (sizeof xig) + (n + m) * sizeof(struct xtcpcb)); if (error != 0) return (error); xig.xig_len = sizeof xig; xig.xig_count = n + m; xig.xig_gen = gencnt; xig.xig_sogen = so_gencnt; error = SYSCTL_OUT(req, &xig, sizeof xig); if (error) return (error); error = syncache_pcblist(req, m, &pcb_count); if (error) return (error); inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK); if (inp_list == NULL) return (ENOMEM); INP_INFO_RLOCK(&V_tcbinfo); for (inp = LIST_FIRST(V_tcbinfo.ipi_listhead), i = 0; inp != NULL && i < n; inp = LIST_NEXT(inp, inp_list)) { INP_WLOCK(inp); if (inp->inp_gencnt <= gencnt) { /* * XXX: This use of cr_cansee(), introduced with * TCP state changes, is not quite right, but for * now, better than nothing. */ if (inp->inp_flags & INP_TIMEWAIT) { if (intotw(inp) != NULL) error = cr_cansee(req->td->td_ucred, intotw(inp)->tw_cred); else error = EINVAL; /* Skip this inp. */ } else error = cr_canseeinpcb(req->td->td_ucred, inp); if (error == 0) { in_pcbref(inp); inp_list[i++] = inp; } } INP_WUNLOCK(inp); } INP_INFO_RUNLOCK(&V_tcbinfo); n = i; error = 0; for (i = 0; i < n; i++) { inp = inp_list[i]; INP_RLOCK(inp); if (inp->inp_gencnt <= gencnt) { struct xtcpcb xt; void *inp_ppcb; bzero(&xt, sizeof(xt)); xt.xt_len = sizeof xt; /* XXX should avoid extra copy */ bcopy(inp, &xt.xt_inp, sizeof *inp); inp_ppcb = inp->inp_ppcb; if (inp_ppcb == NULL) bzero((char *) &xt.xt_tp, sizeof xt.xt_tp); else if (inp->inp_flags & INP_TIMEWAIT) { bzero((char *) &xt.xt_tp, sizeof xt.xt_tp); xt.xt_tp.t_state = TCPS_TIME_WAIT; } else { bcopy(inp_ppcb, &xt.xt_tp, sizeof xt.xt_tp); if (xt.xt_tp.t_timers) tcp_timer_to_xtimer(&xt.xt_tp, xt.xt_tp.t_timers, &xt.xt_timer); } if (inp->inp_socket != NULL) sotoxsocket(inp->inp_socket, &xt.xt_socket); else { bzero(&xt.xt_socket, sizeof xt.xt_socket); xt.xt_socket.xso_protocol = IPPROTO_TCP; } xt.xt_inp.inp_gencnt = inp->inp_gencnt; INP_RUNLOCK(inp); error = SYSCTL_OUT(req, &xt, sizeof xt); } else INP_RUNLOCK(inp); } INP_INFO_WLOCK(&V_tcbinfo); for (i = 0; i < n; i++) { inp = inp_list[i]; - INP_WLOCK(inp); - if (!in_pcbrele(inp)) - INP_WUNLOCK(inp); + INP_RLOCK(inp); + if (!in_pcbrele_rlocked(inp)) + INP_RUNLOCK(inp); } INP_INFO_WUNLOCK(&V_tcbinfo); if (!error) { /* * Give the user an updated idea of our state. * If the generation differs from what we told * her before, she knows that something happened * while we were processing this request, and it * might be necessary to retry. */ INP_INFO_RLOCK(&V_tcbinfo); xig.xig_gen = V_tcbinfo.ipi_gencnt; xig.xig_sogen = so_gencnt; xig.xig_count = V_tcbinfo.ipi_count + pcb_count; INP_INFO_RUNLOCK(&V_tcbinfo); error = SYSCTL_OUT(req, &xig, sizeof xig); } free(inp_list, M_TEMP); return (error); } SYSCTL_PROC(_net_inet_tcp, TCPCTL_PCBLIST, pcblist, CTLTYPE_OPAQUE | CTLFLAG_RD, NULL, 0, tcp_pcblist, "S,xtcpcb", "List of active TCP connections"); #ifdef INET static int tcp_getcred(SYSCTL_HANDLER_ARGS) { struct xucred xuc; struct sockaddr_in addrs[2]; struct inpcb *inp; int error; error = priv_check(req->td, PRIV_NETINET_GETCRED); if (error) return (error); error = SYSCTL_IN(req, addrs, sizeof(addrs)); if (error) return (error); - INP_INFO_RLOCK(&V_tcbinfo); - inp = in_pcblookup_hash(&V_tcbinfo, addrs[1].sin_addr, - addrs[1].sin_port, addrs[0].sin_addr, addrs[0].sin_port, 0, NULL); + inp = in_pcblookup(&V_tcbinfo, addrs[1].sin_addr, addrs[1].sin_port, + addrs[0].sin_addr, addrs[0].sin_port, INPLOOKUP_RLOCKPCB, NULL); if (inp != NULL) { - INP_RLOCK(inp); - INP_INFO_RUNLOCK(&V_tcbinfo); if (inp->inp_socket == NULL) error = ENOENT; if (error == 0) error = cr_canseeinpcb(req->td->td_ucred, inp); if (error == 0) cru2x(inp->inp_cred, &xuc); INP_RUNLOCK(inp); - } else { - INP_INFO_RUNLOCK(&V_tcbinfo); + } else error = ENOENT; - } if (error == 0) error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred)); return (error); } SYSCTL_PROC(_net_inet_tcp, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW|CTLFLAG_PRISON, 0, 0, tcp_getcred, "S,xucred", "Get the xucred of a TCP connection"); #endif /* INET */ #ifdef INET6 static int tcp6_getcred(SYSCTL_HANDLER_ARGS) { struct xucred xuc; struct sockaddr_in6 addrs[2]; struct inpcb *inp; int error; #ifdef INET int mapped = 0; #endif error = priv_check(req->td, PRIV_NETINET_GETCRED); if (error) return (error); error = SYSCTL_IN(req, addrs, sizeof(addrs)); if (error) return (error); if ((error = sa6_embedscope(&addrs[0], V_ip6_use_defzone)) != 0 || (error = sa6_embedscope(&addrs[1], V_ip6_use_defzone)) != 0) { return (error); } if (IN6_IS_ADDR_V4MAPPED(&addrs[0].sin6_addr)) { #ifdef INET if (IN6_IS_ADDR_V4MAPPED(&addrs[1].sin6_addr)) mapped = 1; else #endif return (EINVAL); } - INP_INFO_RLOCK(&V_tcbinfo); #ifdef INET if (mapped == 1) - inp = in_pcblookup_hash(&V_tcbinfo, + inp = in_pcblookup(&V_tcbinfo, *(struct in_addr *)&addrs[1].sin6_addr.s6_addr[12], addrs[1].sin6_port, *(struct in_addr *)&addrs[0].sin6_addr.s6_addr[12], - addrs[0].sin6_port, - 0, NULL); + addrs[0].sin6_port, INPLOOKUP_RLOCKPCB, NULL); else #endif - inp = in6_pcblookup_hash(&V_tcbinfo, + inp = in6_pcblookup(&V_tcbinfo, &addrs[1].sin6_addr, addrs[1].sin6_port, - &addrs[0].sin6_addr, addrs[0].sin6_port, 0, NULL); + &addrs[0].sin6_addr, addrs[0].sin6_port, + INPLOOKUP_RLOCKPCB, NULL); if (inp != NULL) { - INP_RLOCK(inp); - INP_INFO_RUNLOCK(&V_tcbinfo); if (inp->inp_socket == NULL) error = ENOENT; if (error == 0) error = cr_canseeinpcb(req->td->td_ucred, inp); if (error == 0) cru2x(inp->inp_cred, &xuc); INP_RUNLOCK(inp); - } else { - INP_INFO_RUNLOCK(&V_tcbinfo); + } else error = ENOENT; - } if (error == 0) error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred)); return (error); } SYSCTL_PROC(_net_inet6_tcp6, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW|CTLFLAG_PRISON, 0, 0, tcp6_getcred, "S,xucred", "Get the xucred of a TCP6 connection"); #endif /* INET6 */ #ifdef INET void tcp_ctlinput(int cmd, struct sockaddr *sa, void *vip) { struct ip *ip = vip; struct tcphdr *th; struct in_addr faddr; struct inpcb *inp; struct tcpcb *tp; struct inpcb *(*notify)(struct inpcb *, int) = tcp_notify; struct icmp *icp; struct in_conninfo inc; tcp_seq icmp_tcp_seq; int mtu; faddr = ((struct sockaddr_in *)sa)->sin_addr; if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY) return; if (cmd == PRC_MSGSIZE) notify = tcp_mtudisc; else if (V_icmp_may_rst && (cmd == PRC_UNREACH_ADMIN_PROHIB || cmd == PRC_UNREACH_PORT || cmd == PRC_TIMXCEED_INTRANS) && ip) notify = tcp_drop_syn_sent; /* * Redirects don't need to be handled up here. */ else if (PRC_IS_REDIRECT(cmd)) return; /* * Source quench is depreciated. */ else if (cmd == PRC_QUENCH) return; /* * Hostdead is ugly because it goes linearly through all PCBs. * XXX: We never get this from ICMP, otherwise it makes an * excellent DoS attack on machines with many connections. */ else if (cmd == PRC_HOSTDEAD) ip = NULL; else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0) return; if (ip != NULL) { icp = (struct icmp *)((caddr_t)ip - offsetof(struct icmp, icmp_ip)); th = (struct tcphdr *)((caddr_t)ip + (ip->ip_hl << 2)); INP_INFO_WLOCK(&V_tcbinfo); - inp = in_pcblookup_hash(&V_tcbinfo, faddr, th->th_dport, - ip->ip_src, th->th_sport, 0, NULL); + inp = in_pcblookup(&V_tcbinfo, faddr, th->th_dport, + ip->ip_src, th->th_sport, INPLOOKUP_WLOCKPCB, NULL); if (inp != NULL) { - INP_WLOCK(inp); if (!(inp->inp_flags & INP_TIMEWAIT) && !(inp->inp_flags & INP_DROPPED) && !(inp->inp_socket == NULL)) { icmp_tcp_seq = htonl(th->th_seq); tp = intotcpcb(inp); if (SEQ_GEQ(icmp_tcp_seq, tp->snd_una) && SEQ_LT(icmp_tcp_seq, tp->snd_max)) { if (cmd == PRC_MSGSIZE) { /* * MTU discovery: * If we got a needfrag set the MTU * in the route to the suggested new * value (if given) and then notify. */ bzero(&inc, sizeof(inc)); inc.inc_faddr = faddr; inc.inc_fibnum = inp->inp_inc.inc_fibnum; mtu = ntohs(icp->icmp_nextmtu); /* * If no alternative MTU was * proposed, try the next smaller * one. ip->ip_len has already * been swapped in icmp_input(). */ if (!mtu) mtu = ip_next_mtu(ip->ip_len, 1); if (mtu < V_tcp_minmss + sizeof(struct tcpiphdr)) mtu = V_tcp_minmss + sizeof(struct tcpiphdr); /* * Only cache the MTU if it * is smaller than the interface * or route MTU. tcp_mtudisc() * will do right thing by itself. */ if (mtu <= tcp_maxmtu(&inc, NULL)) tcp_hc_updatemtu(&inc, mtu); } inp = (*notify)(inp, inetctlerrmap[cmd]); } } if (inp != NULL) INP_WUNLOCK(inp); } else { bzero(&inc, sizeof(inc)); inc.inc_fport = th->th_dport; inc.inc_lport = th->th_sport; inc.inc_faddr = faddr; inc.inc_laddr = ip->ip_src; syncache_unreach(&inc, th); } INP_INFO_WUNLOCK(&V_tcbinfo); } else in_pcbnotifyall(&V_tcbinfo, faddr, inetctlerrmap[cmd], notify); } #endif /* INET */ #ifdef INET6 void tcp6_ctlinput(int cmd, struct sockaddr *sa, void *d) { struct tcphdr th; struct inpcb *(*notify)(struct inpcb *, int) = tcp_notify; struct ip6_hdr *ip6; struct mbuf *m; struct ip6ctlparam *ip6cp = NULL; const struct sockaddr_in6 *sa6_src = NULL; int off; struct tcp_portonly { u_int16_t th_sport; u_int16_t th_dport; } *thp; if (sa->sa_family != AF_INET6 || sa->sa_len != sizeof(struct sockaddr_in6)) return; if (cmd == PRC_MSGSIZE) notify = tcp_mtudisc; else if (!PRC_IS_REDIRECT(cmd) && ((unsigned)cmd >= PRC_NCMDS || inet6ctlerrmap[cmd] == 0)) return; /* Source quench is depreciated. */ else if (cmd == PRC_QUENCH) return; /* if the parameter is from icmp6, decode it. */ if (d != NULL) { ip6cp = (struct ip6ctlparam *)d; m = ip6cp->ip6c_m; ip6 = ip6cp->ip6c_ip6; off = ip6cp->ip6c_off; sa6_src = ip6cp->ip6c_src; } else { m = NULL; ip6 = NULL; off = 0; /* fool gcc */ sa6_src = &sa6_any; } if (ip6 != NULL) { struct in_conninfo inc; /* * XXX: We assume that when IPV6 is non NULL, * M and OFF are valid. */ /* check if we can safely examine src and dst ports */ if (m->m_pkthdr.len < off + sizeof(*thp)) return; bzero(&th, sizeof(th)); m_copydata(m, off, sizeof(*thp), (caddr_t)&th); in6_pcbnotify(&V_tcbinfo, sa, th.th_dport, (struct sockaddr *)ip6cp->ip6c_src, th.th_sport, cmd, NULL, notify); bzero(&inc, sizeof(inc)); inc.inc_fport = th.th_dport; inc.inc_lport = th.th_sport; inc.inc6_faddr = ((struct sockaddr_in6 *)sa)->sin6_addr; inc.inc6_laddr = ip6cp->ip6c_src->sin6_addr; inc.inc_flags |= INC_ISIPV6; INP_INFO_WLOCK(&V_tcbinfo); syncache_unreach(&inc, &th); INP_INFO_WUNLOCK(&V_tcbinfo); } else in6_pcbnotify(&V_tcbinfo, sa, 0, (const struct sockaddr *)sa6_src, 0, cmd, NULL, notify); } #endif /* INET6 */ /* * Following is where TCP initial sequence number generation occurs. * * There are two places where we must use initial sequence numbers: * 1. In SYN-ACK packets. * 2. In SYN packets. * * All ISNs for SYN-ACK packets are generated by the syncache. See * tcp_syncache.c for details. * * The ISNs in SYN packets must be monotonic; TIME_WAIT recycling * depends on this property. In addition, these ISNs should be * unguessable so as to prevent connection hijacking. To satisfy * the requirements of this situation, the algorithm outlined in * RFC 1948 is used, with only small modifications. * * Implementation details: * * Time is based off the system timer, and is corrected so that it * increases by one megabyte per second. This allows for proper * recycling on high speed LANs while still leaving over an hour * before rollover. * * As reading the *exact* system time is too expensive to be done * whenever setting up a TCP connection, we increment the time * offset in two ways. First, a small random positive increment * is added to isn_offset for each connection that is set up. * Second, the function tcp_isn_tick fires once per clock tick * and increments isn_offset as necessary so that sequence numbers * are incremented at approximately ISN_BYTES_PER_SECOND. The * random positive increments serve only to ensure that the same * exact sequence number is never sent out twice (as could otherwise * happen when a port is recycled in less than the system tick * interval.) * * net.inet.tcp.isn_reseed_interval controls the number of seconds * between seeding of isn_secret. This is normally set to zero, * as reseeding should not be necessary. * * Locking of the global variables isn_secret, isn_last_reseed, isn_offset, * isn_offset_old, and isn_ctx is performed using the TCP pcbinfo lock. In * general, this means holding an exclusive (write) lock. */ #define ISN_BYTES_PER_SECOND 1048576 #define ISN_STATIC_INCREMENT 4096 #define ISN_RANDOM_INCREMENT (4096 - 1) static VNET_DEFINE(u_char, isn_secret[32]); static VNET_DEFINE(int, isn_last); static VNET_DEFINE(int, isn_last_reseed); static VNET_DEFINE(u_int32_t, isn_offset); static VNET_DEFINE(u_int32_t, isn_offset_old); #define V_isn_secret VNET(isn_secret) #define V_isn_last VNET(isn_last) #define V_isn_last_reseed VNET(isn_last_reseed) #define V_isn_offset VNET(isn_offset) #define V_isn_offset_old VNET(isn_offset_old) tcp_seq tcp_new_isn(struct tcpcb *tp) { MD5_CTX isn_ctx; u_int32_t md5_buffer[4]; tcp_seq new_isn; u_int32_t projected_offset; INP_WLOCK_ASSERT(tp->t_inpcb); ISN_LOCK(); /* Seed if this is the first use, reseed if requested. */ if ((V_isn_last_reseed == 0) || ((V_tcp_isn_reseed_interval > 0) && (((u_int)V_isn_last_reseed + (u_int)V_tcp_isn_reseed_interval*hz) < (u_int)ticks))) { read_random(&V_isn_secret, sizeof(V_isn_secret)); V_isn_last_reseed = ticks; } /* Compute the md5 hash and return the ISN. */ MD5Init(&isn_ctx); MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_fport, sizeof(u_short)); MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_lport, sizeof(u_short)); #ifdef INET6 if ((tp->t_inpcb->inp_vflag & INP_IPV6) != 0) { MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->in6p_faddr, sizeof(struct in6_addr)); MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->in6p_laddr, sizeof(struct in6_addr)); } else #endif { MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_faddr, sizeof(struct in_addr)); MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_laddr, sizeof(struct in_addr)); } MD5Update(&isn_ctx, (u_char *) &V_isn_secret, sizeof(V_isn_secret)); MD5Final((u_char *) &md5_buffer, &isn_ctx); new_isn = (tcp_seq) md5_buffer[0]; V_isn_offset += ISN_STATIC_INCREMENT + (arc4random() & ISN_RANDOM_INCREMENT); if (ticks != V_isn_last) { projected_offset = V_isn_offset_old + ISN_BYTES_PER_SECOND / hz * (ticks - V_isn_last); if (SEQ_GT(projected_offset, V_isn_offset)) V_isn_offset = projected_offset; V_isn_offset_old = V_isn_offset; V_isn_last = ticks; } new_isn += V_isn_offset; ISN_UNLOCK(); return (new_isn); } /* * When a specific ICMP unreachable message is received and the * connection state is SYN-SENT, drop the connection. This behavior * is controlled by the icmp_may_rst sysctl. */ struct inpcb * tcp_drop_syn_sent(struct inpcb *inp, int errno) { struct tcpcb *tp; INP_INFO_WLOCK_ASSERT(&V_tcbinfo); INP_WLOCK_ASSERT(inp); if ((inp->inp_flags & INP_TIMEWAIT) || (inp->inp_flags & INP_DROPPED)) return (inp); tp = intotcpcb(inp); if (tp->t_state != TCPS_SYN_SENT) return (inp); tp = tcp_drop(tp, errno); if (tp != NULL) return (inp); else return (NULL); } /* * When `need fragmentation' ICMP is received, update our idea of the MSS * based on the new value in the route. Also nudge TCP to send something, * since we know the packet we just sent was dropped. * This duplicates some code in the tcp_mss() function in tcp_input.c. */ struct inpcb * tcp_mtudisc(struct inpcb *inp, int errno) { struct tcpcb *tp; struct socket *so; INP_WLOCK_ASSERT(inp); if ((inp->inp_flags & INP_TIMEWAIT) || (inp->inp_flags & INP_DROPPED)) return (inp); tp = intotcpcb(inp); KASSERT(tp != NULL, ("tcp_mtudisc: tp == NULL")); tcp_mss_update(tp, -1, NULL, NULL); so = inp->inp_socket; SOCKBUF_LOCK(&so->so_snd); /* If the mss is larger than the socket buffer, decrease the mss. */ if (so->so_snd.sb_hiwat < tp->t_maxseg) tp->t_maxseg = so->so_snd.sb_hiwat; SOCKBUF_UNLOCK(&so->so_snd); TCPSTAT_INC(tcps_mturesent); tp->t_rtttime = 0; tp->snd_nxt = tp->snd_una; tcp_free_sackholes(tp); tp->snd_recover = tp->snd_max; if (tp->t_flags & TF_SACK_PERMIT) EXIT_FASTRECOVERY(tp->t_flags); tcp_output_send(tp); return (inp); } #ifdef INET /* * Look-up the routing entry to the peer of this inpcb. If no route * is found and it cannot be allocated, then return 0. This routine * is called by TCP routines that access the rmx structure and by * tcp_mss_update to get the peer/interface MTU. */ u_long tcp_maxmtu(struct in_conninfo *inc, int *flags) { struct route sro; struct sockaddr_in *dst; struct ifnet *ifp; u_long maxmtu = 0; KASSERT(inc != NULL, ("tcp_maxmtu with NULL in_conninfo pointer")); bzero(&sro, sizeof(sro)); if (inc->inc_faddr.s_addr != INADDR_ANY) { dst = (struct sockaddr_in *)&sro.ro_dst; dst->sin_family = AF_INET; dst->sin_len = sizeof(*dst); dst->sin_addr = inc->inc_faddr; in_rtalloc_ign(&sro, 0, inc->inc_fibnum); } if (sro.ro_rt != NULL) { ifp = sro.ro_rt->rt_ifp; if (sro.ro_rt->rt_rmx.rmx_mtu == 0) maxmtu = ifp->if_mtu; else maxmtu = min(sro.ro_rt->rt_rmx.rmx_mtu, ifp->if_mtu); /* Report additional interface capabilities. */ if (flags != NULL) { if (ifp->if_capenable & IFCAP_TSO4 && ifp->if_hwassist & CSUM_TSO) *flags |= CSUM_TSO; } RTFREE(sro.ro_rt); } return (maxmtu); } #endif /* INET */ #ifdef INET6 u_long tcp_maxmtu6(struct in_conninfo *inc, int *flags) { struct route_in6 sro6; struct ifnet *ifp; u_long maxmtu = 0; KASSERT(inc != NULL, ("tcp_maxmtu6 with NULL in_conninfo pointer")); bzero(&sro6, sizeof(sro6)); if (!IN6_IS_ADDR_UNSPECIFIED(&inc->inc6_faddr)) { sro6.ro_dst.sin6_family = AF_INET6; sro6.ro_dst.sin6_len = sizeof(struct sockaddr_in6); sro6.ro_dst.sin6_addr = inc->inc6_faddr; rtalloc_ign((struct route *)&sro6, 0); } if (sro6.ro_rt != NULL) { ifp = sro6.ro_rt->rt_ifp; if (sro6.ro_rt->rt_rmx.rmx_mtu == 0) maxmtu = IN6_LINKMTU(sro6.ro_rt->rt_ifp); else maxmtu = min(sro6.ro_rt->rt_rmx.rmx_mtu, IN6_LINKMTU(sro6.ro_rt->rt_ifp)); /* Report additional interface capabilities. */ if (flags != NULL) { if (ifp->if_capenable & IFCAP_TSO6 && ifp->if_hwassist & CSUM_TSO) *flags |= CSUM_TSO; } RTFREE(sro6.ro_rt); } return (maxmtu); } #endif /* INET6 */ #ifdef IPSEC /* compute ESP/AH header size for TCP, including outer IP header. */ size_t ipsec_hdrsiz_tcp(struct tcpcb *tp) { struct inpcb *inp; struct mbuf *m; size_t hdrsiz; struct ip *ip; #ifdef INET6 struct ip6_hdr *ip6; #endif struct tcphdr *th; if ((tp == NULL) || ((inp = tp->t_inpcb) == NULL)) return (0); MGETHDR(m, M_DONTWAIT, MT_DATA); if (!m) return (0); #ifdef INET6 if ((inp->inp_vflag & INP_IPV6) != 0) { ip6 = mtod(m, struct ip6_hdr *); th = (struct tcphdr *)(ip6 + 1); m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + sizeof(struct tcphdr); tcpip_fillheaders(inp, ip6, th); hdrsiz = ipsec_hdrsiz(m, IPSEC_DIR_OUTBOUND, inp); } else #endif /* INET6 */ { ip = mtod(m, struct ip *); th = (struct tcphdr *)(ip + 1); m->m_pkthdr.len = m->m_len = sizeof(struct tcpiphdr); tcpip_fillheaders(inp, ip, th); hdrsiz = ipsec_hdrsiz(m, IPSEC_DIR_OUTBOUND, inp); } m_free(m); return (hdrsiz); } #endif /* IPSEC */ #ifdef TCP_SIGNATURE /* * Callback function invoked by m_apply() to digest TCP segment data * contained within an mbuf chain. */ static int tcp_signature_apply(void *fstate, void *data, u_int len) { MD5Update(fstate, (u_char *)data, len); return (0); } /* * Compute TCP-MD5 hash of a TCP segment. (RFC2385) * * Parameters: * m pointer to head of mbuf chain * _unused * len length of TCP segment data, excluding options * optlen length of TCP segment options * buf pointer to storage for computed MD5 digest * direction direction of flow (IPSEC_DIR_INBOUND or OUTBOUND) * * We do this over ip, tcphdr, segment data, and the key in the SADB. * When called from tcp_input(), we can be sure that th_sum has been * zeroed out and verified already. * * Return 0 if successful, otherwise return -1. * * XXX The key is retrieved from the system's PF_KEY SADB, by keying a * search with the destination IP address, and a 'magic SPI' to be * determined by the application. This is hardcoded elsewhere to 1179 * right now. Another branch of this code exists which uses the SPD to * specify per-application flows but it is unstable. */ int tcp_signature_compute(struct mbuf *m, int _unused, int len, int optlen, u_char *buf, u_int direction) { union sockaddr_union dst; #ifdef INET struct ippseudo ippseudo; #endif MD5_CTX ctx; int doff; struct ip *ip; #ifdef INET struct ipovly *ipovly; #endif struct secasvar *sav; struct tcphdr *th; #ifdef INET6 struct ip6_hdr *ip6; struct in6_addr in6; char ip6buf[INET6_ADDRSTRLEN]; uint32_t plen; uint16_t nhdr; #endif u_short savecsum; KASSERT(m != NULL, ("NULL mbuf chain")); KASSERT(buf != NULL, ("NULL signature pointer")); /* Extract the destination from the IP header in the mbuf. */ bzero(&dst, sizeof(union sockaddr_union)); ip = mtod(m, struct ip *); #ifdef INET6 ip6 = NULL; /* Make the compiler happy. */ #endif switch (ip->ip_v) { #ifdef INET case IPVERSION: dst.sa.sa_len = sizeof(struct sockaddr_in); dst.sa.sa_family = AF_INET; dst.sin.sin_addr = (direction == IPSEC_DIR_INBOUND) ? ip->ip_src : ip->ip_dst; break; #endif #ifdef INET6 case (IPV6_VERSION >> 4): ip6 = mtod(m, struct ip6_hdr *); dst.sa.sa_len = sizeof(struct sockaddr_in6); dst.sa.sa_family = AF_INET6; dst.sin6.sin6_addr = (direction == IPSEC_DIR_INBOUND) ? ip6->ip6_src : ip6->ip6_dst; break; #endif default: return (EINVAL); /* NOTREACHED */ break; } /* Look up an SADB entry which matches the address of the peer. */ sav = KEY_ALLOCSA(&dst, IPPROTO_TCP, htonl(TCP_SIG_SPI)); if (sav == NULL) { ipseclog((LOG_ERR, "%s: SADB lookup failed for %s\n", __func__, (ip->ip_v == IPVERSION) ? inet_ntoa(dst.sin.sin_addr) : #ifdef INET6 (ip->ip_v == (IPV6_VERSION >> 4)) ? ip6_sprintf(ip6buf, &dst.sin6.sin6_addr) : #endif "(unsupported)")); return (EINVAL); } MD5Init(&ctx); /* * Step 1: Update MD5 hash with IP(v6) pseudo-header. * * XXX The ippseudo header MUST be digested in network byte order, * or else we'll fail the regression test. Assume all fields we've * been doing arithmetic on have been in host byte order. * XXX One cannot depend on ipovly->ih_len here. When called from * tcp_output(), the underlying ip_len member has not yet been set. */ switch (ip->ip_v) { #ifdef INET case IPVERSION: ipovly = (struct ipovly *)ip; ippseudo.ippseudo_src = ipovly->ih_src; ippseudo.ippseudo_dst = ipovly->ih_dst; ippseudo.ippseudo_pad = 0; ippseudo.ippseudo_p = IPPROTO_TCP; ippseudo.ippseudo_len = htons(len + sizeof(struct tcphdr) + optlen); MD5Update(&ctx, (char *)&ippseudo, sizeof(struct ippseudo)); th = (struct tcphdr *)((u_char *)ip + sizeof(struct ip)); doff = sizeof(struct ip) + sizeof(struct tcphdr) + optlen; break; #endif #ifdef INET6 /* * RFC 2385, 2.0 Proposal * For IPv6, the pseudo-header is as described in RFC 2460, namely the * 128-bit source IPv6 address, 128-bit destination IPv6 address, zero- * extended next header value (to form 32 bits), and 32-bit segment * length. * Note: Upper-Layer Packet Length comes before Next Header. */ case (IPV6_VERSION >> 4): in6 = ip6->ip6_src; in6_clearscope(&in6); MD5Update(&ctx, (char *)&in6, sizeof(struct in6_addr)); in6 = ip6->ip6_dst; in6_clearscope(&in6); MD5Update(&ctx, (char *)&in6, sizeof(struct in6_addr)); plen = htonl(len + sizeof(struct tcphdr) + optlen); MD5Update(&ctx, (char *)&plen, sizeof(uint32_t)); nhdr = 0; MD5Update(&ctx, (char *)&nhdr, sizeof(uint8_t)); MD5Update(&ctx, (char *)&nhdr, sizeof(uint8_t)); MD5Update(&ctx, (char *)&nhdr, sizeof(uint8_t)); nhdr = IPPROTO_TCP; MD5Update(&ctx, (char *)&nhdr, sizeof(uint8_t)); th = (struct tcphdr *)((u_char *)ip6 + sizeof(struct ip6_hdr)); doff = sizeof(struct ip6_hdr) + sizeof(struct tcphdr) + optlen; break; #endif default: return (EINVAL); /* NOTREACHED */ break; } /* * Step 2: Update MD5 hash with TCP header, excluding options. * The TCP checksum must be set to zero. */ savecsum = th->th_sum; th->th_sum = 0; MD5Update(&ctx, (char *)th, sizeof(struct tcphdr)); th->th_sum = savecsum; /* * Step 3: Update MD5 hash with TCP segment data. * Use m_apply() to avoid an early m_pullup(). */ if (len > 0) m_apply(m, doff, len, tcp_signature_apply, &ctx); /* * Step 4: Update MD5 hash with shared secret. */ MD5Update(&ctx, sav->key_auth->key_data, _KEYLEN(sav->key_auth)); MD5Final(buf, &ctx); key_sa_recordxfer(sav, m); KEY_FREESAV(&sav); return (0); } /* * Verify the TCP-MD5 hash of a TCP segment. (RFC2385) * * Parameters: * m pointer to head of mbuf chain * len length of TCP segment data, excluding options * optlen length of TCP segment options * buf pointer to storage for computed MD5 digest * direction direction of flow (IPSEC_DIR_INBOUND or OUTBOUND) * * Return 1 if successful, otherwise return 0. */ int tcp_signature_verify(struct mbuf *m, int off0, int tlen, int optlen, struct tcpopt *to, struct tcphdr *th, u_int tcpbflag) { char tmpdigest[TCP_SIGLEN]; if (tcp_sig_checksigs == 0) return (1); if ((tcpbflag & TF_SIGNATURE) == 0) { if ((to->to_flags & TOF_SIGNATURE) != 0) { /* * If this socket is not expecting signature but * the segment contains signature just fail. */ TCPSTAT_INC(tcps_sig_err_sigopt); TCPSTAT_INC(tcps_sig_rcvbadsig); return (0); } /* Signature is not expected, and not present in segment. */ return (1); } /* * If this socket is expecting signature but the segment does not * contain any just fail. */ if ((to->to_flags & TOF_SIGNATURE) == 0) { TCPSTAT_INC(tcps_sig_err_nosigopt); TCPSTAT_INC(tcps_sig_rcvbadsig); return (0); } if (tcp_signature_compute(m, off0, tlen, optlen, &tmpdigest[0], IPSEC_DIR_INBOUND) == -1) { TCPSTAT_INC(tcps_sig_err_buildsig); TCPSTAT_INC(tcps_sig_rcvbadsig); return (0); } if (bcmp(to->to_signature, &tmpdigest[0], TCP_SIGLEN) != 0) { TCPSTAT_INC(tcps_sig_rcvbadsig); return (0); } TCPSTAT_INC(tcps_sig_rcvgoodsig); return (1); } #endif /* TCP_SIGNATURE */ static int sysctl_drop(SYSCTL_HANDLER_ARGS) { /* addrs[0] is a foreign socket, addrs[1] is a local one. */ struct sockaddr_storage addrs[2]; struct inpcb *inp; struct tcpcb *tp; struct tcptw *tw; struct sockaddr_in *fin, *lin; #ifdef INET6 struct sockaddr_in6 *fin6, *lin6; #endif int error; inp = NULL; fin = lin = NULL; #ifdef INET6 fin6 = lin6 = NULL; #endif error = 0; if (req->oldptr != NULL || req->oldlen != 0) return (EINVAL); if (req->newptr == NULL) return (EPERM); if (req->newlen < sizeof(addrs)) return (ENOMEM); error = SYSCTL_IN(req, &addrs, sizeof(addrs)); if (error) return (error); switch (addrs[0].ss_family) { #ifdef INET6 case AF_INET6: fin6 = (struct sockaddr_in6 *)&addrs[0]; lin6 = (struct sockaddr_in6 *)&addrs[1]; if (fin6->sin6_len != sizeof(struct sockaddr_in6) || lin6->sin6_len != sizeof(struct sockaddr_in6)) return (EINVAL); if (IN6_IS_ADDR_V4MAPPED(&fin6->sin6_addr)) { if (!IN6_IS_ADDR_V4MAPPED(&lin6->sin6_addr)) return (EINVAL); in6_sin6_2_sin_in_sock((struct sockaddr *)&addrs[0]); in6_sin6_2_sin_in_sock((struct sockaddr *)&addrs[1]); fin = (struct sockaddr_in *)&addrs[0]; lin = (struct sockaddr_in *)&addrs[1]; break; } error = sa6_embedscope(fin6, V_ip6_use_defzone); if (error) return (error); error = sa6_embedscope(lin6, V_ip6_use_defzone); if (error) return (error); break; #endif #ifdef INET case AF_INET: fin = (struct sockaddr_in *)&addrs[0]; lin = (struct sockaddr_in *)&addrs[1]; if (fin->sin_len != sizeof(struct sockaddr_in) || lin->sin_len != sizeof(struct sockaddr_in)) return (EINVAL); break; #endif default: return (EINVAL); } INP_INFO_WLOCK(&V_tcbinfo); switch (addrs[0].ss_family) { #ifdef INET6 case AF_INET6: - inp = in6_pcblookup_hash(&V_tcbinfo, &fin6->sin6_addr, - fin6->sin6_port, &lin6->sin6_addr, lin6->sin6_port, 0, - NULL); + inp = in6_pcblookup(&V_tcbinfo, &fin6->sin6_addr, + fin6->sin6_port, &lin6->sin6_addr, lin6->sin6_port, + INPLOOKUP_WLOCKPCB, NULL); break; #endif #ifdef INET case AF_INET: - inp = in_pcblookup_hash(&V_tcbinfo, fin->sin_addr, - fin->sin_port, lin->sin_addr, lin->sin_port, 0, NULL); + inp = in_pcblookup(&V_tcbinfo, fin->sin_addr, fin->sin_port, + lin->sin_addr, lin->sin_port, INPLOOKUP_WLOCKPCB, NULL); break; #endif } if (inp != NULL) { - INP_WLOCK(inp); if (inp->inp_flags & INP_TIMEWAIT) { /* * XXXRW: There currently exists a state where an * inpcb is present, but its timewait state has been * discarded. For now, don't allow dropping of this * type of inpcb. */ tw = intotw(inp); if (tw != NULL) tcp_twclose(tw, 0); else INP_WUNLOCK(inp); } else if (!(inp->inp_flags & INP_DROPPED) && !(inp->inp_socket->so_options & SO_ACCEPTCONN)) { tp = intotcpcb(inp); tp = tcp_drop(tp, ECONNABORTED); if (tp != NULL) INP_WUNLOCK(inp); } else INP_WUNLOCK(inp); } else error = ESRCH; INP_INFO_WUNLOCK(&V_tcbinfo); return (error); } SYSCTL_PROC(_net_inet_tcp, TCPCTL_DROP, drop, CTLTYPE_STRUCT|CTLFLAG_WR|CTLFLAG_SKIP, NULL, 0, sysctl_drop, "", "Drop TCP connection"); /* * Generate a standardized TCP log line for use throughout the * tcp subsystem. Memory allocation is done with M_NOWAIT to * allow use in the interrupt context. * * NB: The caller MUST free(s, M_TCPLOG) the returned string. * NB: The function may return NULL if memory allocation failed. * * Due to header inclusion and ordering limitations the struct ip * and ip6_hdr pointers have to be passed as void pointers. */ char * tcp_log_vain(struct in_conninfo *inc, struct tcphdr *th, void *ip4hdr, const void *ip6hdr) { /* Is logging enabled? */ if (tcp_log_in_vain == 0) return (NULL); return (tcp_log_addr(inc, th, ip4hdr, ip6hdr)); } char * tcp_log_addrs(struct in_conninfo *inc, struct tcphdr *th, void *ip4hdr, const void *ip6hdr) { /* Is logging enabled? */ if (tcp_log_debug == 0) return (NULL); return (tcp_log_addr(inc, th, ip4hdr, ip6hdr)); } static char * tcp_log_addr(struct in_conninfo *inc, struct tcphdr *th, void *ip4hdr, const void *ip6hdr) { char *s, *sp; size_t size; struct ip *ip; #ifdef INET6 const struct ip6_hdr *ip6; ip6 = (const struct ip6_hdr *)ip6hdr; #endif /* INET6 */ ip = (struct ip *)ip4hdr; /* * The log line looks like this: * "TCP: [1.2.3.4]:50332 to [1.2.3.4]:80 tcpflags 0x2" */ size = sizeof("TCP: []:12345 to []:12345 tcpflags 0x2<>") + sizeof(PRINT_TH_FLAGS) + 1 + #ifdef INET6 2 * INET6_ADDRSTRLEN; #else 2 * INET_ADDRSTRLEN; #endif /* INET6 */ s = malloc(size, M_TCPLOG, M_ZERO|M_NOWAIT); if (s == NULL) return (NULL); strcat(s, "TCP: ["); sp = s + strlen(s); if (inc && ((inc->inc_flags & INC_ISIPV6) == 0)) { inet_ntoa_r(inc->inc_faddr, sp); sp = s + strlen(s); sprintf(sp, "]:%i to [", ntohs(inc->inc_fport)); sp = s + strlen(s); inet_ntoa_r(inc->inc_laddr, sp); sp = s + strlen(s); sprintf(sp, "]:%i", ntohs(inc->inc_lport)); #ifdef INET6 } else if (inc) { ip6_sprintf(sp, &inc->inc6_faddr); sp = s + strlen(s); sprintf(sp, "]:%i to [", ntohs(inc->inc_fport)); sp = s + strlen(s); ip6_sprintf(sp, &inc->inc6_laddr); sp = s + strlen(s); sprintf(sp, "]:%i", ntohs(inc->inc_lport)); } else if (ip6 && th) { ip6_sprintf(sp, &ip6->ip6_src); sp = s + strlen(s); sprintf(sp, "]:%i to [", ntohs(th->th_sport)); sp = s + strlen(s); ip6_sprintf(sp, &ip6->ip6_dst); sp = s + strlen(s); sprintf(sp, "]:%i", ntohs(th->th_dport)); #endif /* INET6 */ #ifdef INET } else if (ip && th) { inet_ntoa_r(ip->ip_src, sp); sp = s + strlen(s); sprintf(sp, "]:%i to [", ntohs(th->th_sport)); sp = s + strlen(s); inet_ntoa_r(ip->ip_dst, sp); sp = s + strlen(s); sprintf(sp, "]:%i", ntohs(th->th_dport)); #endif /* INET */ } else { free(s, M_TCPLOG); return (NULL); } sp = s + strlen(s); if (th) sprintf(sp, " tcpflags 0x%b", th->th_flags, PRINT_TH_FLAGS); if (*(s + size - 1) != '\0') panic("%s: string too long", __func__); return (s); } Index: head/sys/netinet/tcp_syncache.c =================================================================== --- head/sys/netinet/tcp_syncache.c (revision 222487) +++ head/sys/netinet/tcp_syncache.c (revision 222488) @@ -1,1830 +1,1835 @@ /*- * Copyright (c) 2001 McAfee, Inc. * Copyright (c) 2006 Andre Oppermann, Internet Business Solutions AG * All rights reserved. * * This software was developed for the FreeBSD Project by Jonathan Lemon * and McAfee Research, the Security Research Division of McAfee, Inc. under * DARPA/SPAWAR contract N66001-01-C-8035 ("CBOSS"), as part of the * DARPA CHATS research program. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include "opt_inet.h" #include "opt_inet6.h" #include "opt_ipsec.h" #include #include #include #include #include #include #include #include #include #include #include /* for proc0 declaration */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef INET6 #include #include #include #include #include #endif #include #include #include #include #include #include #include #ifdef INET6 #include #endif #ifdef IPSEC #include #ifdef INET6 #include #endif #include #endif /*IPSEC*/ #include #include static VNET_DEFINE(int, tcp_syncookies) = 1; #define V_tcp_syncookies VNET(tcp_syncookies) SYSCTL_VNET_INT(_net_inet_tcp, OID_AUTO, syncookies, CTLFLAG_RW, &VNET_NAME(tcp_syncookies), 0, "Use TCP SYN cookies if the syncache overflows"); static VNET_DEFINE(int, tcp_syncookiesonly) = 0; #define V_tcp_syncookiesonly VNET(tcp_syncookiesonly) SYSCTL_VNET_INT(_net_inet_tcp, OID_AUTO, syncookies_only, CTLFLAG_RW, &VNET_NAME(tcp_syncookiesonly), 0, "Use only TCP SYN cookies"); #ifdef TCP_OFFLOAD_DISABLE #define TOEPCB_ISSET(sc) (0) #else #define TOEPCB_ISSET(sc) ((sc)->sc_toepcb != NULL) #endif static void syncache_drop(struct syncache *, struct syncache_head *); static void syncache_free(struct syncache *); static void syncache_insert(struct syncache *, struct syncache_head *); struct syncache *syncache_lookup(struct in_conninfo *, struct syncache_head **); static int syncache_respond(struct syncache *); static struct socket *syncache_socket(struct syncache *, struct socket *, struct mbuf *m); static void syncache_timeout(struct syncache *sc, struct syncache_head *sch, int docallout); static void syncache_timer(void *); static void syncookie_generate(struct syncache_head *, struct syncache *, u_int32_t *); static struct syncache *syncookie_lookup(struct in_conninfo *, struct syncache_head *, struct syncache *, struct tcpopt *, struct tcphdr *, struct socket *); /* * Transmit the SYN,ACK fewer times than TCP_MAXRXTSHIFT specifies. * 3 retransmits corresponds to a timeout of 3 * (1 + 2 + 4 + 8) == 45 seconds, * the odds are that the user has given up attempting to connect by then. */ #define SYNCACHE_MAXREXMTS 3 /* Arbitrary values */ #define TCP_SYNCACHE_HASHSIZE 512 #define TCP_SYNCACHE_BUCKETLIMIT 30 static VNET_DEFINE(struct tcp_syncache, tcp_syncache); #define V_tcp_syncache VNET(tcp_syncache) SYSCTL_NODE(_net_inet_tcp, OID_AUTO, syncache, CTLFLAG_RW, 0, "TCP SYN cache"); SYSCTL_VNET_UINT(_net_inet_tcp_syncache, OID_AUTO, bucketlimit, CTLFLAG_RDTUN, &VNET_NAME(tcp_syncache.bucket_limit), 0, "Per-bucket hash limit for syncache"); SYSCTL_VNET_UINT(_net_inet_tcp_syncache, OID_AUTO, cachelimit, CTLFLAG_RDTUN, &VNET_NAME(tcp_syncache.cache_limit), 0, "Overall entry limit for syncache"); SYSCTL_VNET_UINT(_net_inet_tcp_syncache, OID_AUTO, count, CTLFLAG_RD, &VNET_NAME(tcp_syncache.cache_count), 0, "Current number of entries in syncache"); SYSCTL_VNET_UINT(_net_inet_tcp_syncache, OID_AUTO, hashsize, CTLFLAG_RDTUN, &VNET_NAME(tcp_syncache.hashsize), 0, "Size of TCP syncache hashtable"); SYSCTL_VNET_UINT(_net_inet_tcp_syncache, OID_AUTO, rexmtlimit, CTLFLAG_RW, &VNET_NAME(tcp_syncache.rexmt_limit), 0, "Limit on SYN/ACK retransmissions"); VNET_DEFINE(int, tcp_sc_rst_sock_fail) = 1; SYSCTL_VNET_INT(_net_inet_tcp_syncache, OID_AUTO, rst_on_sock_fail, CTLFLAG_RW, &VNET_NAME(tcp_sc_rst_sock_fail), 0, "Send reset on socket allocation failure"); static MALLOC_DEFINE(M_SYNCACHE, "syncache", "TCP syncache"); #define SYNCACHE_HASH(inc, mask) \ ((V_tcp_syncache.hash_secret ^ \ (inc)->inc_faddr.s_addr ^ \ ((inc)->inc_faddr.s_addr >> 16) ^ \ (inc)->inc_fport ^ (inc)->inc_lport) & mask) #define SYNCACHE_HASH6(inc, mask) \ ((V_tcp_syncache.hash_secret ^ \ (inc)->inc6_faddr.s6_addr32[0] ^ \ (inc)->inc6_faddr.s6_addr32[3] ^ \ (inc)->inc_fport ^ (inc)->inc_lport) & mask) #define ENDPTS_EQ(a, b) ( \ (a)->ie_fport == (b)->ie_fport && \ (a)->ie_lport == (b)->ie_lport && \ (a)->ie_faddr.s_addr == (b)->ie_faddr.s_addr && \ (a)->ie_laddr.s_addr == (b)->ie_laddr.s_addr \ ) #define ENDPTS6_EQ(a, b) (memcmp(a, b, sizeof(*a)) == 0) #define SCH_LOCK(sch) mtx_lock(&(sch)->sch_mtx) #define SCH_UNLOCK(sch) mtx_unlock(&(sch)->sch_mtx) #define SCH_LOCK_ASSERT(sch) mtx_assert(&(sch)->sch_mtx, MA_OWNED) /* * Requires the syncache entry to be already removed from the bucket list. */ static void syncache_free(struct syncache *sc) { if (sc->sc_ipopts) (void) m_free(sc->sc_ipopts); if (sc->sc_cred) crfree(sc->sc_cred); #ifdef MAC mac_syncache_destroy(&sc->sc_label); #endif uma_zfree(V_tcp_syncache.zone, sc); } void syncache_init(void) { int i; V_tcp_syncache.cache_count = 0; V_tcp_syncache.hashsize = TCP_SYNCACHE_HASHSIZE; V_tcp_syncache.bucket_limit = TCP_SYNCACHE_BUCKETLIMIT; V_tcp_syncache.rexmt_limit = SYNCACHE_MAXREXMTS; V_tcp_syncache.hash_secret = arc4random(); TUNABLE_INT_FETCH("net.inet.tcp.syncache.hashsize", &V_tcp_syncache.hashsize); TUNABLE_INT_FETCH("net.inet.tcp.syncache.bucketlimit", &V_tcp_syncache.bucket_limit); if (!powerof2(V_tcp_syncache.hashsize) || V_tcp_syncache.hashsize == 0) { printf("WARNING: syncache hash size is not a power of 2.\n"); V_tcp_syncache.hashsize = TCP_SYNCACHE_HASHSIZE; } V_tcp_syncache.hashmask = V_tcp_syncache.hashsize - 1; /* Set limits. */ V_tcp_syncache.cache_limit = V_tcp_syncache.hashsize * V_tcp_syncache.bucket_limit; TUNABLE_INT_FETCH("net.inet.tcp.syncache.cachelimit", &V_tcp_syncache.cache_limit); /* Allocate the hash table. */ V_tcp_syncache.hashbase = malloc(V_tcp_syncache.hashsize * sizeof(struct syncache_head), M_SYNCACHE, M_WAITOK | M_ZERO); /* Initialize the hash buckets. */ for (i = 0; i < V_tcp_syncache.hashsize; i++) { #ifdef VIMAGE V_tcp_syncache.hashbase[i].sch_vnet = curvnet; #endif TAILQ_INIT(&V_tcp_syncache.hashbase[i].sch_bucket); mtx_init(&V_tcp_syncache.hashbase[i].sch_mtx, "tcp_sc_head", NULL, MTX_DEF); callout_init_mtx(&V_tcp_syncache.hashbase[i].sch_timer, &V_tcp_syncache.hashbase[i].sch_mtx, 0); V_tcp_syncache.hashbase[i].sch_length = 0; } /* Create the syncache entry zone. */ V_tcp_syncache.zone = uma_zcreate("syncache", sizeof(struct syncache), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0); uma_zone_set_max(V_tcp_syncache.zone, V_tcp_syncache.cache_limit); } #ifdef VIMAGE void syncache_destroy(void) { struct syncache_head *sch; struct syncache *sc, *nsc; int i; /* Cleanup hash buckets: stop timers, free entries, destroy locks. */ for (i = 0; i < V_tcp_syncache.hashsize; i++) { sch = &V_tcp_syncache.hashbase[i]; callout_drain(&sch->sch_timer); SCH_LOCK(sch); TAILQ_FOREACH_SAFE(sc, &sch->sch_bucket, sc_hash, nsc) syncache_drop(sc, sch); SCH_UNLOCK(sch); KASSERT(TAILQ_EMPTY(&sch->sch_bucket), ("%s: sch->sch_bucket not empty", __func__)); KASSERT(sch->sch_length == 0, ("%s: sch->sch_length %d not 0", __func__, sch->sch_length)); mtx_destroy(&sch->sch_mtx); } KASSERT(V_tcp_syncache.cache_count == 0, ("%s: cache_count %d not 0", __func__, V_tcp_syncache.cache_count)); /* Free the allocated global resources. */ uma_zdestroy(V_tcp_syncache.zone); free(V_tcp_syncache.hashbase, M_SYNCACHE); } #endif /* * Inserts a syncache entry into the specified bucket row. * Locks and unlocks the syncache_head autonomously. */ static void syncache_insert(struct syncache *sc, struct syncache_head *sch) { struct syncache *sc2; SCH_LOCK(sch); /* * Make sure that we don't overflow the per-bucket limit. * If the bucket is full, toss the oldest element. */ if (sch->sch_length >= V_tcp_syncache.bucket_limit) { KASSERT(!TAILQ_EMPTY(&sch->sch_bucket), ("sch->sch_length incorrect")); sc2 = TAILQ_LAST(&sch->sch_bucket, sch_head); syncache_drop(sc2, sch); TCPSTAT_INC(tcps_sc_bucketoverflow); } /* Put it into the bucket. */ TAILQ_INSERT_HEAD(&sch->sch_bucket, sc, sc_hash); sch->sch_length++; /* Reinitialize the bucket row's timer. */ if (sch->sch_length == 1) sch->sch_nextc = ticks + INT_MAX; syncache_timeout(sc, sch, 1); SCH_UNLOCK(sch); V_tcp_syncache.cache_count++; TCPSTAT_INC(tcps_sc_added); } /* * Remove and free entry from syncache bucket row. * Expects locked syncache head. */ static void syncache_drop(struct syncache *sc, struct syncache_head *sch) { SCH_LOCK_ASSERT(sch); TAILQ_REMOVE(&sch->sch_bucket, sc, sc_hash); sch->sch_length--; #ifndef TCP_OFFLOAD_DISABLE if (sc->sc_tu) sc->sc_tu->tu_syncache_event(TOE_SC_DROP, sc->sc_toepcb); #endif syncache_free(sc); V_tcp_syncache.cache_count--; } /* * Engage/reengage time on bucket row. */ static void syncache_timeout(struct syncache *sc, struct syncache_head *sch, int docallout) { sc->sc_rxttime = ticks + TCPTV_RTOBASE * (tcp_backoff[sc->sc_rxmits]); sc->sc_rxmits++; if (TSTMP_LT(sc->sc_rxttime, sch->sch_nextc)) { sch->sch_nextc = sc->sc_rxttime; if (docallout) callout_reset(&sch->sch_timer, sch->sch_nextc - ticks, syncache_timer, (void *)sch); } } /* * Walk the timer queues, looking for SYN,ACKs that need to be retransmitted. * If we have retransmitted an entry the maximum number of times, expire it. * One separate timer for each bucket row. */ static void syncache_timer(void *xsch) { struct syncache_head *sch = (struct syncache_head *)xsch; struct syncache *sc, *nsc; int tick = ticks; char *s; CURVNET_SET(sch->sch_vnet); /* NB: syncache_head has already been locked by the callout. */ SCH_LOCK_ASSERT(sch); /* * In the following cycle we may remove some entries and/or * advance some timeouts, so re-initialize the bucket timer. */ sch->sch_nextc = tick + INT_MAX; TAILQ_FOREACH_SAFE(sc, &sch->sch_bucket, sc_hash, nsc) { /* * We do not check if the listen socket still exists * and accept the case where the listen socket may be * gone by the time we resend the SYN/ACK. We do * not expect this to happens often. If it does, * then the RST will be sent by the time the remote * host does the SYN/ACK->ACK. */ if (TSTMP_GT(sc->sc_rxttime, tick)) { if (TSTMP_LT(sc->sc_rxttime, sch->sch_nextc)) sch->sch_nextc = sc->sc_rxttime; continue; } if (sc->sc_rxmits > V_tcp_syncache.rexmt_limit) { if ((s = tcp_log_addrs(&sc->sc_inc, NULL, NULL, NULL))) { log(LOG_DEBUG, "%s; %s: Retransmits exhausted, " "giving up and removing syncache entry\n", s, __func__); free(s, M_TCPLOG); } syncache_drop(sc, sch); TCPSTAT_INC(tcps_sc_stale); continue; } if ((s = tcp_log_addrs(&sc->sc_inc, NULL, NULL, NULL))) { log(LOG_DEBUG, "%s; %s: Response timeout, " "retransmitting (%u) SYN|ACK\n", s, __func__, sc->sc_rxmits); free(s, M_TCPLOG); } (void) syncache_respond(sc); TCPSTAT_INC(tcps_sc_retransmitted); syncache_timeout(sc, sch, 0); } if (!TAILQ_EMPTY(&(sch)->sch_bucket)) callout_reset(&(sch)->sch_timer, (sch)->sch_nextc - tick, syncache_timer, (void *)(sch)); CURVNET_RESTORE(); } /* * Find an entry in the syncache. * Returns always with locked syncache_head plus a matching entry or NULL. */ struct syncache * syncache_lookup(struct in_conninfo *inc, struct syncache_head **schp) { struct syncache *sc; struct syncache_head *sch; #ifdef INET6 if (inc->inc_flags & INC_ISIPV6) { sch = &V_tcp_syncache.hashbase[ SYNCACHE_HASH6(inc, V_tcp_syncache.hashmask)]; *schp = sch; SCH_LOCK(sch); /* Circle through bucket row to find matching entry. */ TAILQ_FOREACH(sc, &sch->sch_bucket, sc_hash) { if (ENDPTS6_EQ(&inc->inc_ie, &sc->sc_inc.inc_ie)) return (sc); } } else #endif { sch = &V_tcp_syncache.hashbase[ SYNCACHE_HASH(inc, V_tcp_syncache.hashmask)]; *schp = sch; SCH_LOCK(sch); /* Circle through bucket row to find matching entry. */ TAILQ_FOREACH(sc, &sch->sch_bucket, sc_hash) { #ifdef INET6 if (sc->sc_inc.inc_flags & INC_ISIPV6) continue; #endif if (ENDPTS_EQ(&inc->inc_ie, &sc->sc_inc.inc_ie)) return (sc); } } SCH_LOCK_ASSERT(*schp); return (NULL); /* always returns with locked sch */ } /* * This function is called when we get a RST for a * non-existent connection, so that we can see if the * connection is in the syn cache. If it is, zap it. */ void syncache_chkrst(struct in_conninfo *inc, struct tcphdr *th) { struct syncache *sc; struct syncache_head *sch; char *s = NULL; sc = syncache_lookup(inc, &sch); /* returns locked sch */ SCH_LOCK_ASSERT(sch); /* * Any RST to our SYN|ACK must not carry ACK, SYN or FIN flags. * See RFC 793 page 65, section SEGMENT ARRIVES. */ if (th->th_flags & (TH_ACK|TH_SYN|TH_FIN)) { if ((s = tcp_log_addrs(inc, th, NULL, NULL))) log(LOG_DEBUG, "%s; %s: Spurious RST with ACK, SYN or " "FIN flag set, segment ignored\n", s, __func__); TCPSTAT_INC(tcps_badrst); goto done; } /* * No corresponding connection was found in syncache. * If syncookies are enabled and possibly exclusively * used, or we are under memory pressure, a valid RST * may not find a syncache entry. In that case we're * done and no SYN|ACK retransmissions will happen. * Otherwise the RST was misdirected or spoofed. */ if (sc == NULL) { if ((s = tcp_log_addrs(inc, th, NULL, NULL))) log(LOG_DEBUG, "%s; %s: Spurious RST without matching " "syncache entry (possibly syncookie only), " "segment ignored\n", s, __func__); TCPSTAT_INC(tcps_badrst); goto done; } /* * If the RST bit is set, check the sequence number to see * if this is a valid reset segment. * RFC 793 page 37: * In all states except SYN-SENT, all reset (RST) segments * are validated by checking their SEQ-fields. A reset is * valid if its sequence number is in the window. * * The sequence number in the reset segment is normally an * echo of our outgoing acknowlegement numbers, but some hosts * send a reset with the sequence number at the rightmost edge * of our receive window, and we have to handle this case. */ if (SEQ_GEQ(th->th_seq, sc->sc_irs) && SEQ_LEQ(th->th_seq, sc->sc_irs + sc->sc_wnd)) { syncache_drop(sc, sch); if ((s = tcp_log_addrs(inc, th, NULL, NULL))) log(LOG_DEBUG, "%s; %s: Our SYN|ACK was rejected, " "connection attempt aborted by remote endpoint\n", s, __func__); TCPSTAT_INC(tcps_sc_reset); } else { if ((s = tcp_log_addrs(inc, th, NULL, NULL))) log(LOG_DEBUG, "%s; %s: RST with invalid SEQ %u != " "IRS %u (+WND %u), segment ignored\n", s, __func__, th->th_seq, sc->sc_irs, sc->sc_wnd); TCPSTAT_INC(tcps_badrst); } done: if (s != NULL) free(s, M_TCPLOG); SCH_UNLOCK(sch); } void syncache_badack(struct in_conninfo *inc) { struct syncache *sc; struct syncache_head *sch; sc = syncache_lookup(inc, &sch); /* returns locked sch */ SCH_LOCK_ASSERT(sch); if (sc != NULL) { syncache_drop(sc, sch); TCPSTAT_INC(tcps_sc_badack); } SCH_UNLOCK(sch); } void syncache_unreach(struct in_conninfo *inc, struct tcphdr *th) { struct syncache *sc; struct syncache_head *sch; sc = syncache_lookup(inc, &sch); /* returns locked sch */ SCH_LOCK_ASSERT(sch); if (sc == NULL) goto done; /* If the sequence number != sc_iss, then it's a bogus ICMP msg */ if (ntohl(th->th_seq) != sc->sc_iss) goto done; /* * If we've rertransmitted 3 times and this is our second error, * we remove the entry. Otherwise, we allow it to continue on. * This prevents us from incorrectly nuking an entry during a * spurious network outage. * * See tcp_notify(). */ if ((sc->sc_flags & SCF_UNREACH) == 0 || sc->sc_rxmits < 3 + 1) { sc->sc_flags |= SCF_UNREACH; goto done; } syncache_drop(sc, sch); TCPSTAT_INC(tcps_sc_unreach); done: SCH_UNLOCK(sch); } /* * Build a new TCP socket structure from a syncache entry. */ static struct socket * syncache_socket(struct syncache *sc, struct socket *lso, struct mbuf *m) { struct inpcb *inp = NULL; struct socket *so; struct tcpcb *tp; int error; char *s; INP_INFO_WLOCK_ASSERT(&V_tcbinfo); /* * Ok, create the full blown connection, and set things up * as they would have been set up if we had created the * connection when the SYN arrived. If we can't create * the connection, abort it. */ so = sonewconn(lso, SS_ISCONNECTED); if (so == NULL) { /* * Drop the connection; we will either send a RST or * have the peer retransmit its SYN again after its * RTO and try again. */ TCPSTAT_INC(tcps_listendrop); if ((s = tcp_log_addrs(&sc->sc_inc, NULL, NULL, NULL))) { log(LOG_DEBUG, "%s; %s: Socket create failed " "due to limits or memory shortage\n", s, __func__); free(s, M_TCPLOG); } goto abort2; } #ifdef MAC mac_socketpeer_set_from_mbuf(m, so); #endif inp = sotoinpcb(so); inp->inp_inc.inc_fibnum = so->so_fibnum; INP_WLOCK(inp); + INP_HASH_WLOCK(&V_tcbinfo); /* Insert new socket into PCB hash list. */ inp->inp_inc.inc_flags = sc->sc_inc.inc_flags; #ifdef INET6 if (sc->sc_inc.inc_flags & INC_ISIPV6) { inp->in6p_laddr = sc->sc_inc.inc6_laddr; } else { inp->inp_vflag &= ~INP_IPV6; inp->inp_vflag |= INP_IPV4; #endif inp->inp_laddr = sc->sc_inc.inc_laddr; #ifdef INET6 } #endif inp->inp_lport = sc->sc_inc.inc_lport; if ((error = in_pcbinshash(inp)) != 0) { /* * Undo the assignments above if we failed to * put the PCB on the hash lists. */ #ifdef INET6 if (sc->sc_inc.inc_flags & INC_ISIPV6) inp->in6p_laddr = in6addr_any; else #endif inp->inp_laddr.s_addr = INADDR_ANY; inp->inp_lport = 0; if ((s = tcp_log_addrs(&sc->sc_inc, NULL, NULL, NULL))) { log(LOG_DEBUG, "%s; %s: in_pcbinshash failed " "with error %i\n", s, __func__, error); free(s, M_TCPLOG); } + INP_HASH_WUNLOCK(&V_tcbinfo); goto abort; } #ifdef IPSEC /* Copy old policy into new socket's. */ if (ipsec_copy_policy(sotoinpcb(lso)->inp_sp, inp->inp_sp)) printf("syncache_socket: could not copy policy\n"); #endif #ifdef INET6 if (sc->sc_inc.inc_flags & INC_ISIPV6) { struct inpcb *oinp = sotoinpcb(lso); struct in6_addr laddr6; struct sockaddr_in6 sin6; /* * Inherit socket options from the listening socket. * Note that in6p_inputopts are not (and should not be) * copied, since it stores previously received options and is * used to detect if each new option is different than the * previous one and hence should be passed to a user. * If we copied in6p_inputopts, a user would not be able to * receive options just after calling the accept system call. */ inp->inp_flags |= oinp->inp_flags & INP_CONTROLOPTS; if (oinp->in6p_outputopts) inp->in6p_outputopts = ip6_copypktopts(oinp->in6p_outputopts, M_NOWAIT); sin6.sin6_family = AF_INET6; sin6.sin6_len = sizeof(sin6); sin6.sin6_addr = sc->sc_inc.inc6_faddr; sin6.sin6_port = sc->sc_inc.inc_fport; sin6.sin6_flowinfo = sin6.sin6_scope_id = 0; laddr6 = inp->in6p_laddr; if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) inp->in6p_laddr = sc->sc_inc.inc6_laddr; if ((error = in6_pcbconnect(inp, (struct sockaddr *)&sin6, thread0.td_ucred)) != 0) { inp->in6p_laddr = laddr6; if ((s = tcp_log_addrs(&sc->sc_inc, NULL, NULL, NULL))) { log(LOG_DEBUG, "%s; %s: in6_pcbconnect failed " "with error %i\n", s, __func__, error); free(s, M_TCPLOG); } + INP_HASH_WUNLOCK(&V_tcbinfo); goto abort; } /* Override flowlabel from in6_pcbconnect. */ inp->inp_flow &= ~IPV6_FLOWLABEL_MASK; inp->inp_flow |= sc->sc_flowlabel; } #endif /* INET6 */ #if defined(INET) && defined(INET6) else #endif #ifdef INET { struct in_addr laddr; struct sockaddr_in sin; inp->inp_options = (m) ? ip_srcroute(m) : NULL; if (inp->inp_options == NULL) { inp->inp_options = sc->sc_ipopts; sc->sc_ipopts = NULL; } sin.sin_family = AF_INET; sin.sin_len = sizeof(sin); sin.sin_addr = sc->sc_inc.inc_faddr; sin.sin_port = sc->sc_inc.inc_fport; bzero((caddr_t)sin.sin_zero, sizeof(sin.sin_zero)); laddr = inp->inp_laddr; if (inp->inp_laddr.s_addr == INADDR_ANY) inp->inp_laddr = sc->sc_inc.inc_laddr; if ((error = in_pcbconnect(inp, (struct sockaddr *)&sin, thread0.td_ucred)) != 0) { inp->inp_laddr = laddr; if ((s = tcp_log_addrs(&sc->sc_inc, NULL, NULL, NULL))) { log(LOG_DEBUG, "%s; %s: in_pcbconnect failed " "with error %i\n", s, __func__, error); free(s, M_TCPLOG); } + INP_HASH_WUNLOCK(&V_tcbinfo); goto abort; } } #endif /* INET */ + INP_HASH_WUNLOCK(&V_tcbinfo); tp = intotcpcb(inp); tp->t_state = TCPS_SYN_RECEIVED; tp->iss = sc->sc_iss; tp->irs = sc->sc_irs; tcp_rcvseqinit(tp); tcp_sendseqinit(tp); tp->snd_wl1 = sc->sc_irs; tp->snd_max = tp->iss + 1; tp->snd_nxt = tp->iss + 1; tp->rcv_up = sc->sc_irs + 1; tp->rcv_wnd = sc->sc_wnd; tp->rcv_adv += tp->rcv_wnd; tp->last_ack_sent = tp->rcv_nxt; tp->t_flags = sototcpcb(lso)->t_flags & (TF_NOPUSH|TF_NODELAY); if (sc->sc_flags & SCF_NOOPT) tp->t_flags |= TF_NOOPT; else { if (sc->sc_flags & SCF_WINSCALE) { tp->t_flags |= TF_REQ_SCALE|TF_RCVD_SCALE; tp->snd_scale = sc->sc_requested_s_scale; tp->request_r_scale = sc->sc_requested_r_scale; } if (sc->sc_flags & SCF_TIMESTAMP) { tp->t_flags |= TF_REQ_TSTMP|TF_RCVD_TSTMP; tp->ts_recent = sc->sc_tsreflect; tp->ts_recent_age = ticks; tp->ts_offset = sc->sc_tsoff; } #ifdef TCP_SIGNATURE if (sc->sc_flags & SCF_SIGNATURE) tp->t_flags |= TF_SIGNATURE; #endif if (sc->sc_flags & SCF_SACK) tp->t_flags |= TF_SACK_PERMIT; } if (sc->sc_flags & SCF_ECN) tp->t_flags |= TF_ECN_PERMIT; /* * Set up MSS and get cached values from tcp_hostcache. * This might overwrite some of the defaults we just set. */ tcp_mss(tp, sc->sc_peer_mss); /* * If the SYN,ACK was retransmitted, reset cwnd to 1 segment. * NB: sc_rxmits counts all SYN,ACK transmits, not just retransmits. */ if (sc->sc_rxmits > 1) tp->snd_cwnd = tp->t_maxseg; tcp_timer_activate(tp, TT_KEEP, tcp_keepinit); INP_WUNLOCK(inp); TCPSTAT_INC(tcps_accepts); return (so); abort: INP_WUNLOCK(inp); abort2: if (so != NULL) soabort(so); return (NULL); } /* * This function gets called when we receive an ACK for a * socket in the LISTEN state. We look up the connection * in the syncache, and if its there, we pull it out of * the cache and turn it into a full-blown connection in * the SYN-RECEIVED state. */ int syncache_expand(struct in_conninfo *inc, struct tcpopt *to, struct tcphdr *th, struct socket **lsop, struct mbuf *m) { struct syncache *sc; struct syncache_head *sch; struct syncache scs; char *s; /* * Global TCP locks are held because we manipulate the PCB lists * and create a new socket. */ INP_INFO_WLOCK_ASSERT(&V_tcbinfo); KASSERT((th->th_flags & (TH_RST|TH_ACK|TH_SYN)) == TH_ACK, ("%s: can handle only ACK", __func__)); sc = syncache_lookup(inc, &sch); /* returns locked sch */ SCH_LOCK_ASSERT(sch); if (sc == NULL) { /* * There is no syncache entry, so see if this ACK is * a returning syncookie. To do this, first: * A. See if this socket has had a syncache entry dropped in * the past. We don't want to accept a bogus syncookie * if we've never received a SYN. * B. check that the syncookie is valid. If it is, then * cobble up a fake syncache entry, and return. */ if (!V_tcp_syncookies) { SCH_UNLOCK(sch); if ((s = tcp_log_addrs(inc, th, NULL, NULL))) log(LOG_DEBUG, "%s; %s: Spurious ACK, " "segment rejected (syncookies disabled)\n", s, __func__); goto failed; } bzero(&scs, sizeof(scs)); sc = syncookie_lookup(inc, sch, &scs, to, th, *lsop); SCH_UNLOCK(sch); if (sc == NULL) { if ((s = tcp_log_addrs(inc, th, NULL, NULL))) log(LOG_DEBUG, "%s; %s: Segment failed " "SYNCOOKIE authentication, segment rejected " "(probably spoofed)\n", s, __func__); goto failed; } } else { /* Pull out the entry to unlock the bucket row. */ TAILQ_REMOVE(&sch->sch_bucket, sc, sc_hash); sch->sch_length--; V_tcp_syncache.cache_count--; SCH_UNLOCK(sch); } /* * Segment validation: * ACK must match our initial sequence number + 1 (the SYN|ACK). */ if (th->th_ack != sc->sc_iss + 1 && !TOEPCB_ISSET(sc)) { if ((s = tcp_log_addrs(inc, th, NULL, NULL))) log(LOG_DEBUG, "%s; %s: ACK %u != ISS+1 %u, segment " "rejected\n", s, __func__, th->th_ack, sc->sc_iss); goto failed; } /* * The SEQ must fall in the window starting at the received * initial receive sequence number + 1 (the SYN). */ if ((SEQ_LEQ(th->th_seq, sc->sc_irs) || SEQ_GT(th->th_seq, sc->sc_irs + sc->sc_wnd)) && !TOEPCB_ISSET(sc)) { if ((s = tcp_log_addrs(inc, th, NULL, NULL))) log(LOG_DEBUG, "%s; %s: SEQ %u != IRS+1 %u, segment " "rejected\n", s, __func__, th->th_seq, sc->sc_irs); goto failed; } if (!(sc->sc_flags & SCF_TIMESTAMP) && (to->to_flags & TOF_TS)) { if ((s = tcp_log_addrs(inc, th, NULL, NULL))) log(LOG_DEBUG, "%s; %s: Timestamp not expected, " "segment rejected\n", s, __func__); goto failed; } /* * If timestamps were negotiated the reflected timestamp * must be equal to what we actually sent in the SYN|ACK. */ if ((to->to_flags & TOF_TS) && to->to_tsecr != sc->sc_ts && !TOEPCB_ISSET(sc)) { if ((s = tcp_log_addrs(inc, th, NULL, NULL))) log(LOG_DEBUG, "%s; %s: TSECR %u != TS %u, " "segment rejected\n", s, __func__, to->to_tsecr, sc->sc_ts); goto failed; } *lsop = syncache_socket(sc, *lsop, m); if (*lsop == NULL) TCPSTAT_INC(tcps_sc_aborted); else TCPSTAT_INC(tcps_sc_completed); /* how do we find the inp for the new socket? */ if (sc != &scs) syncache_free(sc); return (1); failed: if (sc != NULL && sc != &scs) syncache_free(sc); if (s != NULL) free(s, M_TCPLOG); *lsop = NULL; return (0); } int tcp_offload_syncache_expand(struct in_conninfo *inc, struct toeopt *toeo, struct tcphdr *th, struct socket **lsop, struct mbuf *m) { struct tcpopt to; int rc; bzero(&to, sizeof(struct tcpopt)); to.to_mss = toeo->to_mss; to.to_wscale = toeo->to_wscale; to.to_flags = toeo->to_flags; INP_INFO_WLOCK(&V_tcbinfo); rc = syncache_expand(inc, &to, th, lsop, m); INP_INFO_WUNLOCK(&V_tcbinfo); return (rc); } /* * Given a LISTEN socket and an inbound SYN request, add * this to the syn cache, and send back a segment: * * to the source. * * IMPORTANT NOTE: We do _NOT_ ACK data that might accompany the SYN. * Doing so would require that we hold onto the data and deliver it * to the application. However, if we are the target of a SYN-flood * DoS attack, an attacker could send data which would eventually * consume all available buffer space if it were ACKed. By not ACKing * the data, we avoid this DoS scenario. */ static void _syncache_add(struct in_conninfo *inc, struct tcpopt *to, struct tcphdr *th, struct inpcb *inp, struct socket **lsop, struct mbuf *m, struct toe_usrreqs *tu, void *toepcb) { struct tcpcb *tp; struct socket *so; struct syncache *sc = NULL; struct syncache_head *sch; struct mbuf *ipopts = NULL; u_int32_t flowtmp; u_int ltflags; int win, sb_hiwat, ip_ttl, ip_tos; char *s; #ifdef INET6 int autoflowlabel = 0; #endif #ifdef MAC struct label *maclabel; #endif struct syncache scs; struct ucred *cred; INP_INFO_WLOCK_ASSERT(&V_tcbinfo); INP_WLOCK_ASSERT(inp); /* listen socket */ KASSERT((th->th_flags & (TH_RST|TH_ACK|TH_SYN)) == TH_SYN, ("%s: unexpected tcp flags", __func__)); /* * Combine all so/tp operations very early to drop the INP lock as * soon as possible. */ so = *lsop; tp = sototcpcb(so); cred = crhold(so->so_cred); #ifdef INET6 if ((inc->inc_flags & INC_ISIPV6) && (inp->inp_flags & IN6P_AUTOFLOWLABEL)) autoflowlabel = 1; #endif ip_ttl = inp->inp_ip_ttl; ip_tos = inp->inp_ip_tos; win = sbspace(&so->so_rcv); sb_hiwat = so->so_rcv.sb_hiwat; ltflags = (tp->t_flags & (TF_NOOPT | TF_SIGNATURE)); /* By the time we drop the lock these should no longer be used. */ so = NULL; tp = NULL; #ifdef MAC if (mac_syncache_init(&maclabel) != 0) { INP_WUNLOCK(inp); INP_INFO_WUNLOCK(&V_tcbinfo); goto done; } else mac_syncache_create(maclabel, inp); #endif INP_WUNLOCK(inp); INP_INFO_WUNLOCK(&V_tcbinfo); /* * Remember the IP options, if any. */ #ifdef INET6 if (!(inc->inc_flags & INC_ISIPV6)) #endif #ifdef INET ipopts = (m) ? ip_srcroute(m) : NULL; #else ipopts = NULL; #endif /* * See if we already have an entry for this connection. * If we do, resend the SYN,ACK, and reset the retransmit timer. * * XXX: should the syncache be re-initialized with the contents * of the new SYN here (which may have different options?) * * XXX: We do not check the sequence number to see if this is a * real retransmit or a new connection attempt. The question is * how to handle such a case; either ignore it as spoofed, or * drop the current entry and create a new one? */ sc = syncache_lookup(inc, &sch); /* returns locked entry */ SCH_LOCK_ASSERT(sch); if (sc != NULL) { #ifndef TCP_OFFLOAD_DISABLE if (sc->sc_tu) sc->sc_tu->tu_syncache_event(TOE_SC_ENTRY_PRESENT, sc->sc_toepcb); #endif TCPSTAT_INC(tcps_sc_dupsyn); if (ipopts) { /* * If we were remembering a previous source route, * forget it and use the new one we've been given. */ if (sc->sc_ipopts) (void) m_free(sc->sc_ipopts); sc->sc_ipopts = ipopts; } /* * Update timestamp if present. */ if ((sc->sc_flags & SCF_TIMESTAMP) && (to->to_flags & TOF_TS)) sc->sc_tsreflect = to->to_tsval; else sc->sc_flags &= ~SCF_TIMESTAMP; #ifdef MAC /* * Since we have already unconditionally allocated label * storage, free it up. The syncache entry will already * have an initialized label we can use. */ mac_syncache_destroy(&maclabel); #endif /* Retransmit SYN|ACK and reset retransmit count. */ if ((s = tcp_log_addrs(&sc->sc_inc, th, NULL, NULL))) { log(LOG_DEBUG, "%s; %s: Received duplicate SYN, " "resetting timer and retransmitting SYN|ACK\n", s, __func__); free(s, M_TCPLOG); } if (!TOEPCB_ISSET(sc) && syncache_respond(sc) == 0) { sc->sc_rxmits = 0; syncache_timeout(sc, sch, 1); TCPSTAT_INC(tcps_sndacks); TCPSTAT_INC(tcps_sndtotal); } SCH_UNLOCK(sch); goto done; } sc = uma_zalloc(V_tcp_syncache.zone, M_NOWAIT | M_ZERO); if (sc == NULL) { /* * The zone allocator couldn't provide more entries. * Treat this as if the cache was full; drop the oldest * entry and insert the new one. */ TCPSTAT_INC(tcps_sc_zonefail); if ((sc = TAILQ_LAST(&sch->sch_bucket, sch_head)) != NULL) syncache_drop(sc, sch); sc = uma_zalloc(V_tcp_syncache.zone, M_NOWAIT | M_ZERO); if (sc == NULL) { if (V_tcp_syncookies) { bzero(&scs, sizeof(scs)); sc = &scs; } else { SCH_UNLOCK(sch); if (ipopts) (void) m_free(ipopts); goto done; } } } /* * Fill in the syncache values. */ #ifdef MAC sc->sc_label = maclabel; #endif sc->sc_cred = cred; cred = NULL; sc->sc_ipopts = ipopts; bcopy(inc, &sc->sc_inc, sizeof(struct in_conninfo)); #ifdef INET6 if (!(inc->inc_flags & INC_ISIPV6)) #endif { sc->sc_ip_tos = ip_tos; sc->sc_ip_ttl = ip_ttl; } #ifndef TCP_OFFLOAD_DISABLE sc->sc_tu = tu; sc->sc_toepcb = toepcb; #endif sc->sc_irs = th->th_seq; sc->sc_iss = arc4random(); sc->sc_flags = 0; sc->sc_flowlabel = 0; /* * Initial receive window: clip sbspace to [0 .. TCP_MAXWIN]. * win was derived from socket earlier in the function. */ win = imax(win, 0); win = imin(win, TCP_MAXWIN); sc->sc_wnd = win; if (V_tcp_do_rfc1323) { /* * A timestamp received in a SYN makes * it ok to send timestamp requests and replies. */ if (to->to_flags & TOF_TS) { sc->sc_tsreflect = to->to_tsval; sc->sc_ts = ticks; sc->sc_flags |= SCF_TIMESTAMP; } if (to->to_flags & TOF_SCALE) { int wscale = 0; /* * Pick the smallest possible scaling factor that * will still allow us to scale up to sb_max, aka * kern.ipc.maxsockbuf. * * We do this because there are broken firewalls that * will corrupt the window scale option, leading to * the other endpoint believing that our advertised * window is unscaled. At scale factors larger than * 5 the unscaled window will drop below 1500 bytes, * leading to serious problems when traversing these * broken firewalls. * * With the default maxsockbuf of 256K, a scale factor * of 3 will be chosen by this algorithm. Those who * choose a larger maxsockbuf should watch out * for the compatiblity problems mentioned above. * * RFC1323: The Window field in a SYN (i.e., a * or ) segment itself is never scaled. */ while (wscale < TCP_MAX_WINSHIFT && (TCP_MAXWIN << wscale) < sb_max) wscale++; sc->sc_requested_r_scale = wscale; sc->sc_requested_s_scale = to->to_wscale; sc->sc_flags |= SCF_WINSCALE; } } #ifdef TCP_SIGNATURE /* * If listening socket requested TCP digests, and received SYN * contains the option, flag this in the syncache so that * syncache_respond() will do the right thing with the SYN+ACK. * XXX: Currently we always record the option by default and will * attempt to use it in syncache_respond(). */ if (to->to_flags & TOF_SIGNATURE || ltflags & TF_SIGNATURE) sc->sc_flags |= SCF_SIGNATURE; #endif if (to->to_flags & TOF_SACKPERM) sc->sc_flags |= SCF_SACK; if (to->to_flags & TOF_MSS) sc->sc_peer_mss = to->to_mss; /* peer mss may be zero */ if (ltflags & TF_NOOPT) sc->sc_flags |= SCF_NOOPT; if ((th->th_flags & (TH_ECE|TH_CWR)) && V_tcp_do_ecn) sc->sc_flags |= SCF_ECN; if (V_tcp_syncookies) { syncookie_generate(sch, sc, &flowtmp); #ifdef INET6 if (autoflowlabel) sc->sc_flowlabel = flowtmp; #endif } else { #ifdef INET6 if (autoflowlabel) sc->sc_flowlabel = (htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK); #endif } SCH_UNLOCK(sch); /* * Do a standard 3-way handshake. */ if (TOEPCB_ISSET(sc) || syncache_respond(sc) == 0) { if (V_tcp_syncookies && V_tcp_syncookiesonly && sc != &scs) syncache_free(sc); else if (sc != &scs) syncache_insert(sc, sch); /* locks and unlocks sch */ TCPSTAT_INC(tcps_sndacks); TCPSTAT_INC(tcps_sndtotal); } else { if (sc != &scs) syncache_free(sc); TCPSTAT_INC(tcps_sc_dropped); } done: if (cred != NULL) crfree(cred); #ifdef MAC if (sc == &scs) mac_syncache_destroy(&maclabel); #endif if (m) { *lsop = NULL; m_freem(m); } } static int syncache_respond(struct syncache *sc) { struct ip *ip = NULL; struct mbuf *m; struct tcphdr *th = NULL; int optlen, error = 0; /* Make compiler happy */ u_int16_t hlen, tlen, mssopt; struct tcpopt to; #ifdef INET6 struct ip6_hdr *ip6 = NULL; #endif hlen = #ifdef INET6 (sc->sc_inc.inc_flags & INC_ISIPV6) ? sizeof(struct ip6_hdr) : #endif sizeof(struct ip); tlen = hlen + sizeof(struct tcphdr); /* Determine MSS we advertize to other end of connection. */ mssopt = tcp_mssopt(&sc->sc_inc); if (sc->sc_peer_mss) mssopt = max( min(sc->sc_peer_mss, mssopt), V_tcp_minmss); /* XXX: Assume that the entire packet will fit in a header mbuf. */ KASSERT(max_linkhdr + tlen + TCP_MAXOLEN <= MHLEN, ("syncache: mbuf too small")); /* Create the IP+TCP header from scratch. */ m = m_gethdr(M_DONTWAIT, MT_DATA); if (m == NULL) return (ENOBUFS); #ifdef MAC mac_syncache_create_mbuf(sc->sc_label, m); #endif m->m_data += max_linkhdr; m->m_len = tlen; m->m_pkthdr.len = tlen; m->m_pkthdr.rcvif = NULL; #ifdef INET6 if (sc->sc_inc.inc_flags & INC_ISIPV6) { ip6 = mtod(m, struct ip6_hdr *); ip6->ip6_vfc = IPV6_VERSION; ip6->ip6_nxt = IPPROTO_TCP; ip6->ip6_src = sc->sc_inc.inc6_laddr; ip6->ip6_dst = sc->sc_inc.inc6_faddr; ip6->ip6_plen = htons(tlen - hlen); /* ip6_hlim is set after checksum */ ip6->ip6_flow &= ~IPV6_FLOWLABEL_MASK; ip6->ip6_flow |= sc->sc_flowlabel; th = (struct tcphdr *)(ip6 + 1); } #endif #if defined(INET6) && defined(INET) else #endif #ifdef INET { ip = mtod(m, struct ip *); ip->ip_v = IPVERSION; ip->ip_hl = sizeof(struct ip) >> 2; ip->ip_len = tlen; ip->ip_id = 0; ip->ip_off = 0; ip->ip_sum = 0; ip->ip_p = IPPROTO_TCP; ip->ip_src = sc->sc_inc.inc_laddr; ip->ip_dst = sc->sc_inc.inc_faddr; ip->ip_ttl = sc->sc_ip_ttl; ip->ip_tos = sc->sc_ip_tos; /* * See if we should do MTU discovery. Route lookups are * expensive, so we will only unset the DF bit if: * * 1) path_mtu_discovery is disabled * 2) the SCF_UNREACH flag has been set */ if (V_path_mtu_discovery && ((sc->sc_flags & SCF_UNREACH) == 0)) ip->ip_off |= IP_DF; th = (struct tcphdr *)(ip + 1); } #endif /* INET */ th->th_sport = sc->sc_inc.inc_lport; th->th_dport = sc->sc_inc.inc_fport; th->th_seq = htonl(sc->sc_iss); th->th_ack = htonl(sc->sc_irs + 1); th->th_off = sizeof(struct tcphdr) >> 2; th->th_x2 = 0; th->th_flags = TH_SYN|TH_ACK; th->th_win = htons(sc->sc_wnd); th->th_urp = 0; if (sc->sc_flags & SCF_ECN) { th->th_flags |= TH_ECE; TCPSTAT_INC(tcps_ecn_shs); } /* Tack on the TCP options. */ if ((sc->sc_flags & SCF_NOOPT) == 0) { to.to_flags = 0; to.to_mss = mssopt; to.to_flags = TOF_MSS; if (sc->sc_flags & SCF_WINSCALE) { to.to_wscale = sc->sc_requested_r_scale; to.to_flags |= TOF_SCALE; } if (sc->sc_flags & SCF_TIMESTAMP) { /* Virgin timestamp or TCP cookie enhanced one. */ to.to_tsval = sc->sc_ts; to.to_tsecr = sc->sc_tsreflect; to.to_flags |= TOF_TS; } if (sc->sc_flags & SCF_SACK) to.to_flags |= TOF_SACKPERM; #ifdef TCP_SIGNATURE if (sc->sc_flags & SCF_SIGNATURE) to.to_flags |= TOF_SIGNATURE; #endif optlen = tcp_addoptions(&to, (u_char *)(th + 1)); /* Adjust headers by option size. */ th->th_off = (sizeof(struct tcphdr) + optlen) >> 2; m->m_len += optlen; m->m_pkthdr.len += optlen; #ifdef TCP_SIGNATURE if (sc->sc_flags & SCF_SIGNATURE) tcp_signature_compute(m, 0, 0, optlen, to.to_signature, IPSEC_DIR_OUTBOUND); #endif #ifdef INET6 if (sc->sc_inc.inc_flags & INC_ISIPV6) ip6->ip6_plen = htons(ntohs(ip6->ip6_plen) + optlen); else #endif ip->ip_len += optlen; } else optlen = 0; M_SETFIB(m, sc->sc_inc.inc_fibnum); #ifdef INET6 if (sc->sc_inc.inc_flags & INC_ISIPV6) { th->th_sum = 0; th->th_sum = in6_cksum(m, IPPROTO_TCP, hlen, tlen + optlen - hlen); ip6->ip6_hlim = in6_selecthlim(NULL, NULL); error = ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL); } #endif #if defined(INET6) && defined(INET) else #endif #ifdef INET { th->th_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr, htons(tlen + optlen - hlen + IPPROTO_TCP)); m->m_pkthdr.csum_flags = CSUM_TCP; m->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum); error = ip_output(m, sc->sc_ipopts, NULL, 0, NULL, NULL); } #endif return (error); } void syncache_add(struct in_conninfo *inc, struct tcpopt *to, struct tcphdr *th, struct inpcb *inp, struct socket **lsop, struct mbuf *m) { _syncache_add(inc, to, th, inp, lsop, m, NULL, NULL); } void tcp_offload_syncache_add(struct in_conninfo *inc, struct toeopt *toeo, struct tcphdr *th, struct inpcb *inp, struct socket **lsop, struct toe_usrreqs *tu, void *toepcb) { struct tcpopt to; bzero(&to, sizeof(struct tcpopt)); to.to_mss = toeo->to_mss; to.to_wscale = toeo->to_wscale; to.to_flags = toeo->to_flags; INP_INFO_WLOCK(&V_tcbinfo); INP_WLOCK(inp); _syncache_add(inc, &to, th, inp, lsop, NULL, tu, toepcb); } /* * The purpose of SYN cookies is to avoid keeping track of all SYN's we * receive and to be able to handle SYN floods from bogus source addresses * (where we will never receive any reply). SYN floods try to exhaust all * our memory and available slots in the SYN cache table to cause a denial * of service to legitimate users of the local host. * * The idea of SYN cookies is to encode and include all necessary information * about the connection setup state within the SYN-ACK we send back and thus * to get along without keeping any local state until the ACK to the SYN-ACK * arrives (if ever). Everything we need to know should be available from * the information we encoded in the SYN-ACK. * * More information about the theory behind SYN cookies and its first * discussion and specification can be found at: * http://cr.yp.to/syncookies.html (overview) * http://cr.yp.to/syncookies/archive (gory details) * * This implementation extends the orginal idea and first implementation * of FreeBSD by using not only the initial sequence number field to store * information but also the timestamp field if present. This way we can * keep track of the entire state we need to know to recreate the session in * its original form. Almost all TCP speakers implement RFC1323 timestamps * these days. For those that do not we still have to live with the known * shortcomings of the ISN only SYN cookies. * * Cookie layers: * * Initial sequence number we send: * 31|................................|0 * DDDDDDDDDDDDDDDDDDDDDDDDDMMMRRRP * D = MD5 Digest (first dword) * M = MSS index * R = Rotation of secret * P = Odd or Even secret * * The MD5 Digest is computed with over following parameters: * a) randomly rotated secret * b) struct in_conninfo containing the remote/local ip/port (IPv4&IPv6) * c) the received initial sequence number from remote host * d) the rotation offset and odd/even bit * * Timestamp we send: * 31|................................|0 * DDDDDDDDDDDDDDDDDDDDDDSSSSRRRRA5 * D = MD5 Digest (third dword) (only as filler) * S = Requested send window scale * R = Requested receive window scale * A = SACK allowed * 5 = TCP-MD5 enabled (not implemented yet) * XORed with MD5 Digest (forth dword) * * The timestamp isn't cryptographically secure and doesn't need to be. * The double use of the MD5 digest dwords ties it to a specific remote/ * local host/port, remote initial sequence number and our local time * limited secret. A received timestamp is reverted (XORed) and then * the contained MD5 dword is compared to the computed one to ensure the * timestamp belongs to the SYN-ACK we sent. The other parameters may * have been tampered with but this isn't different from supplying bogus * values in the SYN in the first place. * * Some problems with SYN cookies remain however: * Consider the problem of a recreated (and retransmitted) cookie. If the * original SYN was accepted, the connection is established. The second * SYN is inflight, and if it arrives with an ISN that falls within the * receive window, the connection is killed. * * Notes: * A heuristic to determine when to accept syn cookies is not necessary. * An ACK flood would cause the syncookie verification to be attempted, * but a SYN flood causes syncookies to be generated. Both are of equal * cost, so there's no point in trying to optimize the ACK flood case. * Also, if you don't process certain ACKs for some reason, then all someone * would have to do is launch a SYN and ACK flood at the same time, which * would stop cookie verification and defeat the entire purpose of syncookies. */ static int tcp_sc_msstab[] = { 0, 256, 468, 536, 996, 1452, 1460, 8960 }; static void syncookie_generate(struct syncache_head *sch, struct syncache *sc, u_int32_t *flowlabel) { MD5_CTX ctx; u_int32_t md5_buffer[MD5_DIGEST_LENGTH / sizeof(u_int32_t)]; u_int32_t data; u_int32_t *secbits; u_int off, pmss, mss; int i; SCH_LOCK_ASSERT(sch); /* Which of the two secrets to use. */ secbits = sch->sch_oddeven ? sch->sch_secbits_odd : sch->sch_secbits_even; /* Reseed secret if too old. */ if (sch->sch_reseed < time_uptime) { sch->sch_oddeven = sch->sch_oddeven ? 0 : 1; /* toggle */ secbits = sch->sch_oddeven ? sch->sch_secbits_odd : sch->sch_secbits_even; for (i = 0; i < SYNCOOKIE_SECRET_SIZE; i++) secbits[i] = arc4random(); sch->sch_reseed = time_uptime + SYNCOOKIE_LIFETIME; } /* Secret rotation offset. */ off = sc->sc_iss & 0x7; /* iss was randomized before */ /* Maximum segment size calculation. */ pmss = max( min(sc->sc_peer_mss, tcp_mssopt(&sc->sc_inc)), V_tcp_minmss); for (mss = sizeof(tcp_sc_msstab) / sizeof(int) - 1; mss > 0; mss--) if (tcp_sc_msstab[mss] <= pmss) break; /* Fold parameters and MD5 digest into the ISN we will send. */ data = sch->sch_oddeven;/* odd or even secret, 1 bit */ data |= off << 1; /* secret offset, derived from iss, 3 bits */ data |= mss << 4; /* mss, 3 bits */ MD5Init(&ctx); MD5Update(&ctx, ((u_int8_t *)secbits) + off, SYNCOOKIE_SECRET_SIZE * sizeof(*secbits) - off); MD5Update(&ctx, secbits, off); MD5Update(&ctx, &sc->sc_inc, sizeof(sc->sc_inc)); MD5Update(&ctx, &sc->sc_irs, sizeof(sc->sc_irs)); MD5Update(&ctx, &data, sizeof(data)); MD5Final((u_int8_t *)&md5_buffer, &ctx); data |= (md5_buffer[0] << 7); sc->sc_iss = data; #ifdef INET6 *flowlabel = md5_buffer[1] & IPV6_FLOWLABEL_MASK; #endif /* Additional parameters are stored in the timestamp if present. */ if (sc->sc_flags & SCF_TIMESTAMP) { data = ((sc->sc_flags & SCF_SIGNATURE) ? 1 : 0); /* TCP-MD5, 1 bit */ data |= ((sc->sc_flags & SCF_SACK) ? 1 : 0) << 1; /* SACK, 1 bit */ data |= sc->sc_requested_s_scale << 2; /* SWIN scale, 4 bits */ data |= sc->sc_requested_r_scale << 6; /* RWIN scale, 4 bits */ data |= md5_buffer[2] << 10; /* more digest bits */ data ^= md5_buffer[3]; sc->sc_ts = data; sc->sc_tsoff = data - ticks; /* after XOR */ } TCPSTAT_INC(tcps_sc_sendcookie); } static struct syncache * syncookie_lookup(struct in_conninfo *inc, struct syncache_head *sch, struct syncache *sc, struct tcpopt *to, struct tcphdr *th, struct socket *so) { MD5_CTX ctx; u_int32_t md5_buffer[MD5_DIGEST_LENGTH / sizeof(u_int32_t)]; u_int32_t data = 0; u_int32_t *secbits; tcp_seq ack, seq; int off, mss, wnd, flags; SCH_LOCK_ASSERT(sch); /* * Pull information out of SYN-ACK/ACK and * revert sequence number advances. */ ack = th->th_ack - 1; seq = th->th_seq - 1; off = (ack >> 1) & 0x7; mss = (ack >> 4) & 0x7; flags = ack & 0x7f; /* Which of the two secrets to use. */ secbits = (flags & 0x1) ? sch->sch_secbits_odd : sch->sch_secbits_even; /* * The secret wasn't updated for the lifetime of a syncookie, * so this SYN-ACK/ACK is either too old (replay) or totally bogus. */ if (sch->sch_reseed + SYNCOOKIE_LIFETIME < time_uptime) { return (NULL); } /* Recompute the digest so we can compare it. */ MD5Init(&ctx); MD5Update(&ctx, ((u_int8_t *)secbits) + off, SYNCOOKIE_SECRET_SIZE * sizeof(*secbits) - off); MD5Update(&ctx, secbits, off); MD5Update(&ctx, inc, sizeof(*inc)); MD5Update(&ctx, &seq, sizeof(seq)); MD5Update(&ctx, &flags, sizeof(flags)); MD5Final((u_int8_t *)&md5_buffer, &ctx); /* Does the digest part of or ACK'ed ISS match? */ if ((ack & (~0x7f)) != (md5_buffer[0] << 7)) return (NULL); /* Does the digest part of our reflected timestamp match? */ if (to->to_flags & TOF_TS) { data = md5_buffer[3] ^ to->to_tsecr; if ((data & (~0x3ff)) != (md5_buffer[2] << 10)) return (NULL); } /* Fill in the syncache values. */ bcopy(inc, &sc->sc_inc, sizeof(struct in_conninfo)); sc->sc_ipopts = NULL; sc->sc_irs = seq; sc->sc_iss = ack; #ifdef INET6 if (inc->inc_flags & INC_ISIPV6) { if (sotoinpcb(so)->inp_flags & IN6P_AUTOFLOWLABEL) sc->sc_flowlabel = md5_buffer[1] & IPV6_FLOWLABEL_MASK; } else #endif { sc->sc_ip_ttl = sotoinpcb(so)->inp_ip_ttl; sc->sc_ip_tos = sotoinpcb(so)->inp_ip_tos; } /* Additional parameters that were encoded in the timestamp. */ if (data) { sc->sc_flags |= SCF_TIMESTAMP; sc->sc_tsreflect = to->to_tsval; sc->sc_ts = to->to_tsecr; sc->sc_tsoff = to->to_tsecr - ticks; sc->sc_flags |= (data & 0x1) ? SCF_SIGNATURE : 0; sc->sc_flags |= ((data >> 1) & 0x1) ? SCF_SACK : 0; sc->sc_requested_s_scale = min((data >> 2) & 0xf, TCP_MAX_WINSHIFT); sc->sc_requested_r_scale = min((data >> 6) & 0xf, TCP_MAX_WINSHIFT); if (sc->sc_requested_s_scale || sc->sc_requested_r_scale) sc->sc_flags |= SCF_WINSCALE; } else sc->sc_flags |= SCF_NOOPT; wnd = sbspace(&so->so_rcv); wnd = imax(wnd, 0); wnd = imin(wnd, TCP_MAXWIN); sc->sc_wnd = wnd; sc->sc_rxmits = 0; sc->sc_peer_mss = tcp_sc_msstab[mss]; TCPSTAT_INC(tcps_sc_recvcookie); return (sc); } /* * Returns the current number of syncache entries. This number * will probably change before you get around to calling * syncache_pcblist. */ int syncache_pcbcount(void) { struct syncache_head *sch; int count, i; for (count = 0, i = 0; i < V_tcp_syncache.hashsize; i++) { /* No need to lock for a read. */ sch = &V_tcp_syncache.hashbase[i]; count += sch->sch_length; } return count; } /* * Exports the syncache entries to userland so that netstat can display * them alongside the other sockets. This function is intended to be * called only from tcp_pcblist. * * Due to concurrency on an active system, the number of pcbs exported * may have no relation to max_pcbs. max_pcbs merely indicates the * amount of space the caller allocated for this function to use. */ int syncache_pcblist(struct sysctl_req *req, int max_pcbs, int *pcbs_exported) { struct xtcpcb xt; struct syncache *sc; struct syncache_head *sch; int count, error, i; for (count = 0, error = 0, i = 0; i < V_tcp_syncache.hashsize; i++) { sch = &V_tcp_syncache.hashbase[i]; SCH_LOCK(sch); TAILQ_FOREACH(sc, &sch->sch_bucket, sc_hash) { if (count >= max_pcbs) { SCH_UNLOCK(sch); goto exit; } if (cr_cansee(req->td->td_ucred, sc->sc_cred) != 0) continue; bzero(&xt, sizeof(xt)); xt.xt_len = sizeof(xt); if (sc->sc_inc.inc_flags & INC_ISIPV6) xt.xt_inp.inp_vflag = INP_IPV6; else xt.xt_inp.inp_vflag = INP_IPV4; bcopy(&sc->sc_inc, &xt.xt_inp.inp_inc, sizeof (struct in_conninfo)); xt.xt_tp.t_inpcb = &xt.xt_inp; xt.xt_tp.t_state = TCPS_SYN_RECEIVED; xt.xt_socket.xso_protocol = IPPROTO_TCP; xt.xt_socket.xso_len = sizeof (struct xsocket); xt.xt_socket.so_type = SOCK_STREAM; xt.xt_socket.so_state = SS_ISCONNECTING; error = SYSCTL_OUT(req, &xt, sizeof xt); if (error) { SCH_UNLOCK(sch); goto exit; } count++; } SCH_UNLOCK(sch); } exit: *pcbs_exported = count; return error; } Index: head/sys/netinet/tcp_timer.c =================================================================== --- head/sys/netinet/tcp_timer.c (revision 222487) +++ head/sys/netinet/tcp_timer.c (revision 222488) @@ -1,674 +1,674 @@ /*- * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)tcp_timer.c 8.2 (Berkeley) 5/24/95 */ #include __FBSDID("$FreeBSD$"); #include "opt_inet6.h" #include "opt_tcpdebug.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef INET6 #include #endif #include #include #include #include #include #ifdef TCPDEBUG #include #endif int tcp_keepinit; SYSCTL_PROC(_net_inet_tcp, TCPCTL_KEEPINIT, keepinit, CTLTYPE_INT|CTLFLAG_RW, &tcp_keepinit, 0, sysctl_msec_to_ticks, "I", "time to establish connection"); int tcp_keepidle; SYSCTL_PROC(_net_inet_tcp, TCPCTL_KEEPIDLE, keepidle, CTLTYPE_INT|CTLFLAG_RW, &tcp_keepidle, 0, sysctl_msec_to_ticks, "I", "time before keepalive probes begin"); int tcp_keepintvl; SYSCTL_PROC(_net_inet_tcp, TCPCTL_KEEPINTVL, keepintvl, CTLTYPE_INT|CTLFLAG_RW, &tcp_keepintvl, 0, sysctl_msec_to_ticks, "I", "time between keepalive probes"); int tcp_delacktime; SYSCTL_PROC(_net_inet_tcp, TCPCTL_DELACKTIME, delacktime, CTLTYPE_INT|CTLFLAG_RW, &tcp_delacktime, 0, sysctl_msec_to_ticks, "I", "Time before a delayed ACK is sent"); int tcp_msl; SYSCTL_PROC(_net_inet_tcp, OID_AUTO, msl, CTLTYPE_INT|CTLFLAG_RW, &tcp_msl, 0, sysctl_msec_to_ticks, "I", "Maximum segment lifetime"); int tcp_rexmit_min; SYSCTL_PROC(_net_inet_tcp, OID_AUTO, rexmit_min, CTLTYPE_INT|CTLFLAG_RW, &tcp_rexmit_min, 0, sysctl_msec_to_ticks, "I", "Minimum Retransmission Timeout"); int tcp_rexmit_slop; SYSCTL_PROC(_net_inet_tcp, OID_AUTO, rexmit_slop, CTLTYPE_INT|CTLFLAG_RW, &tcp_rexmit_slop, 0, sysctl_msec_to_ticks, "I", "Retransmission Timer Slop"); static int always_keepalive = 1; SYSCTL_INT(_net_inet_tcp, OID_AUTO, always_keepalive, CTLFLAG_RW, &always_keepalive , 0, "Assume SO_KEEPALIVE on all TCP connections"); int tcp_fast_finwait2_recycle = 0; SYSCTL_INT(_net_inet_tcp, OID_AUTO, fast_finwait2_recycle, CTLFLAG_RW, &tcp_fast_finwait2_recycle, 0, "Recycle closed FIN_WAIT_2 connections faster"); int tcp_finwait2_timeout; SYSCTL_PROC(_net_inet_tcp, OID_AUTO, finwait2_timeout, CTLTYPE_INT|CTLFLAG_RW, &tcp_finwait2_timeout, 0, sysctl_msec_to_ticks, "I", "FIN-WAIT2 timeout"); static int tcp_keepcnt = TCPTV_KEEPCNT; /* max idle probes */ int tcp_maxpersistidle; /* max idle time in persist */ int tcp_maxidle; static int per_cpu_timers = 0; SYSCTL_INT(_net_inet_tcp, OID_AUTO, per_cpu_timers, CTLFLAG_RW, &per_cpu_timers , 0, "run tcp timers on all cpus"); #define INP_CPU(inp) (per_cpu_timers ? (!CPU_ABSENT(((inp)->inp_flowid % (mp_maxid+1))) ? \ ((inp)->inp_flowid % (mp_maxid+1)) : curcpu) : 0) /* * Tcp protocol timeout routine called every 500 ms. * Updates timestamps used for TCP * causes finite state machine actions if timers expire. */ void tcp_slowtimo(void) { VNET_ITERATOR_DECL(vnet_iter); VNET_LIST_RLOCK_NOSLEEP(); VNET_FOREACH(vnet_iter) { CURVNET_SET(vnet_iter); tcp_maxidle = tcp_keepcnt * tcp_keepintvl; INP_INFO_WLOCK(&V_tcbinfo); (void) tcp_tw_2msl_scan(0); INP_INFO_WUNLOCK(&V_tcbinfo); CURVNET_RESTORE(); } VNET_LIST_RUNLOCK_NOSLEEP(); } int tcp_syn_backoff[TCP_MAXRXTSHIFT + 1] = { 1, 1, 1, 1, 1, 2, 4, 8, 16, 32, 64, 64, 64 }; int tcp_backoff[TCP_MAXRXTSHIFT + 1] = { 1, 2, 4, 8, 16, 32, 64, 128, 256, 512, 512, 512, 512 }; static int tcp_totbackoff = 2559; /* sum of tcp_backoff[] */ static int tcp_timer_race; SYSCTL_INT(_net_inet_tcp, OID_AUTO, timer_race, CTLFLAG_RD, &tcp_timer_race, 0, "Count of t_inpcb races on tcp_discardcb"); /* * TCP timer processing. */ void tcp_timer_delack(void *xtp) { struct tcpcb *tp = xtp; struct inpcb *inp; CURVNET_SET(tp->t_vnet); inp = tp->t_inpcb; /* * XXXRW: While this assert is in fact correct, bugs in the tcpcb * tear-down mean we need it as a work-around for races between * timers and tcp_discardcb(). * * KASSERT(inp != NULL, ("tcp_timer_delack: inp == NULL")); */ if (inp == NULL) { tcp_timer_race++; CURVNET_RESTORE(); return; } INP_WLOCK(inp); if ((inp->inp_flags & INP_DROPPED) || callout_pending(&tp->t_timers->tt_delack) || !callout_active(&tp->t_timers->tt_delack)) { INP_WUNLOCK(inp); CURVNET_RESTORE(); return; } callout_deactivate(&tp->t_timers->tt_delack); tp->t_flags |= TF_ACKNOW; TCPSTAT_INC(tcps_delack); (void) tcp_output(tp); INP_WUNLOCK(inp); CURVNET_RESTORE(); } void tcp_timer_2msl(void *xtp) { struct tcpcb *tp = xtp; struct inpcb *inp; CURVNET_SET(tp->t_vnet); #ifdef TCPDEBUG int ostate; ostate = tp->t_state; #endif /* * XXXRW: Does this actually happen? */ INP_INFO_WLOCK(&V_tcbinfo); inp = tp->t_inpcb; /* * XXXRW: While this assert is in fact correct, bugs in the tcpcb * tear-down mean we need it as a work-around for races between * timers and tcp_discardcb(). * * KASSERT(inp != NULL, ("tcp_timer_2msl: inp == NULL")); */ if (inp == NULL) { tcp_timer_race++; INP_INFO_WUNLOCK(&V_tcbinfo); CURVNET_RESTORE(); return; } INP_WLOCK(inp); tcp_free_sackholes(tp); if ((inp->inp_flags & INP_DROPPED) || callout_pending(&tp->t_timers->tt_2msl) || !callout_active(&tp->t_timers->tt_2msl)) { INP_WUNLOCK(tp->t_inpcb); INP_INFO_WUNLOCK(&V_tcbinfo); CURVNET_RESTORE(); return; } callout_deactivate(&tp->t_timers->tt_2msl); /* * 2 MSL timeout in shutdown went off. If we're closed but * still waiting for peer to close and connection has been idle * too long, or if 2MSL time is up from TIME_WAIT, delete connection * control block. Otherwise, check again in a bit. * * If fastrecycle of FIN_WAIT_2, in FIN_WAIT_2 and receiver has closed, * there's no point in hanging onto FIN_WAIT_2 socket. Just close it. * Ignore fact that there were recent incoming segments. */ if (tcp_fast_finwait2_recycle && tp->t_state == TCPS_FIN_WAIT_2 && tp->t_inpcb && tp->t_inpcb->inp_socket && (tp->t_inpcb->inp_socket->so_rcv.sb_state & SBS_CANTRCVMORE)) { TCPSTAT_INC(tcps_finwait2_drops); tp = tcp_close(tp); } else { if (tp->t_state != TCPS_TIME_WAIT && ticks - tp->t_rcvtime <= tcp_maxidle) callout_reset_on(&tp->t_timers->tt_2msl, tcp_keepintvl, tcp_timer_2msl, tp, INP_CPU(inp)); else tp = tcp_close(tp); } #ifdef TCPDEBUG if (tp != NULL && (tp->t_inpcb->inp_socket->so_options & SO_DEBUG)) tcp_trace(TA_USER, ostate, tp, (void *)0, (struct tcphdr *)0, PRU_SLOWTIMO); #endif if (tp != NULL) INP_WUNLOCK(inp); INP_INFO_WUNLOCK(&V_tcbinfo); CURVNET_RESTORE(); } void tcp_timer_keep(void *xtp) { struct tcpcb *tp = xtp; struct tcptemp *t_template; struct inpcb *inp; CURVNET_SET(tp->t_vnet); #ifdef TCPDEBUG int ostate; ostate = tp->t_state; #endif INP_INFO_WLOCK(&V_tcbinfo); inp = tp->t_inpcb; /* * XXXRW: While this assert is in fact correct, bugs in the tcpcb * tear-down mean we need it as a work-around for races between * timers and tcp_discardcb(). * * KASSERT(inp != NULL, ("tcp_timer_keep: inp == NULL")); */ if (inp == NULL) { tcp_timer_race++; INP_INFO_WUNLOCK(&V_tcbinfo); CURVNET_RESTORE(); return; } INP_WLOCK(inp); if ((inp->inp_flags & INP_DROPPED) || callout_pending(&tp->t_timers->tt_keep) || !callout_active(&tp->t_timers->tt_keep)) { INP_WUNLOCK(inp); INP_INFO_WUNLOCK(&V_tcbinfo); CURVNET_RESTORE(); return; } callout_deactivate(&tp->t_timers->tt_keep); /* * Keep-alive timer went off; send something * or drop connection if idle for too long. */ TCPSTAT_INC(tcps_keeptimeo); if (tp->t_state < TCPS_ESTABLISHED) goto dropit; if ((always_keepalive || inp->inp_socket->so_options & SO_KEEPALIVE) && tp->t_state <= TCPS_CLOSING) { if (ticks - tp->t_rcvtime >= tcp_keepidle + tcp_maxidle) goto dropit; /* * Send a packet designed to force a response * if the peer is up and reachable: * either an ACK if the connection is still alive, * or an RST if the peer has closed the connection * due to timeout or reboot. * Using sequence number tp->snd_una-1 * causes the transmitted zero-length segment * to lie outside the receive window; * by the protocol spec, this requires the * correspondent TCP to respond. */ TCPSTAT_INC(tcps_keepprobe); t_template = tcpip_maketemplate(inp); if (t_template) { tcp_respond(tp, t_template->tt_ipgen, &t_template->tt_t, (struct mbuf *)NULL, tp->rcv_nxt, tp->snd_una - 1, 0); free(t_template, M_TEMP); } callout_reset_on(&tp->t_timers->tt_keep, tcp_keepintvl, tcp_timer_keep, tp, INP_CPU(inp)); } else callout_reset_on(&tp->t_timers->tt_keep, tcp_keepidle, tcp_timer_keep, tp, INP_CPU(inp)); #ifdef TCPDEBUG if (inp->inp_socket->so_options & SO_DEBUG) tcp_trace(TA_USER, ostate, tp, (void *)0, (struct tcphdr *)0, PRU_SLOWTIMO); #endif INP_WUNLOCK(inp); INP_INFO_WUNLOCK(&V_tcbinfo); CURVNET_RESTORE(); return; dropit: TCPSTAT_INC(tcps_keepdrops); tp = tcp_drop(tp, ETIMEDOUT); #ifdef TCPDEBUG if (tp != NULL && (tp->t_inpcb->inp_socket->so_options & SO_DEBUG)) tcp_trace(TA_USER, ostate, tp, (void *)0, (struct tcphdr *)0, PRU_SLOWTIMO); #endif if (tp != NULL) INP_WUNLOCK(tp->t_inpcb); INP_INFO_WUNLOCK(&V_tcbinfo); CURVNET_RESTORE(); } void tcp_timer_persist(void *xtp) { struct tcpcb *tp = xtp; struct inpcb *inp; CURVNET_SET(tp->t_vnet); #ifdef TCPDEBUG int ostate; ostate = tp->t_state; #endif INP_INFO_WLOCK(&V_tcbinfo); inp = tp->t_inpcb; /* * XXXRW: While this assert is in fact correct, bugs in the tcpcb * tear-down mean we need it as a work-around for races between * timers and tcp_discardcb(). * * KASSERT(inp != NULL, ("tcp_timer_persist: inp == NULL")); */ if (inp == NULL) { tcp_timer_race++; INP_INFO_WUNLOCK(&V_tcbinfo); CURVNET_RESTORE(); return; } INP_WLOCK(inp); if ((inp->inp_flags & INP_DROPPED) || callout_pending(&tp->t_timers->tt_persist) || !callout_active(&tp->t_timers->tt_persist)) { INP_WUNLOCK(inp); INP_INFO_WUNLOCK(&V_tcbinfo); CURVNET_RESTORE(); return; } callout_deactivate(&tp->t_timers->tt_persist); /* * Persistance timer into zero window. * Force a byte to be output, if possible. */ TCPSTAT_INC(tcps_persisttimeo); /* * Hack: if the peer is dead/unreachable, we do not * time out if the window is closed. After a full * backoff, drop the connection if the idle time * (no responses to probes) reaches the maximum * backoff that we would use if retransmitting. */ if (tp->t_rxtshift == TCP_MAXRXTSHIFT && (ticks - tp->t_rcvtime >= tcp_maxpersistidle || ticks - tp->t_rcvtime >= TCP_REXMTVAL(tp) * tcp_totbackoff)) { TCPSTAT_INC(tcps_persistdrop); tp = tcp_drop(tp, ETIMEDOUT); goto out; } tcp_setpersist(tp); tp->t_flags |= TF_FORCEDATA; (void) tcp_output(tp); tp->t_flags &= ~TF_FORCEDATA; out: #ifdef TCPDEBUG if (tp != NULL && tp->t_inpcb->inp_socket->so_options & SO_DEBUG) tcp_trace(TA_USER, ostate, tp, NULL, NULL, PRU_SLOWTIMO); #endif if (tp != NULL) INP_WUNLOCK(inp); INP_INFO_WUNLOCK(&V_tcbinfo); CURVNET_RESTORE(); } void tcp_timer_rexmt(void * xtp) { struct tcpcb *tp = xtp; CURVNET_SET(tp->t_vnet); int rexmt; int headlocked; struct inpcb *inp; #ifdef TCPDEBUG int ostate; ostate = tp->t_state; #endif INP_INFO_RLOCK(&V_tcbinfo); inp = tp->t_inpcb; /* * XXXRW: While this assert is in fact correct, bugs in the tcpcb * tear-down mean we need it as a work-around for races between * timers and tcp_discardcb(). * * KASSERT(inp != NULL, ("tcp_timer_rexmt: inp == NULL")); */ if (inp == NULL) { tcp_timer_race++; INP_INFO_RUNLOCK(&V_tcbinfo); CURVNET_RESTORE(); return; } INP_WLOCK(inp); if ((inp->inp_flags & INP_DROPPED) || callout_pending(&tp->t_timers->tt_rexmt) || !callout_active(&tp->t_timers->tt_rexmt)) { INP_WUNLOCK(inp); INP_INFO_RUNLOCK(&V_tcbinfo); CURVNET_RESTORE(); return; } callout_deactivate(&tp->t_timers->tt_rexmt); tcp_free_sackholes(tp); /* * Retransmission timer went off. Message has not * been acked within retransmit interval. Back off * to a longer retransmit interval and retransmit one segment. */ if (++tp->t_rxtshift > TCP_MAXRXTSHIFT) { tp->t_rxtshift = TCP_MAXRXTSHIFT; TCPSTAT_INC(tcps_timeoutdrop); in_pcbref(inp); INP_INFO_RUNLOCK(&V_tcbinfo); INP_WUNLOCK(inp); INP_INFO_WLOCK(&V_tcbinfo); INP_WLOCK(inp); - if (in_pcbrele(inp)) { + if (in_pcbrele_wlocked(inp)) { INP_INFO_WUNLOCK(&V_tcbinfo); CURVNET_RESTORE(); return; } tp = tcp_drop(tp, tp->t_softerror ? tp->t_softerror : ETIMEDOUT); headlocked = 1; goto out; } INP_INFO_RUNLOCK(&V_tcbinfo); headlocked = 0; if (tp->t_rxtshift == 1) { /* * first retransmit; record ssthresh and cwnd so they can * be recovered if this turns out to be a "bad" retransmit. * A retransmit is considered "bad" if an ACK for this * segment is received within RTT/2 interval; the assumption * here is that the ACK was already in flight. See * "On Estimating End-to-End Network Path Properties" by * Allman and Paxson for more details. */ tp->snd_cwnd_prev = tp->snd_cwnd; tp->snd_ssthresh_prev = tp->snd_ssthresh; tp->snd_recover_prev = tp->snd_recover; if (IN_FASTRECOVERY(tp->t_flags)) tp->t_flags |= TF_WASFRECOVERY; else tp->t_flags &= ~TF_WASFRECOVERY; if (IN_CONGRECOVERY(tp->t_flags)) tp->t_flags |= TF_WASCRECOVERY; else tp->t_flags &= ~TF_WASCRECOVERY; tp->t_badrxtwin = ticks + (tp->t_srtt >> (TCP_RTT_SHIFT + 1)); tp->t_flags |= TF_PREVVALID; } else tp->t_flags &= ~TF_PREVVALID; TCPSTAT_INC(tcps_rexmttimeo); if (tp->t_state == TCPS_SYN_SENT) rexmt = TCP_REXMTVAL(tp) * tcp_syn_backoff[tp->t_rxtshift]; else rexmt = TCP_REXMTVAL(tp) * tcp_backoff[tp->t_rxtshift]; TCPT_RANGESET(tp->t_rxtcur, rexmt, tp->t_rttmin, TCPTV_REXMTMAX); /* * Disable rfc1323 if we haven't got any response to * our third SYN to work-around some broken terminal servers * (most of which have hopefully been retired) that have bad VJ * header compression code which trashes TCP segments containing * unknown-to-them TCP options. */ if ((tp->t_state == TCPS_SYN_SENT) && (tp->t_rxtshift == 3)) tp->t_flags &= ~(TF_REQ_SCALE|TF_REQ_TSTMP); /* * If we backed off this far, our srtt estimate is probably bogus. * Clobber it so we'll take the next rtt measurement as our srtt; * move the current srtt into rttvar to keep the current * retransmit times until then. */ if (tp->t_rxtshift > TCP_MAXRXTSHIFT / 4) { #ifdef INET6 if ((tp->t_inpcb->inp_vflag & INP_IPV6) != 0) in6_losing(tp->t_inpcb); else #endif tp->t_rttvar += (tp->t_srtt >> TCP_RTT_SHIFT); tp->t_srtt = 0; } tp->snd_nxt = tp->snd_una; tp->snd_recover = tp->snd_max; /* * Force a segment to be sent. */ tp->t_flags |= TF_ACKNOW; /* * If timing a segment in this window, stop the timer. */ tp->t_rtttime = 0; cc_cong_signal(tp, NULL, CC_RTO); (void) tcp_output(tp); out: #ifdef TCPDEBUG if (tp != NULL && (tp->t_inpcb->inp_socket->so_options & SO_DEBUG)) tcp_trace(TA_USER, ostate, tp, (void *)0, (struct tcphdr *)0, PRU_SLOWTIMO); #endif if (tp != NULL) INP_WUNLOCK(inp); if (headlocked) INP_INFO_WUNLOCK(&V_tcbinfo); CURVNET_RESTORE(); } void tcp_timer_activate(struct tcpcb *tp, int timer_type, u_int delta) { struct callout *t_callout; void *f_callout; struct inpcb *inp = tp->t_inpcb; int cpu = INP_CPU(inp); switch (timer_type) { case TT_DELACK: t_callout = &tp->t_timers->tt_delack; f_callout = tcp_timer_delack; break; case TT_REXMT: t_callout = &tp->t_timers->tt_rexmt; f_callout = tcp_timer_rexmt; break; case TT_PERSIST: t_callout = &tp->t_timers->tt_persist; f_callout = tcp_timer_persist; break; case TT_KEEP: t_callout = &tp->t_timers->tt_keep; f_callout = tcp_timer_keep; break; case TT_2MSL: t_callout = &tp->t_timers->tt_2msl; f_callout = tcp_timer_2msl; break; default: panic("bad timer_type"); } if (delta == 0) { callout_stop(t_callout); } else { callout_reset_on(t_callout, delta, f_callout, tp, cpu); } } int tcp_timer_active(struct tcpcb *tp, int timer_type) { struct callout *t_callout; switch (timer_type) { case TT_DELACK: t_callout = &tp->t_timers->tt_delack; break; case TT_REXMT: t_callout = &tp->t_timers->tt_rexmt; break; case TT_PERSIST: t_callout = &tp->t_timers->tt_persist; break; case TT_KEEP: t_callout = &tp->t_timers->tt_keep; break; case TT_2MSL: t_callout = &tp->t_timers->tt_2msl; break; default: panic("bad timer_type"); } return callout_active(t_callout); } #define ticks_to_msecs(t) (1000*(t) / hz) void tcp_timer_to_xtimer(struct tcpcb *tp, struct tcp_timer *timer, struct xtcp_timer *xtimer) { bzero(xtimer, sizeof(struct xtcp_timer)); if (timer == NULL) return; if (callout_active(&timer->tt_delack)) xtimer->tt_delack = ticks_to_msecs(timer->tt_delack.c_time - ticks); if (callout_active(&timer->tt_rexmt)) xtimer->tt_rexmt = ticks_to_msecs(timer->tt_rexmt.c_time - ticks); if (callout_active(&timer->tt_persist)) xtimer->tt_persist = ticks_to_msecs(timer->tt_persist.c_time - ticks); if (callout_active(&timer->tt_keep)) xtimer->tt_keep = ticks_to_msecs(timer->tt_keep.c_time - ticks); if (callout_active(&timer->tt_2msl)) xtimer->tt_2msl = ticks_to_msecs(timer->tt_2msl.c_time - ticks); xtimer->t_rcvtime = ticks_to_msecs(ticks - tp->t_rcvtime); } Index: head/sys/netinet/tcp_usrreq.c =================================================================== --- head/sys/netinet/tcp_usrreq.c (revision 222487) +++ head/sys/netinet/tcp_usrreq.c (revision 222488) @@ -1,1959 +1,1965 @@ /*- * Copyright (c) 1982, 1986, 1988, 1993 * The Regents of the University of California. * Copyright (c) 2006-2007 Robert N. M. Watson + * Copyright (c) 2010-2011 Juniper Networks, Inc. * All rights reserved. * + * Portions of this software were developed by Robert N. M. Watson under + * contract to Juniper Networks, Inc. + * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * From: @(#)tcp_usrreq.c 8.2 (Berkeley) 1/3/94 */ #include __FBSDID("$FreeBSD$"); #include "opt_ddb.h" #include "opt_inet.h" #include "opt_inet6.h" #include "opt_tcpdebug.h" #include #include #include #include #include #include #ifdef INET6 #include #endif /* INET6 */ #include #include #include #include #include #ifdef DDB #include #endif #include #include #include #include #include #include #include #include #include #ifdef INET6 #include #include #include #include #endif #include #include #include #include #include #ifdef TCPDEBUG #include #endif #include /* * TCP protocol interface to socket abstraction. */ static int tcp_attach(struct socket *); #ifdef INET static int tcp_connect(struct tcpcb *, struct sockaddr *, struct thread *td); #endif /* INET */ #ifdef INET6 static int tcp6_connect(struct tcpcb *, struct sockaddr *, struct thread *td); #endif /* INET6 */ static void tcp_disconnect(struct tcpcb *); static void tcp_usrclosed(struct tcpcb *); static void tcp_fill_info(struct tcpcb *, struct tcp_info *); #ifdef TCPDEBUG #define TCPDEBUG0 int ostate = 0 #define TCPDEBUG1() ostate = tp ? tp->t_state : 0 #define TCPDEBUG2(req) if (tp && (so->so_options & SO_DEBUG)) \ tcp_trace(TA_USER, ostate, tp, 0, 0, req) #else #define TCPDEBUG0 #define TCPDEBUG1() #define TCPDEBUG2(req) #endif /* * TCP attaches to socket via pru_attach(), reserving space, * and an internet control block. */ static int tcp_usr_attach(struct socket *so, int proto, struct thread *td) { struct inpcb *inp; struct tcpcb *tp = NULL; int error; TCPDEBUG0; inp = sotoinpcb(so); KASSERT(inp == NULL, ("tcp_usr_attach: inp != NULL")); TCPDEBUG1(); error = tcp_attach(so); if (error) goto out; if ((so->so_options & SO_LINGER) && so->so_linger == 0) so->so_linger = TCP_LINGERTIME; inp = sotoinpcb(so); tp = intotcpcb(inp); out: TCPDEBUG2(PRU_ATTACH); return error; } /* * tcp_detach is called when the socket layer loses its final reference * to the socket, be it a file descriptor reference, a reference from TCP, * etc. At this point, there is only one case in which we will keep around * inpcb state: time wait. * * This function can probably be re-absorbed back into tcp_usr_detach() now * that there is a single detach path. */ static void tcp_detach(struct socket *so, struct inpcb *inp) { struct tcpcb *tp; INP_INFO_WLOCK_ASSERT(&V_tcbinfo); INP_WLOCK_ASSERT(inp); KASSERT(so->so_pcb == inp, ("tcp_detach: so_pcb != inp")); KASSERT(inp->inp_socket == so, ("tcp_detach: inp_socket != so")); tp = intotcpcb(inp); if (inp->inp_flags & INP_TIMEWAIT) { /* * There are two cases to handle: one in which the time wait * state is being discarded (INP_DROPPED), and one in which * this connection will remain in timewait. In the former, * it is time to discard all state (except tcptw, which has * already been discarded by the timewait close code, which * should be further up the call stack somewhere). In the * latter case, we detach from the socket, but leave the pcb * present until timewait ends. * * XXXRW: Would it be cleaner to free the tcptw here? */ if (inp->inp_flags & INP_DROPPED) { KASSERT(tp == NULL, ("tcp_detach: INP_TIMEWAIT && " "INP_DROPPED && tp != NULL")); in_pcbdetach(inp); in_pcbfree(inp); } else { in_pcbdetach(inp); INP_WUNLOCK(inp); } } else { /* * If the connection is not in timewait, we consider two * two conditions: one in which no further processing is * necessary (dropped || embryonic), and one in which TCP is * not yet done, but no longer requires the socket, so the * pcb will persist for the time being. * * XXXRW: Does the second case still occur? */ if (inp->inp_flags & INP_DROPPED || tp->t_state < TCPS_SYN_SENT) { tcp_discardcb(tp); in_pcbdetach(inp); in_pcbfree(inp); } else in_pcbdetach(inp); } } /* * pru_detach() detaches the TCP protocol from the socket. * If the protocol state is non-embryonic, then can't * do this directly: have to initiate a pru_disconnect(), * which may finish later; embryonic TCB's can just * be discarded here. */ static void tcp_usr_detach(struct socket *so) { struct inpcb *inp; inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_detach: inp == NULL")); INP_INFO_WLOCK(&V_tcbinfo); INP_WLOCK(inp); KASSERT(inp->inp_socket != NULL, ("tcp_usr_detach: inp_socket == NULL")); tcp_detach(so, inp); INP_INFO_WUNLOCK(&V_tcbinfo); } #ifdef INET /* * Give the socket an address. */ static int tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td) { int error = 0; struct inpcb *inp; struct tcpcb *tp = NULL; struct sockaddr_in *sinp; sinp = (struct sockaddr_in *)nam; if (nam->sa_len != sizeof (*sinp)) return (EINVAL); /* * Must check for multicast addresses and disallow binding * to them. */ if (sinp->sin_family == AF_INET && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) return (EAFNOSUPPORT); TCPDEBUG0; - INP_INFO_WLOCK(&V_tcbinfo); inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_bind: inp == NULL")); INP_WLOCK(inp); if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { error = EINVAL; goto out; } tp = intotcpcb(inp); TCPDEBUG1(); + INP_HASH_WLOCK(&V_tcbinfo); error = in_pcbbind(inp, nam, td->td_ucred); + INP_HASH_WUNLOCK(&V_tcbinfo); out: TCPDEBUG2(PRU_BIND); INP_WUNLOCK(inp); - INP_INFO_WUNLOCK(&V_tcbinfo); return (error); } #endif /* INET */ #ifdef INET6 static int tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td) { int error = 0; struct inpcb *inp; struct tcpcb *tp = NULL; struct sockaddr_in6 *sin6p; sin6p = (struct sockaddr_in6 *)nam; if (nam->sa_len != sizeof (*sin6p)) return (EINVAL); /* * Must check for multicast addresses and disallow binding * to them. */ if (sin6p->sin6_family == AF_INET6 && IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) return (EAFNOSUPPORT); TCPDEBUG0; - INP_INFO_WLOCK(&V_tcbinfo); inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp6_usr_bind: inp == NULL")); INP_WLOCK(inp); if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { error = EINVAL; goto out; } tp = intotcpcb(inp); TCPDEBUG1(); + INP_HASH_WLOCK(&V_tcbinfo); inp->inp_vflag &= ~INP_IPV4; inp->inp_vflag |= INP_IPV6; #ifdef INET if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) { if (IN6_IS_ADDR_UNSPECIFIED(&sin6p->sin6_addr)) inp->inp_vflag |= INP_IPV4; else if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) { struct sockaddr_in sin; in6_sin6_2_sin(&sin, sin6p); inp->inp_vflag |= INP_IPV4; inp->inp_vflag &= ~INP_IPV6; error = in_pcbbind(inp, (struct sockaddr *)&sin, td->td_ucred); + INP_HASH_WUNLOCK(&V_tcbinfo); goto out; } } #endif error = in6_pcbbind(inp, nam, td->td_ucred); + INP_HASH_WUNLOCK(&V_tcbinfo); out: TCPDEBUG2(PRU_BIND); INP_WUNLOCK(inp); - INP_INFO_WUNLOCK(&V_tcbinfo); return (error); } #endif /* INET6 */ #ifdef INET /* * Prepare to accept connections. */ static int tcp_usr_listen(struct socket *so, int backlog, struct thread *td) { int error = 0; struct inpcb *inp; struct tcpcb *tp = NULL; TCPDEBUG0; - INP_INFO_WLOCK(&V_tcbinfo); inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_listen: inp == NULL")); INP_WLOCK(inp); if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { error = EINVAL; goto out; } tp = intotcpcb(inp); TCPDEBUG1(); SOCK_LOCK(so); error = solisten_proto_check(so); + INP_HASH_WLOCK(&V_tcbinfo); if (error == 0 && inp->inp_lport == 0) error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred); + INP_HASH_WUNLOCK(&V_tcbinfo); if (error == 0) { tp->t_state = TCPS_LISTEN; solisten_proto(so, backlog); tcp_offload_listen_open(tp); } SOCK_UNLOCK(so); out: TCPDEBUG2(PRU_LISTEN); INP_WUNLOCK(inp); - INP_INFO_WUNLOCK(&V_tcbinfo); return (error); } #endif /* INET */ #ifdef INET6 static int tcp6_usr_listen(struct socket *so, int backlog, struct thread *td) { int error = 0; struct inpcb *inp; struct tcpcb *tp = NULL; TCPDEBUG0; - INP_INFO_WLOCK(&V_tcbinfo); inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp6_usr_listen: inp == NULL")); INP_WLOCK(inp); if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { error = EINVAL; goto out; } tp = intotcpcb(inp); TCPDEBUG1(); SOCK_LOCK(so); error = solisten_proto_check(so); + INP_HASH_WLOCK(&V_tcbinfo); if (error == 0 && inp->inp_lport == 0) { inp->inp_vflag &= ~INP_IPV4; if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) inp->inp_vflag |= INP_IPV4; error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred); } + INP_HASH_WUNLOCK(&V_tcbinfo); if (error == 0) { tp->t_state = TCPS_LISTEN; solisten_proto(so, backlog); } SOCK_UNLOCK(so); out: TCPDEBUG2(PRU_LISTEN); INP_WUNLOCK(inp); - INP_INFO_WUNLOCK(&V_tcbinfo); return (error); } #endif /* INET6 */ #ifdef INET /* * Initiate connection to peer. * Create a template for use in transmissions on this connection. * Enter SYN_SENT state, and mark socket as connecting. * Start keep-alive timer, and seed output sequence space. * Send initial segment on connection. */ static int tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td) { int error = 0; struct inpcb *inp; struct tcpcb *tp = NULL; struct sockaddr_in *sinp; sinp = (struct sockaddr_in *)nam; if (nam->sa_len != sizeof (*sinp)) return (EINVAL); /* * Must disallow TCP ``connections'' to multicast addresses. */ if (sinp->sin_family == AF_INET && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) return (EAFNOSUPPORT); if ((error = prison_remote_ip4(td->td_ucred, &sinp->sin_addr)) != 0) return (error); TCPDEBUG0; - INP_INFO_WLOCK(&V_tcbinfo); inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_connect: inp == NULL")); INP_WLOCK(inp); if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { error = EINVAL; goto out; } tp = intotcpcb(inp); TCPDEBUG1(); if ((error = tcp_connect(tp, nam, td)) != 0) goto out; error = tcp_output_connect(so, nam); out: TCPDEBUG2(PRU_CONNECT); INP_WUNLOCK(inp); - INP_INFO_WUNLOCK(&V_tcbinfo); return (error); } #endif /* INET */ #ifdef INET6 static int tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td) { int error = 0; struct inpcb *inp; struct tcpcb *tp = NULL; struct sockaddr_in6 *sin6p; TCPDEBUG0; sin6p = (struct sockaddr_in6 *)nam; if (nam->sa_len != sizeof (*sin6p)) return (EINVAL); /* * Must disallow TCP ``connections'' to multicast addresses. */ if (sin6p->sin6_family == AF_INET6 && IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) return (EAFNOSUPPORT); - INP_INFO_WLOCK(&V_tcbinfo); inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp6_usr_connect: inp == NULL")); INP_WLOCK(inp); if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { error = EINVAL; goto out; } tp = intotcpcb(inp); TCPDEBUG1(); #ifdef INET + /* + * XXXRW: Some confusion: V4/V6 flags relate to binding, and + * therefore probably require the hash lock, which isn't held here. + * Is this a significant problem? + */ if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) { struct sockaddr_in sin; if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) { error = EINVAL; goto out; } in6_sin6_2_sin(&sin, sin6p); inp->inp_vflag |= INP_IPV4; inp->inp_vflag &= ~INP_IPV6; if ((error = prison_remote_ip4(td->td_ucred, &sin.sin_addr)) != 0) goto out; if ((error = tcp_connect(tp, (struct sockaddr *)&sin, td)) != 0) goto out; error = tcp_output_connect(so, nam); goto out; } #endif inp->inp_vflag &= ~INP_IPV4; inp->inp_vflag |= INP_IPV6; inp->inp_inc.inc_flags |= INC_ISIPV6; if ((error = prison_remote_ip6(td->td_ucred, &sin6p->sin6_addr)) != 0) goto out; if ((error = tcp6_connect(tp, nam, td)) != 0) goto out; error = tcp_output_connect(so, nam); out: TCPDEBUG2(PRU_CONNECT); INP_WUNLOCK(inp); - INP_INFO_WUNLOCK(&V_tcbinfo); return (error); } #endif /* INET6 */ /* * Initiate disconnect from peer. * If connection never passed embryonic stage, just drop; * else if don't need to let data drain, then can just drop anyways, * else have to begin TCP shutdown process: mark socket disconnecting, * drain unread data, state switch to reflect user close, and * send segment (e.g. FIN) to peer. Socket will be really disconnected * when peer sends FIN and acks ours. * * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB. */ static int tcp_usr_disconnect(struct socket *so) { struct inpcb *inp; struct tcpcb *tp = NULL; int error = 0; TCPDEBUG0; INP_INFO_WLOCK(&V_tcbinfo); inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_disconnect: inp == NULL")); INP_WLOCK(inp); if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { error = ECONNRESET; goto out; } tp = intotcpcb(inp); TCPDEBUG1(); tcp_disconnect(tp); out: TCPDEBUG2(PRU_DISCONNECT); INP_WUNLOCK(inp); INP_INFO_WUNLOCK(&V_tcbinfo); return (error); } #ifdef INET /* * Accept a connection. Essentially all the work is done at higher levels; * just return the address of the peer, storing through addr. * * The rationale for acquiring the tcbinfo lock here is somewhat complicated, * and is described in detail in the commit log entry for r175612. Acquiring * it delays an accept(2) racing with sonewconn(), which inserts the socket * before the inpcb address/port fields are initialized. A better fix would * prevent the socket from being placed in the listen queue until all fields * are fully initialized. */ static int tcp_usr_accept(struct socket *so, struct sockaddr **nam) { int error = 0; struct inpcb *inp = NULL; struct tcpcb *tp = NULL; struct in_addr addr; in_port_t port = 0; TCPDEBUG0; if (so->so_state & SS_ISDISCONNECTED) return (ECONNABORTED); inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_accept: inp == NULL")); INP_INFO_RLOCK(&V_tcbinfo); INP_WLOCK(inp); if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { error = ECONNABORTED; goto out; } tp = intotcpcb(inp); TCPDEBUG1(); /* * We inline in_getpeeraddr and COMMON_END here, so that we can * copy the data of interest and defer the malloc until after we * release the lock. */ port = inp->inp_fport; addr = inp->inp_faddr; out: TCPDEBUG2(PRU_ACCEPT); INP_WUNLOCK(inp); INP_INFO_RUNLOCK(&V_tcbinfo); if (error == 0) *nam = in_sockaddr(port, &addr); return error; } #endif /* INET */ #ifdef INET6 static int tcp6_usr_accept(struct socket *so, struct sockaddr **nam) { struct inpcb *inp = NULL; int error = 0; struct tcpcb *tp = NULL; struct in_addr addr; struct in6_addr addr6; in_port_t port = 0; int v4 = 0; TCPDEBUG0; if (so->so_state & SS_ISDISCONNECTED) return (ECONNABORTED); inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp6_usr_accept: inp == NULL")); + INP_INFO_RLOCK(&V_tcbinfo); INP_WLOCK(inp); if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { error = ECONNABORTED; goto out; } tp = intotcpcb(inp); TCPDEBUG1(); /* * We inline in6_mapped_peeraddr and COMMON_END here, so that we can * copy the data of interest and defer the malloc until after we * release the lock. */ if (inp->inp_vflag & INP_IPV4) { v4 = 1; port = inp->inp_fport; addr = inp->inp_faddr; } else { port = inp->inp_fport; addr6 = inp->in6p_faddr; } out: TCPDEBUG2(PRU_ACCEPT); INP_WUNLOCK(inp); + INP_INFO_RUNLOCK(&V_tcbinfo); if (error == 0) { if (v4) *nam = in6_v4mapsin6_sockaddr(port, &addr); else *nam = in6_sockaddr(port, &addr6); } return error; } #endif /* INET6 */ /* * Mark the connection as being incapable of further output. */ static int tcp_usr_shutdown(struct socket *so) { int error = 0; struct inpcb *inp; struct tcpcb *tp = NULL; TCPDEBUG0; INP_INFO_WLOCK(&V_tcbinfo); inp = sotoinpcb(so); KASSERT(inp != NULL, ("inp == NULL")); INP_WLOCK(inp); if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { error = ECONNRESET; goto out; } tp = intotcpcb(inp); TCPDEBUG1(); socantsendmore(so); tcp_usrclosed(tp); if (!(inp->inp_flags & INP_DROPPED)) error = tcp_output_disconnect(tp); out: TCPDEBUG2(PRU_SHUTDOWN); INP_WUNLOCK(inp); INP_INFO_WUNLOCK(&V_tcbinfo); return (error); } /* * After a receive, possibly send window update to peer. */ static int tcp_usr_rcvd(struct socket *so, int flags) { struct inpcb *inp; struct tcpcb *tp = NULL; int error = 0; TCPDEBUG0; inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_rcvd: inp == NULL")); INP_WLOCK(inp); if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { error = ECONNRESET; goto out; } tp = intotcpcb(inp); TCPDEBUG1(); tcp_output_rcvd(tp); out: TCPDEBUG2(PRU_RCVD); INP_WUNLOCK(inp); return (error); } /* * Do a send by putting data in output queue and updating urgent * marker if URG set. Possibly send more data. Unlike the other * pru_*() routines, the mbuf chains are our responsibility. We * must either enqueue them or free them. The other pru_* routines * generally are caller-frees. */ static int tcp_usr_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, struct mbuf *control, struct thread *td) { int error = 0; struct inpcb *inp; struct tcpcb *tp = NULL; - int headlocked = 0; #ifdef INET6 int isipv6; #endif TCPDEBUG0; /* - * We require the pcbinfo lock in two cases: - * - * (1) An implied connect is taking place, which can result in - * binding IPs and ports and hence modification of the pcb hash - * chains. - * - * (2) PRUS_EOF is set, resulting in explicit close on the send. + * We require the pcbinfo lock if we will close the socket as part of + * this call. */ - if ((nam != NULL) || (flags & PRUS_EOF)) { + if (flags & PRUS_EOF) INP_INFO_WLOCK(&V_tcbinfo); - headlocked = 1; - } inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_send: inp == NULL")); INP_WLOCK(inp); if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { if (control) m_freem(control); if (m) m_freem(m); error = ECONNRESET; goto out; } #ifdef INET6 isipv6 = nam && nam->sa_family == AF_INET6; #endif /* INET6 */ tp = intotcpcb(inp); TCPDEBUG1(); if (control) { /* TCP doesn't do control messages (rights, creds, etc) */ if (control->m_len) { m_freem(control); if (m) m_freem(m); error = EINVAL; goto out; } m_freem(control); /* empty control, just free it */ } if (!(flags & PRUS_OOB)) { sbappendstream(&so->so_snd, m); if (nam && tp->t_state < TCPS_SYN_SENT) { /* * Do implied connect if not yet connected, * initialize window to default value, and * initialize maxseg/maxopd using peer's cached * MSS. */ - INP_INFO_WLOCK_ASSERT(&V_tcbinfo); #ifdef INET6 if (isipv6) error = tcp6_connect(tp, nam, td); #endif /* INET6 */ #if defined(INET6) && defined(INET) else #endif #ifdef INET error = tcp_connect(tp, nam, td); #endif if (error) goto out; tp->snd_wnd = TTCP_CLIENT_SND_WND; tcp_mss(tp, -1); } if (flags & PRUS_EOF) { /* * Close the send side of the connection after * the data is sent. */ INP_INFO_WLOCK_ASSERT(&V_tcbinfo); socantsendmore(so); tcp_usrclosed(tp); } - if (headlocked) { - INP_INFO_WUNLOCK(&V_tcbinfo); - headlocked = 0; - } if (!(inp->inp_flags & INP_DROPPED)) { if (flags & PRUS_MORETOCOME) tp->t_flags |= TF_MORETOCOME; error = tcp_output_send(tp); if (flags & PRUS_MORETOCOME) tp->t_flags &= ~TF_MORETOCOME; } } else { /* * XXXRW: PRUS_EOF not implemented with PRUS_OOB? */ SOCKBUF_LOCK(&so->so_snd); if (sbspace(&so->so_snd) < -512) { SOCKBUF_UNLOCK(&so->so_snd); m_freem(m); error = ENOBUFS; goto out; } /* * According to RFC961 (Assigned Protocols), * the urgent pointer points to the last octet * of urgent data. We continue, however, * to consider it to indicate the first octet * of data past the urgent section. * Otherwise, snd_up should be one lower. */ sbappendstream_locked(&so->so_snd, m); SOCKBUF_UNLOCK(&so->so_snd); if (nam && tp->t_state < TCPS_SYN_SENT) { /* * Do implied connect if not yet connected, * initialize window to default value, and * initialize maxseg/maxopd using peer's cached * MSS. */ - INP_INFO_WLOCK_ASSERT(&V_tcbinfo); #ifdef INET6 if (isipv6) error = tcp6_connect(tp, nam, td); #endif /* INET6 */ #if defined(INET6) && defined(INET) else #endif #ifdef INET error = tcp_connect(tp, nam, td); #endif if (error) goto out; tp->snd_wnd = TTCP_CLIENT_SND_WND; tcp_mss(tp, -1); - INP_INFO_WUNLOCK(&V_tcbinfo); - headlocked = 0; - } else if (nam) { - INP_INFO_WUNLOCK(&V_tcbinfo); - headlocked = 0; } tp->snd_up = tp->snd_una + so->so_snd.sb_cc; tp->t_flags |= TF_FORCEDATA; error = tcp_output_send(tp); tp->t_flags &= ~TF_FORCEDATA; } out: TCPDEBUG2((flags & PRUS_OOB) ? PRU_SENDOOB : ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND)); INP_WUNLOCK(inp); - if (headlocked) + if (flags & PRUS_EOF) INP_INFO_WUNLOCK(&V_tcbinfo); return (error); } /* * Abort the TCP. Drop the connection abruptly. */ static void tcp_usr_abort(struct socket *so) { struct inpcb *inp; struct tcpcb *tp = NULL; TCPDEBUG0; inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_abort: inp == NULL")); INP_INFO_WLOCK(&V_tcbinfo); INP_WLOCK(inp); KASSERT(inp->inp_socket != NULL, ("tcp_usr_abort: inp_socket == NULL")); /* * If we still have full TCP state, and we're not dropped, drop. */ if (!(inp->inp_flags & INP_TIMEWAIT) && !(inp->inp_flags & INP_DROPPED)) { tp = intotcpcb(inp); TCPDEBUG1(); tcp_drop(tp, ECONNABORTED); TCPDEBUG2(PRU_ABORT); } if (!(inp->inp_flags & INP_DROPPED)) { SOCK_LOCK(so); so->so_state |= SS_PROTOREF; SOCK_UNLOCK(so); inp->inp_flags |= INP_SOCKREF; } INP_WUNLOCK(inp); INP_INFO_WUNLOCK(&V_tcbinfo); } /* * TCP socket is closed. Start friendly disconnect. */ static void tcp_usr_close(struct socket *so) { struct inpcb *inp; struct tcpcb *tp = NULL; TCPDEBUG0; inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_close: inp == NULL")); INP_INFO_WLOCK(&V_tcbinfo); INP_WLOCK(inp); KASSERT(inp->inp_socket != NULL, ("tcp_usr_close: inp_socket == NULL")); /* * If we still have full TCP state, and we're not dropped, initiate * a disconnect. */ if (!(inp->inp_flags & INP_TIMEWAIT) && !(inp->inp_flags & INP_DROPPED)) { tp = intotcpcb(inp); TCPDEBUG1(); tcp_disconnect(tp); TCPDEBUG2(PRU_CLOSE); } if (!(inp->inp_flags & INP_DROPPED)) { SOCK_LOCK(so); so->so_state |= SS_PROTOREF; SOCK_UNLOCK(so); inp->inp_flags |= INP_SOCKREF; } INP_WUNLOCK(inp); INP_INFO_WUNLOCK(&V_tcbinfo); } /* * Receive out-of-band data. */ static int tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags) { int error = 0; struct inpcb *inp; struct tcpcb *tp = NULL; TCPDEBUG0; inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_rcvoob: inp == NULL")); INP_WLOCK(inp); if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { error = ECONNRESET; goto out; } tp = intotcpcb(inp); TCPDEBUG1(); if ((so->so_oobmark == 0 && (so->so_rcv.sb_state & SBS_RCVATMARK) == 0) || so->so_options & SO_OOBINLINE || tp->t_oobflags & TCPOOB_HADDATA) { error = EINVAL; goto out; } if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) { error = EWOULDBLOCK; goto out; } m->m_len = 1; *mtod(m, caddr_t) = tp->t_iobc; if ((flags & MSG_PEEK) == 0) tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA); out: TCPDEBUG2(PRU_RCVOOB); INP_WUNLOCK(inp); return (error); } #ifdef INET struct pr_usrreqs tcp_usrreqs = { .pru_abort = tcp_usr_abort, .pru_accept = tcp_usr_accept, .pru_attach = tcp_usr_attach, .pru_bind = tcp_usr_bind, .pru_connect = tcp_usr_connect, .pru_control = in_control, .pru_detach = tcp_usr_detach, .pru_disconnect = tcp_usr_disconnect, .pru_listen = tcp_usr_listen, .pru_peeraddr = in_getpeeraddr, .pru_rcvd = tcp_usr_rcvd, .pru_rcvoob = tcp_usr_rcvoob, .pru_send = tcp_usr_send, .pru_shutdown = tcp_usr_shutdown, .pru_sockaddr = in_getsockaddr, .pru_sosetlabel = in_pcbsosetlabel, .pru_close = tcp_usr_close, }; #endif /* INET */ #ifdef INET6 struct pr_usrreqs tcp6_usrreqs = { .pru_abort = tcp_usr_abort, .pru_accept = tcp6_usr_accept, .pru_attach = tcp_usr_attach, .pru_bind = tcp6_usr_bind, .pru_connect = tcp6_usr_connect, .pru_control = in6_control, .pru_detach = tcp_usr_detach, .pru_disconnect = tcp_usr_disconnect, .pru_listen = tcp6_usr_listen, .pru_peeraddr = in6_mapped_peeraddr, .pru_rcvd = tcp_usr_rcvd, .pru_rcvoob = tcp_usr_rcvoob, .pru_send = tcp_usr_send, .pru_shutdown = tcp_usr_shutdown, .pru_sockaddr = in6_mapped_sockaddr, .pru_sosetlabel = in_pcbsosetlabel, .pru_close = tcp_usr_close, }; #endif /* INET6 */ #ifdef INET /* * Common subroutine to open a TCP connection to remote host specified * by struct sockaddr_in in mbuf *nam. Call in_pcbbind to assign a local * port number if needed. Call in_pcbconnect_setup to do the routing and * to choose a local host address (interface). If there is an existing * incarnation of the same connection in TIME-WAIT state and if the remote * host was sending CC options and if the connection duration was < MSL, then * truncate the previous TIME-WAIT state and proceed. * Initialize connection parameters and enter SYN-SENT state. */ static int tcp_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td) { struct inpcb *inp = tp->t_inpcb, *oinp; struct socket *so = inp->inp_socket; struct in_addr laddr; u_short lport; int error; - INP_INFO_WLOCK_ASSERT(&V_tcbinfo); INP_WLOCK_ASSERT(inp); + INP_HASH_WLOCK(&V_tcbinfo); if (inp->inp_lport == 0) { error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred); if (error) - return error; + goto out; } /* * Cannot simply call in_pcbconnect, because there might be an * earlier incarnation of this same connection still in * TIME_WAIT state, creating an ADDRINUSE error. */ laddr = inp->inp_laddr; lport = inp->inp_lport; error = in_pcbconnect_setup(inp, nam, &laddr.s_addr, &lport, &inp->inp_faddr.s_addr, &inp->inp_fport, &oinp, td->td_ucred); if (error && oinp == NULL) - return error; - if (oinp) - return EADDRINUSE; + goto out; + if (oinp) { + error = EADDRINUSE; + goto out; + } inp->inp_laddr = laddr; in_pcbrehash(inp); + INP_HASH_WUNLOCK(&V_tcbinfo); /* * Compute window scaling to request: * Scale to fit into sweet spot. See tcp_syncache.c. * XXX: This should move to tcp_output(). */ while (tp->request_r_scale < TCP_MAX_WINSHIFT && (TCP_MAXWIN << tp->request_r_scale) < sb_max) tp->request_r_scale++; soisconnecting(so); TCPSTAT_INC(tcps_connattempt); tp->t_state = TCPS_SYN_SENT; tcp_timer_activate(tp, TT_KEEP, tcp_keepinit); tp->iss = tcp_new_isn(tp); tcp_sendseqinit(tp); return 0; + +out: + INP_HASH_WUNLOCK(&V_tcbinfo); + return (error); } #endif /* INET */ #ifdef INET6 static int tcp6_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td) { struct inpcb *inp = tp->t_inpcb, *oinp; struct socket *so = inp->inp_socket; struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam; struct in6_addr addr6; int error; - INP_INFO_WLOCK_ASSERT(&V_tcbinfo); INP_WLOCK_ASSERT(inp); + INP_HASH_WLOCK(&V_tcbinfo); if (inp->inp_lport == 0) { error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred); if (error) - return error; + goto out; } /* * Cannot simply call in_pcbconnect, because there might be an * earlier incarnation of this same connection still in * TIME_WAIT state, creating an ADDRINUSE error. * in6_pcbladdr() also handles scope zone IDs. + * + * XXXRW: We wouldn't need to expose in6_pcblookup_hash_locked() + * outside of in6_pcb.c if there were an in6_pcbconnect_setup(). */ error = in6_pcbladdr(inp, nam, &addr6); if (error) return error; - oinp = in6_pcblookup_hash(inp->inp_pcbinfo, + oinp = in6_pcblookup_hash_locked(inp->inp_pcbinfo, &sin6->sin6_addr, sin6->sin6_port, IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) ? &addr6 : &inp->in6p_laddr, inp->inp_lport, 0, NULL); - if (oinp) - return EADDRINUSE; + if (oinp) { + error = EADDRINUSE; + goto out; + } if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) inp->in6p_laddr = addr6; inp->in6p_faddr = sin6->sin6_addr; inp->inp_fport = sin6->sin6_port; /* update flowinfo - draft-itojun-ipv6-flowlabel-api-00 */ inp->inp_flow &= ~IPV6_FLOWLABEL_MASK; if (inp->inp_flags & IN6P_AUTOFLOWLABEL) inp->inp_flow |= (htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK); in_pcbrehash(inp); + INP_HASH_WUNLOCK(&V_tcbinfo); /* Compute window scaling to request. */ while (tp->request_r_scale < TCP_MAX_WINSHIFT && (TCP_MAXWIN << tp->request_r_scale) < sb_max) tp->request_r_scale++; soisconnecting(so); TCPSTAT_INC(tcps_connattempt); tp->t_state = TCPS_SYN_SENT; tcp_timer_activate(tp, TT_KEEP, tcp_keepinit); tp->iss = tcp_new_isn(tp); tcp_sendseqinit(tp); return 0; + +out: + INP_HASH_WUNLOCK(&V_tcbinfo); + return error; } #endif /* INET6 */ /* * Export TCP internal state information via a struct tcp_info, based on the * Linux 2.6 API. Not ABI compatible as our constants are mapped differently * (TCP state machine, etc). We export all information using FreeBSD-native * constants -- for example, the numeric values for tcpi_state will differ * from Linux. */ static void tcp_fill_info(struct tcpcb *tp, struct tcp_info *ti) { INP_WLOCK_ASSERT(tp->t_inpcb); bzero(ti, sizeof(*ti)); ti->tcpi_state = tp->t_state; if ((tp->t_flags & TF_REQ_TSTMP) && (tp->t_flags & TF_RCVD_TSTMP)) ti->tcpi_options |= TCPI_OPT_TIMESTAMPS; if (tp->t_flags & TF_SACK_PERMIT) ti->tcpi_options |= TCPI_OPT_SACK; if ((tp->t_flags & TF_REQ_SCALE) && (tp->t_flags & TF_RCVD_SCALE)) { ti->tcpi_options |= TCPI_OPT_WSCALE; ti->tcpi_snd_wscale = tp->snd_scale; ti->tcpi_rcv_wscale = tp->rcv_scale; } ti->tcpi_rto = tp->t_rxtcur * tick; ti->tcpi_last_data_recv = (long)(ticks - (int)tp->t_rcvtime) * tick; ti->tcpi_rtt = ((u_int64_t)tp->t_srtt * tick) >> TCP_RTT_SHIFT; ti->tcpi_rttvar = ((u_int64_t)tp->t_rttvar * tick) >> TCP_RTTVAR_SHIFT; ti->tcpi_snd_ssthresh = tp->snd_ssthresh; ti->tcpi_snd_cwnd = tp->snd_cwnd; /* * FreeBSD-specific extension fields for tcp_info. */ ti->tcpi_rcv_space = tp->rcv_wnd; ti->tcpi_rcv_nxt = tp->rcv_nxt; ti->tcpi_snd_wnd = tp->snd_wnd; ti->tcpi_snd_bwnd = 0; /* Unused, kept for compat. */ ti->tcpi_snd_nxt = tp->snd_nxt; ti->tcpi_snd_mss = tp->t_maxseg; ti->tcpi_rcv_mss = tp->t_maxseg; if (tp->t_flags & TF_TOE) ti->tcpi_options |= TCPI_OPT_TOE; ti->tcpi_snd_rexmitpack = tp->t_sndrexmitpack; ti->tcpi_rcv_ooopack = tp->t_rcvoopack; ti->tcpi_snd_zerowin = tp->t_sndzerowin; } /* * tcp_ctloutput() must drop the inpcb lock before performing copyin on * socket option arguments. When it re-acquires the lock after the copy, it * has to revalidate that the connection is still valid for the socket * option. */ #define INP_WLOCK_RECHECK(inp) do { \ INP_WLOCK(inp); \ if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { \ INP_WUNLOCK(inp); \ return (ECONNRESET); \ } \ tp = intotcpcb(inp); \ } while(0) int tcp_ctloutput(struct socket *so, struct sockopt *sopt) { int error, opt, optval; struct inpcb *inp; struct tcpcb *tp; struct tcp_info ti; char buf[TCP_CA_NAME_MAX]; struct cc_algo *algo; error = 0; inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_ctloutput: inp == NULL")); INP_WLOCK(inp); if (sopt->sopt_level != IPPROTO_TCP) { #ifdef INET6 if (inp->inp_vflag & INP_IPV6PROTO) { INP_WUNLOCK(inp); error = ip6_ctloutput(so, sopt); } #endif /* INET6 */ #if defined(INET6) && defined(INET) else #endif #ifdef INET { INP_WUNLOCK(inp); error = ip_ctloutput(so, sopt); } #endif return (error); } if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { INP_WUNLOCK(inp); return (ECONNRESET); } switch (sopt->sopt_dir) { case SOPT_SET: switch (sopt->sopt_name) { #ifdef TCP_SIGNATURE case TCP_MD5SIG: INP_WUNLOCK(inp); error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); if (error) return (error); INP_WLOCK_RECHECK(inp); if (optval > 0) tp->t_flags |= TF_SIGNATURE; else tp->t_flags &= ~TF_SIGNATURE; INP_WUNLOCK(inp); break; #endif /* TCP_SIGNATURE */ case TCP_NODELAY: case TCP_NOOPT: INP_WUNLOCK(inp); error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); if (error) return (error); INP_WLOCK_RECHECK(inp); switch (sopt->sopt_name) { case TCP_NODELAY: opt = TF_NODELAY; break; case TCP_NOOPT: opt = TF_NOOPT; break; default: opt = 0; /* dead code to fool gcc */ break; } if (optval) tp->t_flags |= opt; else tp->t_flags &= ~opt; INP_WUNLOCK(inp); break; case TCP_NOPUSH: INP_WUNLOCK(inp); error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); if (error) return (error); INP_WLOCK_RECHECK(inp); if (optval) tp->t_flags |= TF_NOPUSH; else if (tp->t_flags & TF_NOPUSH) { tp->t_flags &= ~TF_NOPUSH; if (TCPS_HAVEESTABLISHED(tp->t_state)) error = tcp_output(tp); } INP_WUNLOCK(inp); break; case TCP_MAXSEG: INP_WUNLOCK(inp); error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); if (error) return (error); INP_WLOCK_RECHECK(inp); if (optval > 0 && optval <= tp->t_maxseg && optval + 40 >= V_tcp_minmss) tp->t_maxseg = optval; else error = EINVAL; INP_WUNLOCK(inp); break; case TCP_INFO: INP_WUNLOCK(inp); error = EINVAL; break; case TCP_CONGESTION: INP_WUNLOCK(inp); bzero(buf, sizeof(buf)); error = sooptcopyin(sopt, &buf, sizeof(buf), 1); if (error) break; INP_WLOCK_RECHECK(inp); /* * Return EINVAL if we can't find the requested cc algo. */ error = EINVAL; CC_LIST_RLOCK(); STAILQ_FOREACH(algo, &cc_list, entries) { if (strncmp(buf, algo->name, TCP_CA_NAME_MAX) == 0) { /* We've found the requested algo. */ error = 0; /* * We hold a write lock over the tcb * so it's safe to do these things * without ordering concerns. */ if (CC_ALGO(tp)->cb_destroy != NULL) CC_ALGO(tp)->cb_destroy(tp->ccv); CC_ALGO(tp) = algo; /* * If something goes pear shaped * initialising the new algo, * fall back to newreno (which * does not require initialisation). */ if (algo->cb_init != NULL) if (algo->cb_init(tp->ccv) > 0) { CC_ALGO(tp) = &newreno_cc_algo; /* * The only reason init * should fail is * because of malloc. */ error = ENOMEM; } break; /* Break the STAILQ_FOREACH. */ } } CC_LIST_RUNLOCK(); INP_WUNLOCK(inp); break; default: INP_WUNLOCK(inp); error = ENOPROTOOPT; break; } break; case SOPT_GET: tp = intotcpcb(inp); switch (sopt->sopt_name) { #ifdef TCP_SIGNATURE case TCP_MD5SIG: optval = (tp->t_flags & TF_SIGNATURE) ? 1 : 0; INP_WUNLOCK(inp); error = sooptcopyout(sopt, &optval, sizeof optval); break; #endif case TCP_NODELAY: optval = tp->t_flags & TF_NODELAY; INP_WUNLOCK(inp); error = sooptcopyout(sopt, &optval, sizeof optval); break; case TCP_MAXSEG: optval = tp->t_maxseg; INP_WUNLOCK(inp); error = sooptcopyout(sopt, &optval, sizeof optval); break; case TCP_NOOPT: optval = tp->t_flags & TF_NOOPT; INP_WUNLOCK(inp); error = sooptcopyout(sopt, &optval, sizeof optval); break; case TCP_NOPUSH: optval = tp->t_flags & TF_NOPUSH; INP_WUNLOCK(inp); error = sooptcopyout(sopt, &optval, sizeof optval); break; case TCP_INFO: tcp_fill_info(tp, &ti); INP_WUNLOCK(inp); error = sooptcopyout(sopt, &ti, sizeof ti); break; case TCP_CONGESTION: bzero(buf, sizeof(buf)); strlcpy(buf, CC_ALGO(tp)->name, TCP_CA_NAME_MAX); INP_WUNLOCK(inp); error = sooptcopyout(sopt, buf, TCP_CA_NAME_MAX); break; default: INP_WUNLOCK(inp); error = ENOPROTOOPT; break; } break; } return (error); } #undef INP_WLOCK_RECHECK /* * tcp_sendspace and tcp_recvspace are the default send and receive window * sizes, respectively. These are obsolescent (this information should * be set by the route). */ u_long tcp_sendspace = 1024*32; SYSCTL_ULONG(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace, CTLFLAG_RW, &tcp_sendspace , 0, "Maximum outgoing TCP datagram size"); u_long tcp_recvspace = 1024*64; SYSCTL_ULONG(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace, CTLFLAG_RW, &tcp_recvspace , 0, "Maximum incoming TCP datagram size"); /* * Attach TCP protocol to socket, allocating * internet protocol control block, tcp control block, * bufer space, and entering LISTEN state if to accept connections. */ static int tcp_attach(struct socket *so) { struct tcpcb *tp; struct inpcb *inp; int error; if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { error = soreserve(so, tcp_sendspace, tcp_recvspace); if (error) return (error); } so->so_rcv.sb_flags |= SB_AUTOSIZE; so->so_snd.sb_flags |= SB_AUTOSIZE; INP_INFO_WLOCK(&V_tcbinfo); error = in_pcballoc(so, &V_tcbinfo); if (error) { INP_INFO_WUNLOCK(&V_tcbinfo); return (error); } inp = sotoinpcb(so); #ifdef INET6 if (inp->inp_vflag & INP_IPV6PROTO) { inp->inp_vflag |= INP_IPV6; inp->in6p_hops = -1; /* use kernel default */ } else #endif inp->inp_vflag |= INP_IPV4; tp = tcp_newtcpcb(inp); if (tp == NULL) { in_pcbdetach(inp); in_pcbfree(inp); INP_INFO_WUNLOCK(&V_tcbinfo); return (ENOBUFS); } tp->t_state = TCPS_CLOSED; INP_WUNLOCK(inp); INP_INFO_WUNLOCK(&V_tcbinfo); return (0); } /* * Initiate (or continue) disconnect. * If embryonic state, just send reset (once). * If in ``let data drain'' option and linger null, just drop. * Otherwise (hard), mark socket disconnecting and drop * current input data; switch states based on user close, and * send segment to peer (with FIN). */ static void tcp_disconnect(struct tcpcb *tp) { struct inpcb *inp = tp->t_inpcb; struct socket *so = inp->inp_socket; INP_INFO_WLOCK_ASSERT(&V_tcbinfo); INP_WLOCK_ASSERT(inp); /* * Neither tcp_close() nor tcp_drop() should return NULL, as the * socket is still open. */ if (tp->t_state < TCPS_ESTABLISHED) { tp = tcp_close(tp); KASSERT(tp != NULL, ("tcp_disconnect: tcp_close() returned NULL")); } else if ((so->so_options & SO_LINGER) && so->so_linger == 0) { tp = tcp_drop(tp, 0); KASSERT(tp != NULL, ("tcp_disconnect: tcp_drop() returned NULL")); } else { soisdisconnecting(so); sbflush(&so->so_rcv); tcp_usrclosed(tp); if (!(inp->inp_flags & INP_DROPPED)) tcp_output_disconnect(tp); } } /* * User issued close, and wish to trail through shutdown states: * if never received SYN, just forget it. If got a SYN from peer, * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN. * If already got a FIN from peer, then almost done; go to LAST_ACK * state. In all other cases, have already sent FIN to peer (e.g. * after PRU_SHUTDOWN), and just have to play tedious game waiting * for peer to send FIN or not respond to keep-alives, etc. * We can let the user exit from the close as soon as the FIN is acked. */ static void tcp_usrclosed(struct tcpcb *tp) { INP_INFO_WLOCK_ASSERT(&V_tcbinfo); INP_WLOCK_ASSERT(tp->t_inpcb); switch (tp->t_state) { case TCPS_LISTEN: tcp_offload_listen_close(tp); /* FALLTHROUGH */ case TCPS_CLOSED: tp->t_state = TCPS_CLOSED; tp = tcp_close(tp); /* * tcp_close() should never return NULL here as the socket is * still open. */ KASSERT(tp != NULL, ("tcp_usrclosed: tcp_close() returned NULL")); break; case TCPS_SYN_SENT: case TCPS_SYN_RECEIVED: tp->t_flags |= TF_NEEDFIN; break; case TCPS_ESTABLISHED: tp->t_state = TCPS_FIN_WAIT_1; break; case TCPS_CLOSE_WAIT: tp->t_state = TCPS_LAST_ACK; break; } if (tp->t_state >= TCPS_FIN_WAIT_2) { soisdisconnected(tp->t_inpcb->inp_socket); /* Prevent the connection hanging in FIN_WAIT_2 forever. */ if (tp->t_state == TCPS_FIN_WAIT_2) { int timeout; timeout = (tcp_fast_finwait2_recycle) ? tcp_finwait2_timeout : tcp_maxidle; tcp_timer_activate(tp, TT_2MSL, timeout); } } } #ifdef DDB static void db_print_indent(int indent) { int i; for (i = 0; i < indent; i++) db_printf(" "); } static void db_print_tstate(int t_state) { switch (t_state) { case TCPS_CLOSED: db_printf("TCPS_CLOSED"); return; case TCPS_LISTEN: db_printf("TCPS_LISTEN"); return; case TCPS_SYN_SENT: db_printf("TCPS_SYN_SENT"); return; case TCPS_SYN_RECEIVED: db_printf("TCPS_SYN_RECEIVED"); return; case TCPS_ESTABLISHED: db_printf("TCPS_ESTABLISHED"); return; case TCPS_CLOSE_WAIT: db_printf("TCPS_CLOSE_WAIT"); return; case TCPS_FIN_WAIT_1: db_printf("TCPS_FIN_WAIT_1"); return; case TCPS_CLOSING: db_printf("TCPS_CLOSING"); return; case TCPS_LAST_ACK: db_printf("TCPS_LAST_ACK"); return; case TCPS_FIN_WAIT_2: db_printf("TCPS_FIN_WAIT_2"); return; case TCPS_TIME_WAIT: db_printf("TCPS_TIME_WAIT"); return; default: db_printf("unknown"); return; } } static void db_print_tflags(u_int t_flags) { int comma; comma = 0; if (t_flags & TF_ACKNOW) { db_printf("%sTF_ACKNOW", comma ? ", " : ""); comma = 1; } if (t_flags & TF_DELACK) { db_printf("%sTF_DELACK", comma ? ", " : ""); comma = 1; } if (t_flags & TF_NODELAY) { db_printf("%sTF_NODELAY", comma ? ", " : ""); comma = 1; } if (t_flags & TF_NOOPT) { db_printf("%sTF_NOOPT", comma ? ", " : ""); comma = 1; } if (t_flags & TF_SENTFIN) { db_printf("%sTF_SENTFIN", comma ? ", " : ""); comma = 1; } if (t_flags & TF_REQ_SCALE) { db_printf("%sTF_REQ_SCALE", comma ? ", " : ""); comma = 1; } if (t_flags & TF_RCVD_SCALE) { db_printf("%sTF_RECVD_SCALE", comma ? ", " : ""); comma = 1; } if (t_flags & TF_REQ_TSTMP) { db_printf("%sTF_REQ_TSTMP", comma ? ", " : ""); comma = 1; } if (t_flags & TF_RCVD_TSTMP) { db_printf("%sTF_RCVD_TSTMP", comma ? ", " : ""); comma = 1; } if (t_flags & TF_SACK_PERMIT) { db_printf("%sTF_SACK_PERMIT", comma ? ", " : ""); comma = 1; } if (t_flags & TF_NEEDSYN) { db_printf("%sTF_NEEDSYN", comma ? ", " : ""); comma = 1; } if (t_flags & TF_NEEDFIN) { db_printf("%sTF_NEEDFIN", comma ? ", " : ""); comma = 1; } if (t_flags & TF_NOPUSH) { db_printf("%sTF_NOPUSH", comma ? ", " : ""); comma = 1; } if (t_flags & TF_MORETOCOME) { db_printf("%sTF_MORETOCOME", comma ? ", " : ""); comma = 1; } if (t_flags & TF_LQ_OVERFLOW) { db_printf("%sTF_LQ_OVERFLOW", comma ? ", " : ""); comma = 1; } if (t_flags & TF_LASTIDLE) { db_printf("%sTF_LASTIDLE", comma ? ", " : ""); comma = 1; } if (t_flags & TF_RXWIN0SENT) { db_printf("%sTF_RXWIN0SENT", comma ? ", " : ""); comma = 1; } if (t_flags & TF_FASTRECOVERY) { db_printf("%sTF_FASTRECOVERY", comma ? ", " : ""); comma = 1; } if (t_flags & TF_CONGRECOVERY) { db_printf("%sTF_CONGRECOVERY", comma ? ", " : ""); comma = 1; } if (t_flags & TF_WASFRECOVERY) { db_printf("%sTF_WASFRECOVERY", comma ? ", " : ""); comma = 1; } if (t_flags & TF_SIGNATURE) { db_printf("%sTF_SIGNATURE", comma ? ", " : ""); comma = 1; } if (t_flags & TF_FORCEDATA) { db_printf("%sTF_FORCEDATA", comma ? ", " : ""); comma = 1; } if (t_flags & TF_TSO) { db_printf("%sTF_TSO", comma ? ", " : ""); comma = 1; } if (t_flags & TF_ECN_PERMIT) { db_printf("%sTF_ECN_PERMIT", comma ? ", " : ""); comma = 1; } } static void db_print_toobflags(char t_oobflags) { int comma; comma = 0; if (t_oobflags & TCPOOB_HAVEDATA) { db_printf("%sTCPOOB_HAVEDATA", comma ? ", " : ""); comma = 1; } if (t_oobflags & TCPOOB_HADDATA) { db_printf("%sTCPOOB_HADDATA", comma ? ", " : ""); comma = 1; } } static void db_print_tcpcb(struct tcpcb *tp, const char *name, int indent) { db_print_indent(indent); db_printf("%s at %p\n", name, tp); indent += 2; db_print_indent(indent); db_printf("t_segq first: %p t_segqlen: %d t_dupacks: %d\n", LIST_FIRST(&tp->t_segq), tp->t_segqlen, tp->t_dupacks); db_print_indent(indent); db_printf("tt_rexmt: %p tt_persist: %p tt_keep: %p\n", &tp->t_timers->tt_rexmt, &tp->t_timers->tt_persist, &tp->t_timers->tt_keep); db_print_indent(indent); db_printf("tt_2msl: %p tt_delack: %p t_inpcb: %p\n", &tp->t_timers->tt_2msl, &tp->t_timers->tt_delack, tp->t_inpcb); db_print_indent(indent); db_printf("t_state: %d (", tp->t_state); db_print_tstate(tp->t_state); db_printf(")\n"); db_print_indent(indent); db_printf("t_flags: 0x%x (", tp->t_flags); db_print_tflags(tp->t_flags); db_printf(")\n"); db_print_indent(indent); db_printf("snd_una: 0x%08x snd_max: 0x%08x snd_nxt: x0%08x\n", tp->snd_una, tp->snd_max, tp->snd_nxt); db_print_indent(indent); db_printf("snd_up: 0x%08x snd_wl1: 0x%08x snd_wl2: 0x%08x\n", tp->snd_up, tp->snd_wl1, tp->snd_wl2); db_print_indent(indent); db_printf("iss: 0x%08x irs: 0x%08x rcv_nxt: 0x%08x\n", tp->iss, tp->irs, tp->rcv_nxt); db_print_indent(indent); db_printf("rcv_adv: 0x%08x rcv_wnd: %lu rcv_up: 0x%08x\n", tp->rcv_adv, tp->rcv_wnd, tp->rcv_up); db_print_indent(indent); db_printf("snd_wnd: %lu snd_cwnd: %lu\n", tp->snd_wnd, tp->snd_cwnd); db_print_indent(indent); db_printf("snd_ssthresh: %lu snd_recover: " "0x%08x\n", tp->snd_ssthresh, tp->snd_recover); db_print_indent(indent); db_printf("t_maxopd: %u t_rcvtime: %u t_startime: %u\n", tp->t_maxopd, tp->t_rcvtime, tp->t_starttime); db_print_indent(indent); db_printf("t_rttime: %u t_rtsq: 0x%08x\n", tp->t_rtttime, tp->t_rtseq); db_print_indent(indent); db_printf("t_rxtcur: %d t_maxseg: %u t_srtt: %d\n", tp->t_rxtcur, tp->t_maxseg, tp->t_srtt); db_print_indent(indent); db_printf("t_rttvar: %d t_rxtshift: %d t_rttmin: %u " "t_rttbest: %u\n", tp->t_rttvar, tp->t_rxtshift, tp->t_rttmin, tp->t_rttbest); db_print_indent(indent); db_printf("t_rttupdated: %lu max_sndwnd: %lu t_softerror: %d\n", tp->t_rttupdated, tp->max_sndwnd, tp->t_softerror); db_print_indent(indent); db_printf("t_oobflags: 0x%x (", tp->t_oobflags); db_print_toobflags(tp->t_oobflags); db_printf(") t_iobc: 0x%02x\n", tp->t_iobc); db_print_indent(indent); db_printf("snd_scale: %u rcv_scale: %u request_r_scale: %u\n", tp->snd_scale, tp->rcv_scale, tp->request_r_scale); db_print_indent(indent); db_printf("ts_recent: %u ts_recent_age: %u\n", tp->ts_recent, tp->ts_recent_age); db_print_indent(indent); db_printf("ts_offset: %u last_ack_sent: 0x%08x snd_cwnd_prev: " "%lu\n", tp->ts_offset, tp->last_ack_sent, tp->snd_cwnd_prev); db_print_indent(indent); db_printf("snd_ssthresh_prev: %lu snd_recover_prev: 0x%08x " "t_badrxtwin: %u\n", tp->snd_ssthresh_prev, tp->snd_recover_prev, tp->t_badrxtwin); db_print_indent(indent); db_printf("snd_numholes: %d snd_holes first: %p\n", tp->snd_numholes, TAILQ_FIRST(&tp->snd_holes)); db_print_indent(indent); db_printf("snd_fack: 0x%08x rcv_numsacks: %d sack_newdata: " "0x%08x\n", tp->snd_fack, tp->rcv_numsacks, tp->sack_newdata); /* Skip sackblks, sackhint. */ db_print_indent(indent); db_printf("t_rttlow: %d rfbuf_ts: %u rfbuf_cnt: %d\n", tp->t_rttlow, tp->rfbuf_ts, tp->rfbuf_cnt); } DB_SHOW_COMMAND(tcpcb, db_show_tcpcb) { struct tcpcb *tp; if (!have_addr) { db_printf("usage: show tcpcb \n"); return; } tp = (struct tcpcb *)addr; db_print_tcpcb(tp, "tcpcb", 0); } #endif Index: head/sys/netinet/udp_usrreq.c =================================================================== --- head/sys/netinet/udp_usrreq.c (revision 222487) +++ head/sys/netinet/udp_usrreq.c (revision 222488) @@ -1,1602 +1,1604 @@ /*- * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995 * The Regents of the University of California. * Copyright (c) 2008 Robert N. M. Watson + * Copyright (c) 2010-2011 Juniper Networks, Inc. * All rights reserved. * + * Portions of this software were developed by Robert N. M. Watson under + * contract to Juniper Networks, Inc. + * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)udp_usrreq.c 8.6 (Berkeley) 5/23/95 */ #include __FBSDID("$FreeBSD$"); #include "opt_ipfw.h" #include "opt_inet.h" #include "opt_inet6.h" #include "opt_ipsec.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef INET6 #include #endif #include #include #include #include #ifdef INET6 #include #endif #include #include #ifdef IPSEC #include #include #endif #include #include /* * UDP protocol implementation. * Per RFC 768, August, 1980. */ /* * BSD 4.2 defaulted the udp checksum to be off. Turning off udp checksums * removes the only data integrity mechanism for packets and malformed * packets that would otherwise be discarded due to bad checksums, and may * cause problems (especially for NFS data blocks). */ static int udp_cksum = 1; SYSCTL_INT(_net_inet_udp, UDPCTL_CHECKSUM, checksum, CTLFLAG_RW, &udp_cksum, 0, "compute udp checksum"); int udp_log_in_vain = 0; SYSCTL_INT(_net_inet_udp, OID_AUTO, log_in_vain, CTLFLAG_RW, &udp_log_in_vain, 0, "Log all incoming UDP packets"); VNET_DEFINE(int, udp_blackhole) = 0; SYSCTL_VNET_INT(_net_inet_udp, OID_AUTO, blackhole, CTLFLAG_RW, &VNET_NAME(udp_blackhole), 0, "Do not send port unreachables for refused connects"); u_long udp_sendspace = 9216; /* really max datagram size */ /* 40 1K datagrams */ SYSCTL_ULONG(_net_inet_udp, UDPCTL_MAXDGRAM, maxdgram, CTLFLAG_RW, &udp_sendspace, 0, "Maximum outgoing UDP datagram size"); u_long udp_recvspace = 40 * (1024 + #ifdef INET6 sizeof(struct sockaddr_in6) #else sizeof(struct sockaddr_in) #endif ); SYSCTL_ULONG(_net_inet_udp, UDPCTL_RECVSPACE, recvspace, CTLFLAG_RW, &udp_recvspace, 0, "Maximum space for incoming UDP datagrams"); VNET_DEFINE(struct inpcbhead, udb); /* from udp_var.h */ VNET_DEFINE(struct inpcbinfo, udbinfo); static VNET_DEFINE(uma_zone_t, udpcb_zone); #define V_udpcb_zone VNET(udpcb_zone) #ifndef UDBHASHSIZE #define UDBHASHSIZE 128 #endif VNET_DEFINE(struct udpstat, udpstat); /* from udp_var.h */ SYSCTL_VNET_STRUCT(_net_inet_udp, UDPCTL_STATS, stats, CTLFLAG_RW, &VNET_NAME(udpstat), udpstat, "UDP statistics (struct udpstat, netinet/udp_var.h)"); #ifdef INET static void udp_detach(struct socket *so); static int udp_output(struct inpcb *, struct mbuf *, struct sockaddr *, struct mbuf *, struct thread *); #endif #ifdef IPSEC #ifdef IPSEC_NAT_T #define UF_ESPINUDP_ALL (UF_ESPINUDP_NON_IKE|UF_ESPINUDP) #ifdef INET static struct mbuf *udp4_espdecap(struct inpcb *, struct mbuf *, int); #endif #endif /* IPSEC_NAT_T */ #endif /* IPSEC */ static void udp_zone_change(void *tag) { uma_zone_set_max(V_udbinfo.ipi_zone, maxsockets); uma_zone_set_max(V_udpcb_zone, maxsockets); } static int udp_inpcb_init(void *mem, int size, int flags) { struct inpcb *inp; inp = mem; INP_LOCK_INIT(inp, "inp", "udpinp"); return (0); } void udp_init(void) { in_pcbinfo_init(&V_udbinfo, "udp", &V_udb, UDBHASHSIZE, UDBHASHSIZE, "udp_inpcb", udp_inpcb_init, NULL, UMA_ZONE_NOFREE); V_udpcb_zone = uma_zcreate("udpcb", sizeof(struct udpcb), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); uma_zone_set_max(V_udpcb_zone, maxsockets); EVENTHANDLER_REGISTER(maxsockets_change, udp_zone_change, NULL, EVENTHANDLER_PRI_ANY); } /* * Kernel module interface for updating udpstat. The argument is an index * into udpstat treated as an array of u_long. While this encodes the * general layout of udpstat into the caller, it doesn't encode its location, * so that future changes to add, for example, per-CPU stats support won't * cause binary compatibility problems for kernel modules. */ void kmod_udpstat_inc(int statnum) { (*((u_long *)&V_udpstat + statnum))++; } int udp_newudpcb(struct inpcb *inp) { struct udpcb *up; up = uma_zalloc(V_udpcb_zone, M_NOWAIT | M_ZERO); if (up == NULL) return (ENOBUFS); inp->inp_ppcb = up; return (0); } void udp_discardcb(struct udpcb *up) { uma_zfree(V_udpcb_zone, up); } #ifdef VIMAGE void udp_destroy(void) { in_pcbinfo_destroy(&V_udbinfo); uma_zdestroy(V_udpcb_zone); } #endif #ifdef INET /* * Subroutine of udp_input(), which appends the provided mbuf chain to the * passed pcb/socket. The caller must provide a sockaddr_in via udp_in that * contains the source address. If the socket ends up being an IPv6 socket, * udp_append() will convert to a sockaddr_in6 before passing the address * into the socket code. */ static void udp_append(struct inpcb *inp, struct ip *ip, struct mbuf *n, int off, struct sockaddr_in *udp_in) { struct sockaddr *append_sa; struct socket *so; struct mbuf *opts = 0; #ifdef INET6 struct sockaddr_in6 udp_in6; #endif struct udpcb *up; - INP_RLOCK_ASSERT(inp); + INP_LOCK_ASSERT(inp); /* * Engage the tunneling protocol. */ up = intoudpcb(inp); if (up->u_tun_func != NULL) { (*up->u_tun_func)(n, off, inp); return; } if (n == NULL) return; off += sizeof(struct udphdr); #ifdef IPSEC /* Check AH/ESP integrity. */ if (ipsec4_in_reject(n, inp)) { m_freem(n); V_ipsec4stat.in_polvio++; return; } #ifdef IPSEC_NAT_T up = intoudpcb(inp); KASSERT(up != NULL, ("%s: udpcb NULL", __func__)); if (up->u_flags & UF_ESPINUDP_ALL) { /* IPSec UDP encaps. */ n = udp4_espdecap(inp, n, off); if (n == NULL) /* Consumed. */ return; } #endif /* IPSEC_NAT_T */ #endif /* IPSEC */ #ifdef MAC if (mac_inpcb_check_deliver(inp, n) != 0) { m_freem(n); return; } #endif /* MAC */ if (inp->inp_flags & INP_CONTROLOPTS || inp->inp_socket->so_options & (SO_TIMESTAMP | SO_BINTIME)) { #ifdef INET6 if (inp->inp_vflag & INP_IPV6) (void)ip6_savecontrol_v4(inp, n, &opts, NULL); else #endif /* INET6 */ ip_savecontrol(inp, &opts, ip, n); } #ifdef INET6 if (inp->inp_vflag & INP_IPV6) { bzero(&udp_in6, sizeof(udp_in6)); udp_in6.sin6_len = sizeof(udp_in6); udp_in6.sin6_family = AF_INET6; in6_sin_2_v4mapsin6(udp_in, &udp_in6); append_sa = (struct sockaddr *)&udp_in6; } else #endif /* INET6 */ append_sa = (struct sockaddr *)udp_in; m_adj(n, off); so = inp->inp_socket; SOCKBUF_LOCK(&so->so_rcv); if (sbappendaddr_locked(&so->so_rcv, append_sa, n, opts) == 0) { SOCKBUF_UNLOCK(&so->so_rcv); m_freem(n); if (opts) m_freem(opts); UDPSTAT_INC(udps_fullsock); } else sorwakeup_locked(so); } void udp_input(struct mbuf *m, int off) { int iphlen = off; struct ip *ip; struct udphdr *uh; struct ifnet *ifp; struct inpcb *inp; int len; struct ip save_ip; struct sockaddr_in udp_in; #ifdef IPFIREWALL_FORWARD struct m_tag *fwd_tag; #endif ifp = m->m_pkthdr.rcvif; UDPSTAT_INC(udps_ipackets); /* * Strip IP options, if any; should skip this, make available to * user, and use on returned packets, but we don't yet have a way to * check the checksum with options still present. */ if (iphlen > sizeof (struct ip)) { ip_stripoptions(m, (struct mbuf *)0); iphlen = sizeof(struct ip); } /* * Get IP and UDP header together in first mbuf. */ ip = mtod(m, struct ip *); if (m->m_len < iphlen + sizeof(struct udphdr)) { if ((m = m_pullup(m, iphlen + sizeof(struct udphdr))) == 0) { UDPSTAT_INC(udps_hdrops); return; } ip = mtod(m, struct ip *); } uh = (struct udphdr *)((caddr_t)ip + iphlen); /* * Destination port of 0 is illegal, based on RFC768. */ if (uh->uh_dport == 0) goto badunlocked; /* * Construct sockaddr format source address. Stuff source address * and datagram in user buffer. */ bzero(&udp_in, sizeof(udp_in)); udp_in.sin_len = sizeof(udp_in); udp_in.sin_family = AF_INET; udp_in.sin_port = uh->uh_sport; udp_in.sin_addr = ip->ip_src; /* * Make mbuf data length reflect UDP length. If not enough data to * reflect UDP length, drop. */ len = ntohs((u_short)uh->uh_ulen); if (ip->ip_len != len) { if (len > ip->ip_len || len < sizeof(struct udphdr)) { UDPSTAT_INC(udps_badlen); goto badunlocked; } m_adj(m, len - ip->ip_len); /* ip->ip_len = len; */ } /* * Save a copy of the IP header in case we want restore it for * sending an ICMP error message in response. */ if (!V_udp_blackhole) save_ip = *ip; else memset(&save_ip, 0, sizeof(save_ip)); /* * Checksum extended UDP header and data. */ if (uh->uh_sum) { u_short uh_sum; if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) { if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) uh_sum = m->m_pkthdr.csum_data; else uh_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr, htonl((u_short)len + m->m_pkthdr.csum_data + IPPROTO_UDP)); uh_sum ^= 0xffff; } else { char b[9]; bcopy(((struct ipovly *)ip)->ih_x1, b, 9); bzero(((struct ipovly *)ip)->ih_x1, 9); ((struct ipovly *)ip)->ih_len = uh->uh_ulen; uh_sum = in_cksum(m, len + sizeof (struct ip)); bcopy(b, ((struct ipovly *)ip)->ih_x1, 9); } if (uh_sum) { UDPSTAT_INC(udps_badsum); m_freem(m); return; } } else UDPSTAT_INC(udps_nosum); #ifdef IPFIREWALL_FORWARD /* * Grab info from PACKET_TAG_IPFORWARD tag prepended to the chain. */ fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL); if (fwd_tag != NULL) { struct sockaddr_in *next_hop; /* * Do the hack. */ next_hop = (struct sockaddr_in *)(fwd_tag + 1); ip->ip_dst = next_hop->sin_addr; uh->uh_dport = ntohs(next_hop->sin_port); /* * Remove the tag from the packet. We don't need it anymore. */ m_tag_delete(m, fwd_tag); } #endif - INP_INFO_RLOCK(&V_udbinfo); if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) || in_broadcast(ip->ip_dst, ifp)) { struct inpcb *last; struct ip_moptions *imo; + INP_INFO_RLOCK(&V_udbinfo); last = NULL; LIST_FOREACH(inp, &V_udb, inp_list) { if (inp->inp_lport != uh->uh_dport) continue; #ifdef INET6 if ((inp->inp_vflag & INP_IPV4) == 0) continue; #endif if (inp->inp_laddr.s_addr != INADDR_ANY && inp->inp_laddr.s_addr != ip->ip_dst.s_addr) continue; if (inp->inp_faddr.s_addr != INADDR_ANY && inp->inp_faddr.s_addr != ip->ip_src.s_addr) continue; if (inp->inp_fport != 0 && inp->inp_fport != uh->uh_sport) continue; INP_RLOCK(inp); /* + * XXXRW: Because we weren't holding either the inpcb + * or the hash lock when we checked for a match + * before, we should probably recheck now that the + * inpcb lock is held. + */ + + /* * Handle socket delivery policy for any-source * and source-specific multicast. [RFC3678] */ imo = inp->inp_moptions; if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { struct sockaddr_in group; int blocked; if (imo == NULL) { INP_RUNLOCK(inp); continue; } bzero(&group, sizeof(struct sockaddr_in)); group.sin_len = sizeof(struct sockaddr_in); group.sin_family = AF_INET; group.sin_addr = ip->ip_dst; blocked = imo_multi_filter(imo, ifp, (struct sockaddr *)&group, (struct sockaddr *)&udp_in); if (blocked != MCAST_PASS) { if (blocked == MCAST_NOTGMEMBER) IPSTAT_INC(ips_notmember); if (blocked == MCAST_NOTSMEMBER || blocked == MCAST_MUTED) UDPSTAT_INC(udps_filtermcast); INP_RUNLOCK(inp); continue; } } if (last != NULL) { struct mbuf *n; n = m_copy(m, 0, M_COPYALL); udp_append(last, ip, n, iphlen, &udp_in); INP_RUNLOCK(last); } last = inp; /* * Don't look for additional matches if this one does * not have either the SO_REUSEPORT or SO_REUSEADDR * socket options set. This heuristic avoids * searching through all pcbs in the common case of a * non-shared port. It assumes that an application * will never clear these options after setting them. */ if ((last->inp_socket->so_options & (SO_REUSEPORT|SO_REUSEADDR)) == 0) break; } if (last == NULL) { /* * No matching pcb found; discard datagram. (No need * to send an ICMP Port Unreachable for a broadcast * or multicast datgram.) */ UDPSTAT_INC(udps_noportbcast); - goto badheadlocked; + if (inp) + INP_RUNLOCK(inp); + INP_INFO_RUNLOCK(&V_udbinfo); + goto badunlocked; } udp_append(last, ip, m, iphlen, &udp_in); INP_RUNLOCK(last); INP_INFO_RUNLOCK(&V_udbinfo); return; } /* * Locate pcb for datagram. */ - inp = in_pcblookup_hash(&V_udbinfo, ip->ip_src, uh->uh_sport, - ip->ip_dst, uh->uh_dport, 1, ifp); + inp = in_pcblookup(&V_udbinfo, ip->ip_src, uh->uh_sport, + ip->ip_dst, uh->uh_dport, INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, + ifp); if (inp == NULL) { if (udp_log_in_vain) { char buf[4*sizeof "123"]; strcpy(buf, inet_ntoa(ip->ip_dst)); log(LOG_INFO, "Connection attempt to UDP %s:%d from %s:%d\n", buf, ntohs(uh->uh_dport), inet_ntoa(ip->ip_src), ntohs(uh->uh_sport)); } UDPSTAT_INC(udps_noport); if (m->m_flags & (M_BCAST | M_MCAST)) { UDPSTAT_INC(udps_noportbcast); - goto badheadlocked; + goto badunlocked; } if (V_udp_blackhole) - goto badheadlocked; + goto badunlocked; if (badport_bandlim(BANDLIM_ICMP_UNREACH) < 0) - goto badheadlocked; + goto badunlocked; *ip = save_ip; ip->ip_len += iphlen; icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0); - INP_INFO_RUNLOCK(&V_udbinfo); return; } /* * Check the minimum TTL for socket. */ - INP_RLOCK(inp); - INP_INFO_RUNLOCK(&V_udbinfo); + INP_RLOCK_ASSERT(inp); if (inp->inp_ip_minttl && inp->inp_ip_minttl > ip->ip_ttl) { INP_RUNLOCK(inp); - goto badunlocked; + m_freem(m); + return; } udp_append(inp, ip, m, iphlen, &udp_in); INP_RUNLOCK(inp); return; -badheadlocked: - if (inp) - INP_RUNLOCK(inp); - INP_INFO_RUNLOCK(&V_udbinfo); badunlocked: m_freem(m); } #endif /* INET */ /* * Notify a udp user of an asynchronous error; just wake up so that they can * collect error status. */ struct inpcb * udp_notify(struct inpcb *inp, int errno) { /* * While udp_ctlinput() always calls udp_notify() with a read lock * when invoking it directly, in_pcbnotifyall() currently uses write * locks due to sharing code with TCP. For now, accept either a read * or a write lock, but a read lock is sufficient. */ INP_LOCK_ASSERT(inp); inp->inp_socket->so_error = errno; sorwakeup(inp->inp_socket); sowwakeup(inp->inp_socket); return (inp); } #ifdef INET void udp_ctlinput(int cmd, struct sockaddr *sa, void *vip) { struct ip *ip = vip; struct udphdr *uh; struct in_addr faddr; struct inpcb *inp; faddr = ((struct sockaddr_in *)sa)->sin_addr; if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY) return; /* * Redirects don't need to be handled up here. */ if (PRC_IS_REDIRECT(cmd)) return; /* * Hostdead is ugly because it goes linearly through all PCBs. * * XXX: We never get this from ICMP, otherwise it makes an excellent * DoS attack on machines with many connections. */ if (cmd == PRC_HOSTDEAD) ip = NULL; else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0) return; if (ip != NULL) { uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2)); - INP_INFO_RLOCK(&V_udbinfo); - inp = in_pcblookup_hash(&V_udbinfo, faddr, uh->uh_dport, - ip->ip_src, uh->uh_sport, 0, NULL); + inp = in_pcblookup(&V_udbinfo, faddr, uh->uh_dport, + ip->ip_src, uh->uh_sport, INPLOOKUP_RLOCKPCB, NULL); if (inp != NULL) { - INP_RLOCK(inp); + INP_RLOCK_ASSERT(inp); if (inp->inp_socket != NULL) { udp_notify(inp, inetctlerrmap[cmd]); } INP_RUNLOCK(inp); } - INP_INFO_RUNLOCK(&V_udbinfo); } else in_pcbnotifyall(&V_udbinfo, faddr, inetctlerrmap[cmd], udp_notify); } #endif /* INET */ static int udp_pcblist(SYSCTL_HANDLER_ARGS) { int error, i, n; struct inpcb *inp, **inp_list; inp_gen_t gencnt; struct xinpgen xig; /* * The process of preparing the PCB list is too time-consuming and * resource-intensive to repeat twice on every request. */ if (req->oldptr == 0) { n = V_udbinfo.ipi_count; n += imax(n / 8, 10); req->oldidx = 2 * (sizeof xig) + n * sizeof(struct xinpcb); return (0); } if (req->newptr != 0) return (EPERM); /* * OK, now we're committed to doing something. */ INP_INFO_RLOCK(&V_udbinfo); gencnt = V_udbinfo.ipi_gencnt; n = V_udbinfo.ipi_count; INP_INFO_RUNLOCK(&V_udbinfo); 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); inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK); if (inp_list == 0) return (ENOMEM); INP_INFO_RLOCK(&V_udbinfo); for (inp = LIST_FIRST(V_udbinfo.ipi_listhead), i = 0; inp && i < n; inp = LIST_NEXT(inp, inp_list)) { INP_WLOCK(inp); if (inp->inp_gencnt <= gencnt && cr_canseeinpcb(req->td->td_ucred, inp) == 0) { in_pcbref(inp); inp_list[i++] = inp; } INP_WUNLOCK(inp); } INP_INFO_RUNLOCK(&V_udbinfo); n = i; error = 0; for (i = 0; i < n; i++) { inp = inp_list[i]; INP_RLOCK(inp); if (inp->inp_gencnt <= gencnt) { struct xinpcb xi; bzero(&xi, sizeof(xi)); xi.xi_len = sizeof xi; /* XXX should avoid extra copy */ bcopy(inp, &xi.xi_inp, sizeof *inp); if (inp->inp_socket) sotoxsocket(inp->inp_socket, &xi.xi_socket); xi.xi_inp.inp_gencnt = inp->inp_gencnt; INP_RUNLOCK(inp); error = SYSCTL_OUT(req, &xi, sizeof xi); } else INP_RUNLOCK(inp); } INP_INFO_WLOCK(&V_udbinfo); for (i = 0; i < n; i++) { inp = inp_list[i]; - INP_WLOCK(inp); - if (!in_pcbrele(inp)) - INP_WUNLOCK(inp); + INP_RLOCK(inp); + if (!in_pcbrele_rlocked(inp)) + INP_RUNLOCK(inp); } INP_INFO_WUNLOCK(&V_udbinfo); if (!error) { /* * Give the user an updated idea of our state. If the * generation differs from what we told her before, she knows * that something happened while we were processing this * request, and it might be necessary to retry. */ INP_INFO_RLOCK(&V_udbinfo); xig.xig_gen = V_udbinfo.ipi_gencnt; xig.xig_sogen = so_gencnt; xig.xig_count = V_udbinfo.ipi_count; INP_INFO_RUNLOCK(&V_udbinfo); error = SYSCTL_OUT(req, &xig, sizeof xig); } free(inp_list, M_TEMP); return (error); } SYSCTL_PROC(_net_inet_udp, UDPCTL_PCBLIST, pcblist, CTLTYPE_OPAQUE | CTLFLAG_RD, NULL, 0, udp_pcblist, "S,xinpcb", "List of active UDP sockets"); #ifdef INET static int udp_getcred(SYSCTL_HANDLER_ARGS) { struct xucred xuc; struct sockaddr_in addrs[2]; struct inpcb *inp; int error; error = priv_check(req->td, PRIV_NETINET_GETCRED); if (error) return (error); error = SYSCTL_IN(req, addrs, sizeof(addrs)); if (error) return (error); - INP_INFO_RLOCK(&V_udbinfo); - inp = in_pcblookup_hash(&V_udbinfo, addrs[1].sin_addr, addrs[1].sin_port, - addrs[0].sin_addr, addrs[0].sin_port, 1, NULL); + inp = in_pcblookup(&V_udbinfo, addrs[1].sin_addr, addrs[1].sin_port, + addrs[0].sin_addr, addrs[0].sin_port, + INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, NULL); if (inp != NULL) { - INP_RLOCK(inp); - INP_INFO_RUNLOCK(&V_udbinfo); + INP_RLOCK_ASSERT(inp); if (inp->inp_socket == NULL) error = ENOENT; if (error == 0) error = cr_canseeinpcb(req->td->td_ucred, inp); if (error == 0) cru2x(inp->inp_cred, &xuc); INP_RUNLOCK(inp); - } else { - INP_INFO_RUNLOCK(&V_udbinfo); + } else error = ENOENT; - } if (error == 0) error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred)); return (error); } SYSCTL_PROC(_net_inet_udp, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW|CTLFLAG_PRISON, 0, 0, udp_getcred, "S,xucred", "Get the xucred of a UDP connection"); #endif /* INET */ int udp_ctloutput(struct socket *so, struct sockopt *sopt) { int error = 0, optval; struct inpcb *inp; #ifdef IPSEC_NAT_T struct udpcb *up; #endif inp = sotoinpcb(so); KASSERT(inp != NULL, ("%s: inp == NULL", __func__)); INP_WLOCK(inp); if (sopt->sopt_level != IPPROTO_UDP) { #ifdef INET6 if (INP_CHECK_SOCKAF(so, AF_INET6)) { INP_WUNLOCK(inp); error = ip6_ctloutput(so, sopt); } #endif #if defined(INET) && defined(INET6) else #endif #ifdef INET { INP_WUNLOCK(inp); error = ip_ctloutput(so, sopt); } #endif return (error); } switch (sopt->sopt_dir) { case SOPT_SET: switch (sopt->sopt_name) { case UDP_ENCAP: INP_WUNLOCK(inp); error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); if (error) break; inp = sotoinpcb(so); KASSERT(inp != NULL, ("%s: inp == NULL", __func__)); INP_WLOCK(inp); #ifdef IPSEC_NAT_T up = intoudpcb(inp); KASSERT(up != NULL, ("%s: up == NULL", __func__)); #endif switch (optval) { case 0: /* Clear all UDP encap. */ #ifdef IPSEC_NAT_T up->u_flags &= ~UF_ESPINUDP_ALL; #endif break; #ifdef IPSEC_NAT_T case UDP_ENCAP_ESPINUDP: case UDP_ENCAP_ESPINUDP_NON_IKE: up->u_flags &= ~UF_ESPINUDP_ALL; if (optval == UDP_ENCAP_ESPINUDP) up->u_flags |= UF_ESPINUDP; else if (optval == UDP_ENCAP_ESPINUDP_NON_IKE) up->u_flags |= UF_ESPINUDP_NON_IKE; break; #endif default: error = EINVAL; break; } INP_WUNLOCK(inp); break; default: INP_WUNLOCK(inp); error = ENOPROTOOPT; break; } break; case SOPT_GET: switch (sopt->sopt_name) { #ifdef IPSEC_NAT_T case UDP_ENCAP: up = intoudpcb(inp); KASSERT(up != NULL, ("%s: up == NULL", __func__)); optval = up->u_flags & UF_ESPINUDP_ALL; INP_WUNLOCK(inp); error = sooptcopyout(sopt, &optval, sizeof optval); break; #endif default: INP_WUNLOCK(inp); error = ENOPROTOOPT; break; } break; } return (error); } #ifdef INET +#define UH_WLOCKED 2 +#define UH_RLOCKED 1 +#define UH_UNLOCKED 0 static int udp_output(struct inpcb *inp, struct mbuf *m, struct sockaddr *addr, struct mbuf *control, struct thread *td) { struct udpiphdr *ui; int len = m->m_pkthdr.len; struct in_addr faddr, laddr; struct cmsghdr *cm; struct sockaddr_in *sin, src; int error = 0; int ipflags; u_short fport, lport; int unlock_udbinfo; /* * udp_output() may need to temporarily bind or connect the current * inpcb. As such, we don't know up front whether we will need the * pcbinfo lock or not. Do any work to decide what is needed up * front before acquiring any locks. */ if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) { if (control) m_freem(control); m_freem(m); return (EMSGSIZE); } src.sin_family = 0; if (control != NULL) { /* * XXX: Currently, we assume all the optional information is * stored in a single mbuf. */ if (control->m_next) { m_freem(control); m_freem(m); return (EINVAL); } for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len), control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { cm = mtod(control, struct cmsghdr *); if (control->m_len < sizeof(*cm) || cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) { error = EINVAL; break; } if (cm->cmsg_level != IPPROTO_IP) continue; switch (cm->cmsg_type) { case IP_SENDSRCADDR: if (cm->cmsg_len != CMSG_LEN(sizeof(struct in_addr))) { error = EINVAL; break; } bzero(&src, sizeof(src)); src.sin_family = AF_INET; src.sin_len = sizeof(src); src.sin_port = inp->inp_lport; src.sin_addr = *(struct in_addr *)CMSG_DATA(cm); break; default: error = ENOPROTOOPT; break; } if (error) break; } m_freem(control); } if (error) { m_freem(m); return (error); } /* * Depending on whether or not the application has bound or connected * the socket, we may have to do varying levels of work. The optimal * case is for a connected UDP socket, as a global lock isn't * required at all. * * In order to decide which we need, we require stability of the * inpcb binding, which we ensure by acquiring a read lock on the * inpcb. This doesn't strictly follow the lock order, so we play * the trylock and retry game; note that we may end up with more * conservative locks than required the second time around, so later * assertions have to accept that. Further analysis of the number of * misses under contention is required. + * + * XXXRW: Check that hash locking update here is correct. */ sin = (struct sockaddr_in *)addr; INP_RLOCK(inp); if (sin != NULL && (inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0)) { INP_RUNLOCK(inp); - INP_INFO_WLOCK(&V_udbinfo); INP_WLOCK(inp); - unlock_udbinfo = 2; + INP_HASH_WLOCK(&V_udbinfo); + unlock_udbinfo = UH_WLOCKED; } else if ((sin != NULL && ( (sin->sin_addr.s_addr == INADDR_ANY) || (sin->sin_addr.s_addr == INADDR_BROADCAST) || (inp->inp_laddr.s_addr == INADDR_ANY) || (inp->inp_lport == 0))) || (src.sin_family == AF_INET)) { - if (!INP_INFO_TRY_RLOCK(&V_udbinfo)) { - INP_RUNLOCK(inp); - INP_INFO_RLOCK(&V_udbinfo); - INP_RLOCK(inp); - } - unlock_udbinfo = 1; + INP_HASH_RLOCK(&V_udbinfo); + unlock_udbinfo = UH_RLOCKED; } else - unlock_udbinfo = 0; + unlock_udbinfo = UH_UNLOCKED; /* * If the IP_SENDSRCADDR control message was specified, override the * source address for this datagram. Its use is invalidated if the * address thus specified is incomplete or clobbers other inpcbs. */ laddr = inp->inp_laddr; lport = inp->inp_lport; if (src.sin_family == AF_INET) { - INP_INFO_LOCK_ASSERT(&V_udbinfo); + INP_HASH_LOCK_ASSERT(&V_udbinfo); if ((lport == 0) || (laddr.s_addr == INADDR_ANY && src.sin_addr.s_addr == INADDR_ANY)) { error = EINVAL; goto release; } error = in_pcbbind_setup(inp, (struct sockaddr *)&src, &laddr.s_addr, &lport, td->td_ucred); if (error) goto release; } /* * If a UDP socket has been connected, then a local address/port will * have been selected and bound. * * If a UDP socket has not been connected to, then an explicit * destination address must be used, in which case a local * address/port may not have been selected and bound. */ if (sin != NULL) { INP_LOCK_ASSERT(inp); if (inp->inp_faddr.s_addr != INADDR_ANY) { error = EISCONN; goto release; } /* * Jail may rewrite the destination address, so let it do * that before we use it. */ error = prison_remote_ip4(td->td_ucred, &sin->sin_addr); if (error) goto release; /* * If a local address or port hasn't yet been selected, or if * the destination address needs to be rewritten due to using * a special INADDR_ constant, invoke in_pcbconnect_setup() * to do the heavy lifting. Once a port is selected, we * commit the binding back to the socket; we also commit the * binding of the address if in jail. * * If we already have a valid binding and we're not * requesting a destination address rewrite, use a fast path. */ if (inp->inp_laddr.s_addr == INADDR_ANY || inp->inp_lport == 0 || sin->sin_addr.s_addr == INADDR_ANY || sin->sin_addr.s_addr == INADDR_BROADCAST) { - INP_INFO_LOCK_ASSERT(&V_udbinfo); + INP_HASH_LOCK_ASSERT(&V_udbinfo); error = in_pcbconnect_setup(inp, addr, &laddr.s_addr, &lport, &faddr.s_addr, &fport, NULL, td->td_ucred); if (error) goto release; /* * XXXRW: Why not commit the port if the address is * !INADDR_ANY? */ /* Commit the local port if newly assigned. */ if (inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0) { - INP_INFO_WLOCK_ASSERT(&V_udbinfo); INP_WLOCK_ASSERT(inp); + INP_HASH_WLOCK_ASSERT(&V_udbinfo); /* * Remember addr if jailed, to prevent * rebinding. */ if (prison_flag(td->td_ucred, PR_IP4)) inp->inp_laddr = laddr; inp->inp_lport = lport; if (in_pcbinshash(inp) != 0) { inp->inp_lport = 0; error = EAGAIN; goto release; } inp->inp_flags |= INP_ANONPORT; } } else { faddr = sin->sin_addr; fport = sin->sin_port; } } else { INP_LOCK_ASSERT(inp); faddr = inp->inp_faddr; fport = inp->inp_fport; if (faddr.s_addr == INADDR_ANY) { error = ENOTCONN; goto release; } } /* * Calculate data length and get a mbuf for UDP, IP, and possible * link-layer headers. Immediate slide the data pointer back forward * since we won't use that space at this layer. */ M_PREPEND(m, sizeof(struct udpiphdr) + max_linkhdr, M_DONTWAIT); if (m == NULL) { error = ENOBUFS; goto release; } m->m_data += max_linkhdr; m->m_len -= max_linkhdr; m->m_pkthdr.len -= max_linkhdr; /* * Fill in mbuf with extended UDP header and addresses and length put * into network format. */ ui = mtod(m, struct udpiphdr *); bzero(ui->ui_x1, sizeof(ui->ui_x1)); /* XXX still needed? */ ui->ui_pr = IPPROTO_UDP; ui->ui_src = laddr; ui->ui_dst = faddr; ui->ui_sport = lport; ui->ui_dport = fport; ui->ui_ulen = htons((u_short)len + sizeof(struct udphdr)); /* * Set the Don't Fragment bit in the IP header. */ if (inp->inp_flags & INP_DONTFRAG) { struct ip *ip; ip = (struct ip *)&ui->ui_i; ip->ip_off |= IP_DF; } ipflags = 0; if (inp->inp_socket->so_options & SO_DONTROUTE) ipflags |= IP_ROUTETOIF; if (inp->inp_socket->so_options & SO_BROADCAST) ipflags |= IP_ALLOWBROADCAST; if (inp->inp_flags & INP_ONESBCAST) ipflags |= IP_SENDONES; #ifdef MAC mac_inpcb_create_mbuf(inp, m); #endif /* * Set up checksum and output datagram. */ if (udp_cksum) { if (inp->inp_flags & INP_ONESBCAST) faddr.s_addr = INADDR_BROADCAST; ui->ui_sum = in_pseudo(ui->ui_src.s_addr, faddr.s_addr, htons((u_short)len + sizeof(struct udphdr) + IPPROTO_UDP)); m->m_pkthdr.csum_flags = CSUM_UDP; m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum); } else ui->ui_sum = 0; ((struct ip *)ui)->ip_len = sizeof (struct udpiphdr) + len; ((struct ip *)ui)->ip_ttl = inp->inp_ip_ttl; /* XXX */ ((struct ip *)ui)->ip_tos = inp->inp_ip_tos; /* XXX */ UDPSTAT_INC(udps_opackets); - if (unlock_udbinfo == 2) - INP_INFO_WUNLOCK(&V_udbinfo); - else if (unlock_udbinfo == 1) - INP_INFO_RUNLOCK(&V_udbinfo); + if (unlock_udbinfo == UH_WLOCKED) + INP_HASH_WUNLOCK(&V_udbinfo); + else if (unlock_udbinfo == UH_RLOCKED) + INP_HASH_RUNLOCK(&V_udbinfo); error = ip_output(m, inp->inp_options, NULL, ipflags, inp->inp_moptions, inp); - if (unlock_udbinfo == 2) + if (unlock_udbinfo == UH_WLOCKED) INP_WUNLOCK(inp); else INP_RUNLOCK(inp); return (error); release: - if (unlock_udbinfo == 2) { + if (unlock_udbinfo == UH_WLOCKED) { + INP_HASH_WUNLOCK(&V_udbinfo); INP_WUNLOCK(inp); - INP_INFO_WUNLOCK(&V_udbinfo); - } else if (unlock_udbinfo == 1) { + } else if (unlock_udbinfo == UH_RLOCKED) { + INP_HASH_RUNLOCK(&V_udbinfo); INP_RUNLOCK(inp); - INP_INFO_RUNLOCK(&V_udbinfo); } else INP_RUNLOCK(inp); m_freem(m); return (error); } #if defined(IPSEC) && defined(IPSEC_NAT_T) /* * Potentially decap ESP in UDP frame. Check for an ESP header * and optional marker; if present, strip the UDP header and * push the result through IPSec. * * Returns mbuf to be processed (potentially re-allocated) or * NULL if consumed and/or processed. */ static struct mbuf * udp4_espdecap(struct inpcb *inp, struct mbuf *m, int off) { size_t minlen, payload, skip, iphlen; caddr_t data; struct udpcb *up; struct m_tag *tag; struct udphdr *udphdr; struct ip *ip; INP_RLOCK_ASSERT(inp); /* * Pull up data so the longest case is contiguous: * IP/UDP hdr + non ESP marker + ESP hdr. */ minlen = off + sizeof(uint64_t) + sizeof(struct esp); if (minlen > m->m_pkthdr.len) minlen = m->m_pkthdr.len; if ((m = m_pullup(m, minlen)) == NULL) { V_ipsec4stat.in_inval++; return (NULL); /* Bypass caller processing. */ } data = mtod(m, caddr_t); /* Points to ip header. */ payload = m->m_len - off; /* Size of payload. */ if (payload == 1 && data[off] == '\xff') return (m); /* NB: keepalive packet, no decap. */ up = intoudpcb(inp); KASSERT(up != NULL, ("%s: udpcb NULL", __func__)); KASSERT((up->u_flags & UF_ESPINUDP_ALL) != 0, ("u_flags 0x%x", up->u_flags)); /* * Check that the payload is large enough to hold an * ESP header and compute the amount of data to remove. * * NB: the caller has already done a pullup for us. * XXX can we assume alignment and eliminate bcopys? */ if (up->u_flags & UF_ESPINUDP_NON_IKE) { /* * draft-ietf-ipsec-nat-t-ike-0[01].txt and * draft-ietf-ipsec-udp-encaps-(00/)01.txt, ignoring * possible AH mode non-IKE marker+non-ESP marker * from draft-ietf-ipsec-udp-encaps-00.txt. */ uint64_t marker; if (payload <= sizeof(uint64_t) + sizeof(struct esp)) return (m); /* NB: no decap. */ bcopy(data + off, &marker, sizeof(uint64_t)); if (marker != 0) /* Non-IKE marker. */ return (m); /* NB: no decap. */ skip = sizeof(uint64_t) + sizeof(struct udphdr); } else { uint32_t spi; if (payload <= sizeof(struct esp)) { V_ipsec4stat.in_inval++; m_freem(m); return (NULL); /* Discard. */ } bcopy(data + off, &spi, sizeof(uint32_t)); if (spi == 0) /* Non-ESP marker. */ return (m); /* NB: no decap. */ skip = sizeof(struct udphdr); } /* * Setup a PACKET_TAG_IPSEC_NAT_T_PORT tag to remember * the UDP ports. This is required if we want to select * the right SPD for multiple hosts behind same NAT. * * NB: ports are maintained in network byte order everywhere * in the NAT-T code. */ tag = m_tag_get(PACKET_TAG_IPSEC_NAT_T_PORTS, 2 * sizeof(uint16_t), M_NOWAIT); if (tag == NULL) { V_ipsec4stat.in_nomem++; m_freem(m); return (NULL); /* Discard. */ } iphlen = off - sizeof(struct udphdr); udphdr = (struct udphdr *)(data + iphlen); ((uint16_t *)(tag + 1))[0] = udphdr->uh_sport; ((uint16_t *)(tag + 1))[1] = udphdr->uh_dport; m_tag_prepend(m, tag); /* * Remove the UDP header (and possibly the non ESP marker) * IP header length is iphlen * Before: * <--- off ---> * +----+------+-----+ * | IP | UDP | ESP | * +----+------+-----+ * <-skip-> * After: * +----+-----+ * | IP | ESP | * +----+-----+ * <-skip-> */ ovbcopy(data, data + skip, iphlen); m_adj(m, skip); ip = mtod(m, struct ip *); ip->ip_len -= skip; ip->ip_p = IPPROTO_ESP; /* * We cannot yet update the cksums so clear any * h/w cksum flags as they are no longer valid. */ if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) m->m_pkthdr.csum_flags &= ~(CSUM_DATA_VALID|CSUM_PSEUDO_HDR); (void) ipsec4_common_input(m, iphlen, ip->ip_p); return (NULL); /* NB: consumed, bypass processing. */ } #endif /* defined(IPSEC) && defined(IPSEC_NAT_T) */ static void udp_abort(struct socket *so) { struct inpcb *inp; inp = sotoinpcb(so); KASSERT(inp != NULL, ("udp_abort: inp == NULL")); - INP_INFO_WLOCK(&V_udbinfo); INP_WLOCK(inp); if (inp->inp_faddr.s_addr != INADDR_ANY) { + INP_HASH_WLOCK(&V_udbinfo); in_pcbdisconnect(inp); inp->inp_laddr.s_addr = INADDR_ANY; + INP_HASH_WUNLOCK(&V_udbinfo); soisdisconnected(so); } INP_WUNLOCK(inp); - INP_INFO_WUNLOCK(&V_udbinfo); } static int udp_attach(struct socket *so, int proto, struct thread *td) { struct inpcb *inp; int error; inp = sotoinpcb(so); KASSERT(inp == NULL, ("udp_attach: inp != NULL")); error = soreserve(so, udp_sendspace, udp_recvspace); if (error) return (error); INP_INFO_WLOCK(&V_udbinfo); error = in_pcballoc(so, &V_udbinfo); if (error) { INP_INFO_WUNLOCK(&V_udbinfo); return (error); } inp = sotoinpcb(so); inp->inp_vflag |= INP_IPV4; inp->inp_ip_ttl = V_ip_defttl; error = udp_newudpcb(inp); if (error) { in_pcbdetach(inp); in_pcbfree(inp); INP_INFO_WUNLOCK(&V_udbinfo); return (error); } INP_WUNLOCK(inp); INP_INFO_WUNLOCK(&V_udbinfo); return (0); } #endif /* INET */ int udp_set_kernel_tunneling(struct socket *so, udp_tun_func_t f) { struct inpcb *inp; struct udpcb *up; KASSERT(so->so_type == SOCK_DGRAM, ("udp_set_kernel_tunneling: !dgram")); inp = sotoinpcb(so); KASSERT(inp != NULL, ("udp_set_kernel_tunneling: inp == NULL")); INP_WLOCK(inp); up = intoudpcb(inp); if (up->u_tun_func != NULL) { INP_WUNLOCK(inp); return (EBUSY); } up->u_tun_func = f; INP_WUNLOCK(inp); return (0); } #ifdef INET static int udp_bind(struct socket *so, struct sockaddr *nam, struct thread *td) { struct inpcb *inp; int error; inp = sotoinpcb(so); KASSERT(inp != NULL, ("udp_bind: inp == NULL")); - INP_INFO_WLOCK(&V_udbinfo); INP_WLOCK(inp); + INP_HASH_WLOCK(&V_udbinfo); error = in_pcbbind(inp, nam, td->td_ucred); + INP_HASH_WUNLOCK(&V_udbinfo); INP_WUNLOCK(inp); - INP_INFO_WUNLOCK(&V_udbinfo); return (error); } static void udp_close(struct socket *so) { struct inpcb *inp; inp = sotoinpcb(so); KASSERT(inp != NULL, ("udp_close: inp == NULL")); - INP_INFO_WLOCK(&V_udbinfo); INP_WLOCK(inp); if (inp->inp_faddr.s_addr != INADDR_ANY) { + INP_HASH_WLOCK(&V_udbinfo); in_pcbdisconnect(inp); inp->inp_laddr.s_addr = INADDR_ANY; + INP_HASH_WUNLOCK(&V_udbinfo); soisdisconnected(so); } INP_WUNLOCK(inp); - INP_INFO_WUNLOCK(&V_udbinfo); } static int udp_connect(struct socket *so, struct sockaddr *nam, struct thread *td) { struct inpcb *inp; int error; struct sockaddr_in *sin; inp = sotoinpcb(so); KASSERT(inp != NULL, ("udp_connect: inp == NULL")); - INP_INFO_WLOCK(&V_udbinfo); INP_WLOCK(inp); if (inp->inp_faddr.s_addr != INADDR_ANY) { INP_WUNLOCK(inp); - INP_INFO_WUNLOCK(&V_udbinfo); return (EISCONN); } sin = (struct sockaddr_in *)nam; error = prison_remote_ip4(td->td_ucred, &sin->sin_addr); if (error != 0) { INP_WUNLOCK(inp); - INP_INFO_WUNLOCK(&V_udbinfo); return (error); } + INP_HASH_WLOCK(&V_udbinfo); error = in_pcbconnect(inp, nam, td->td_ucred); + INP_HASH_WUNLOCK(&V_udbinfo); if (error == 0) soisconnected(so); INP_WUNLOCK(inp); - INP_INFO_WUNLOCK(&V_udbinfo); return (error); } static void udp_detach(struct socket *so) { struct inpcb *inp; struct udpcb *up; inp = sotoinpcb(so); KASSERT(inp != NULL, ("udp_detach: inp == NULL")); KASSERT(inp->inp_faddr.s_addr == INADDR_ANY, ("udp_detach: not disconnected")); INP_INFO_WLOCK(&V_udbinfo); INP_WLOCK(inp); up = intoudpcb(inp); KASSERT(up != NULL, ("%s: up == NULL", __func__)); inp->inp_ppcb = NULL; in_pcbdetach(inp); in_pcbfree(inp); INP_INFO_WUNLOCK(&V_udbinfo); udp_discardcb(up); } static int udp_disconnect(struct socket *so) { struct inpcb *inp; inp = sotoinpcb(so); KASSERT(inp != NULL, ("udp_disconnect: inp == NULL")); - INP_INFO_WLOCK(&V_udbinfo); INP_WLOCK(inp); if (inp->inp_faddr.s_addr == INADDR_ANY) { INP_WUNLOCK(inp); - INP_INFO_WUNLOCK(&V_udbinfo); return (ENOTCONN); } - + INP_HASH_WLOCK(&V_udbinfo); in_pcbdisconnect(inp); inp->inp_laddr.s_addr = INADDR_ANY; + INP_HASH_WUNLOCK(&V_udbinfo); SOCK_LOCK(so); so->so_state &= ~SS_ISCONNECTED; /* XXX */ SOCK_UNLOCK(so); INP_WUNLOCK(inp); - INP_INFO_WUNLOCK(&V_udbinfo); return (0); } static int udp_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr, struct mbuf *control, struct thread *td) { struct inpcb *inp; inp = sotoinpcb(so); KASSERT(inp != NULL, ("udp_send: inp == NULL")); return (udp_output(inp, m, addr, control, td)); } #endif /* INET */ int udp_shutdown(struct socket *so) { struct inpcb *inp; inp = sotoinpcb(so); KASSERT(inp != NULL, ("udp_shutdown: inp == NULL")); INP_WLOCK(inp); socantsendmore(so); INP_WUNLOCK(inp); return (0); } #ifdef INET struct pr_usrreqs udp_usrreqs = { .pru_abort = udp_abort, .pru_attach = udp_attach, .pru_bind = udp_bind, .pru_connect = udp_connect, .pru_control = in_control, .pru_detach = udp_detach, .pru_disconnect = udp_disconnect, .pru_peeraddr = in_getpeeraddr, .pru_send = udp_send, .pru_soreceive = soreceive_dgram, .pru_sosend = sosend_dgram, .pru_shutdown = udp_shutdown, .pru_sockaddr = in_getsockaddr, .pru_sosetlabel = in_pcbsosetlabel, .pru_close = udp_close, }; #endif /* INET */ Index: head/sys/netinet6/in6_pcb.c =================================================================== --- head/sys/netinet6/in6_pcb.c (revision 222487) +++ head/sys/netinet6/in6_pcb.c (revision 222488) @@ -1,951 +1,1007 @@ /*- * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. + * Copyright (c) 2010-2011 Juniper Networks, Inc. * All rights reserved. * + * Portions of this software were developed by Robert N. M. Watson under + * contract to Juniper Networks, Inc. + * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. Neither the name of the project nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * $KAME: in6_pcb.c,v 1.31 2001/05/21 05:45:10 jinmei Exp $ */ /*- * Copyright (c) 1982, 1986, 1991, 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. * * @(#)in_pcb.c 8.2 (Berkeley) 1/4/94 */ #include __FBSDID("$FreeBSD$"); #include "opt_inet.h" #include "opt_inet6.h" #include "opt_ipsec.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include struct in6_addr zeroin6_addr; int in6_pcbbind(register struct inpcb *inp, struct sockaddr *nam, struct ucred *cred) { struct socket *so = inp->inp_socket; struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)NULL; struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; u_short lport = 0; int error, lookupflags = 0; int reuseport = (so->so_options & SO_REUSEPORT); - INP_INFO_WLOCK_ASSERT(pcbinfo); INP_WLOCK_ASSERT(inp); + INP_HASH_WLOCK_ASSERT(pcbinfo); if (TAILQ_EMPTY(&V_in6_ifaddrhead)) /* XXX broken! */ return (EADDRNOTAVAIL); if (inp->inp_lport || !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) return (EINVAL); if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0) lookupflags = INPLOOKUP_WILDCARD; if (nam == NULL) { if ((error = prison_local_ip6(cred, &inp->in6p_laddr, ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0) return (error); } else { sin6 = (struct sockaddr_in6 *)nam; if (nam->sa_len != sizeof(*sin6)) return (EINVAL); /* * family check. */ if (nam->sa_family != AF_INET6) return (EAFNOSUPPORT); if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0) return(error); if ((error = prison_local_ip6(cred, &sin6->sin6_addr, ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0) return (error); lport = sin6->sin6_port; if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) { /* * Treat SO_REUSEADDR as SO_REUSEPORT for multicast; * allow compepte duplication of binding if * SO_REUSEPORT is set, or if SO_REUSEADDR is set * and a multicast address is bound on both * new and duplicated sockets. */ if (so->so_options & SO_REUSEADDR) reuseport = SO_REUSEADDR|SO_REUSEPORT; } else if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { struct ifaddr *ifa; sin6->sin6_port = 0; /* yech... */ if ((ifa = ifa_ifwithaddr((struct sockaddr *)sin6)) == NULL && (inp->inp_flags & INP_BINDANY) == 0) { return (EADDRNOTAVAIL); } /* * XXX: bind to an anycast address might accidentally * cause sending a packet with anycast source address. * We should allow to bind to a deprecated address, since * the application dares to use it. */ if (ifa != NULL && ((struct in6_ifaddr *)ifa)->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY|IN6_IFF_DETACHED)) { ifa_free(ifa); return (EADDRNOTAVAIL); } if (ifa != NULL) ifa_free(ifa); } if (lport) { struct inpcb *t; /* GROSS */ if (ntohs(lport) <= V_ipport_reservedhigh && ntohs(lport) >= V_ipport_reservedlow && priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT, 0)) return (EACCES); if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr) && priv_check_cred(inp->inp_cred, PRIV_NETINET_REUSEPORT, 0) != 0) { t = in6_pcblookup_local(pcbinfo, &sin6->sin6_addr, lport, INPLOOKUP_WILDCARD, cred); if (t && ((t->inp_flags & INP_TIMEWAIT) == 0) && (so->so_type != SOCK_STREAM || IN6_IS_ADDR_UNSPECIFIED(&t->in6p_faddr)) && (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) || !IN6_IS_ADDR_UNSPECIFIED(&t->in6p_laddr) || (t->inp_socket->so_options & SO_REUSEPORT) == 0) && (inp->inp_cred->cr_uid != t->inp_cred->cr_uid)) return (EADDRINUSE); #ifdef INET if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0 && IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { struct sockaddr_in sin; in6_sin6_2_sin(&sin, sin6); t = in_pcblookup_local(pcbinfo, sin.sin_addr, lport, INPLOOKUP_WILDCARD, cred); if (t && ((t->inp_flags & INP_TIMEWAIT) == 0) && (so->so_type != SOCK_STREAM || ntohl(t->inp_faddr.s_addr) == INADDR_ANY) && (inp->inp_cred->cr_uid != t->inp_cred->cr_uid)) return (EADDRINUSE); } #endif } t = in6_pcblookup_local(pcbinfo, &sin6->sin6_addr, lport, lookupflags, cred); if (t && (reuseport & ((t->inp_flags & INP_TIMEWAIT) ? intotw(t)->tw_so_options : t->inp_socket->so_options)) == 0) return (EADDRINUSE); #ifdef INET if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0 && IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { struct sockaddr_in sin; in6_sin6_2_sin(&sin, sin6); t = in_pcblookup_local(pcbinfo, sin.sin_addr, lport, lookupflags, cred); if (t && t->inp_flags & INP_TIMEWAIT) { if ((reuseport & intotw(t)->tw_so_options) == 0 && (ntohl(t->inp_laddr.s_addr) != INADDR_ANY || ((inp->inp_vflag & INP_IPV6PROTO) == (t->inp_vflag & INP_IPV6PROTO)))) return (EADDRINUSE); } else if (t && (reuseport & t->inp_socket->so_options) == 0 && (ntohl(t->inp_laddr.s_addr) != INADDR_ANY || INP_SOCKAF(so) == INP_SOCKAF(t->inp_socket))) return (EADDRINUSE); } #endif } inp->in6p_laddr = sin6->sin6_addr; } if (lport == 0) { if ((error = in6_pcbsetport(&inp->in6p_laddr, inp, cred)) != 0) { /* Undo an address bind that may have occurred. */ inp->in6p_laddr = in6addr_any; return (error); } } else { inp->inp_lport = lport; if (in_pcbinshash(inp) != 0) { inp->in6p_laddr = in6addr_any; inp->inp_lport = 0; return (EAGAIN); } } return (0); } /* * Transform old in6_pcbconnect() into an inner subroutine for new * in6_pcbconnect(): Do some validity-checking on the remote * address (in mbuf 'nam') and then determine local host address * (i.e., which interface) to use to access that remote host. * * This preserves definition of in6_pcbconnect(), while supporting a * slightly different version for T/TCP. (This is more than * a bit of a kludge, but cleaning up the internal interfaces would * have forced minor changes in every protocol). */ int in6_pcbladdr(register struct inpcb *inp, struct sockaddr *nam, struct in6_addr *plocal_addr6) { register struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam; int error = 0; struct ifnet *ifp = NULL; int scope_ambiguous = 0; struct in6_addr in6a; - INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo); INP_WLOCK_ASSERT(inp); + INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo); /* XXXRW: why? */ if (nam->sa_len != sizeof (*sin6)) return (EINVAL); if (sin6->sin6_family != AF_INET6) return (EAFNOSUPPORT); if (sin6->sin6_port == 0) return (EADDRNOTAVAIL); if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone) scope_ambiguous = 1; if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0) return(error); if (!TAILQ_EMPTY(&V_in6_ifaddrhead)) { /* * If the destination address is UNSPECIFIED addr, * use the loopback addr, e.g ::1. */ if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) sin6->sin6_addr = in6addr_loopback; } if ((error = prison_remote_ip6(inp->inp_cred, &sin6->sin6_addr)) != 0) return (error); error = in6_selectsrc(sin6, inp->in6p_outputopts, inp, NULL, inp->inp_cred, &ifp, &in6a); if (error) return (error); if (ifp && scope_ambiguous && (error = in6_setscope(&sin6->sin6_addr, ifp, NULL)) != 0) { return(error); } /* * Do not update this earlier, in case we return with an error. * * XXX: this in6_selectsrc result might replace the bound local * address with the address specified by setsockopt(IPV6_PKTINFO). * Is it the intended behavior? */ *plocal_addr6 = in6a; /* * Don't do pcblookup call here; return interface in * plocal_addr6 * and exit to caller, that will do the lookup. */ return (0); } /* * Outer subroutine: * Connect from a socket to a specified address. * Both address and port must be specified in argument sin. * If don't have a local address for this socket yet, * then pick one. */ int in6_pcbconnect(register struct inpcb *inp, struct sockaddr *nam, struct ucred *cred) { + struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; register struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam; struct in6_addr addr6; int error; - INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo); INP_WLOCK_ASSERT(inp); + INP_HASH_WLOCK_ASSERT(pcbinfo); /* * Call inner routine, to assign local interface address. * in6_pcbladdr() may automatically fill in sin6_scope_id. */ if ((error = in6_pcbladdr(inp, nam, &addr6)) != 0) return (error); - if (in6_pcblookup_hash(inp->inp_pcbinfo, &sin6->sin6_addr, + if (in6_pcblookup_hash_locked(pcbinfo, &sin6->sin6_addr, sin6->sin6_port, IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) ? &addr6 : &inp->in6p_laddr, inp->inp_lport, 0, NULL) != NULL) { return (EADDRINUSE); } if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) { if (inp->inp_lport == 0) { error = in6_pcbbind(inp, (struct sockaddr *)0, cred); if (error) return (error); } inp->in6p_laddr = addr6; } inp->in6p_faddr = sin6->sin6_addr; inp->inp_fport = sin6->sin6_port; /* update flowinfo - draft-itojun-ipv6-flowlabel-api-00 */ inp->inp_flow &= ~IPV6_FLOWLABEL_MASK; if (inp->inp_flags & IN6P_AUTOFLOWLABEL) inp->inp_flow |= (htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK); in_pcbrehash(inp); return (0); } void in6_pcbdisconnect(struct inpcb *inp) { - INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo); INP_WLOCK_ASSERT(inp); + INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo); bzero((caddr_t)&inp->in6p_faddr, sizeof(inp->in6p_faddr)); inp->inp_fport = 0; /* clear flowinfo - draft-itojun-ipv6-flowlabel-api-00 */ inp->inp_flow &= ~IPV6_FLOWLABEL_MASK; in_pcbrehash(inp); } struct sockaddr * in6_sockaddr(in_port_t port, struct in6_addr *addr_p) { struct sockaddr_in6 *sin6; sin6 = malloc(sizeof *sin6, M_SONAME, M_WAITOK); bzero(sin6, sizeof *sin6); sin6->sin6_family = AF_INET6; sin6->sin6_len = sizeof(*sin6); sin6->sin6_port = port; sin6->sin6_addr = *addr_p; (void)sa6_recoverscope(sin6); /* XXX: should catch errors */ return (struct sockaddr *)sin6; } struct sockaddr * in6_v4mapsin6_sockaddr(in_port_t port, struct in_addr *addr_p) { struct sockaddr_in sin; struct sockaddr_in6 *sin6_p; bzero(&sin, sizeof sin); sin.sin_family = AF_INET; sin.sin_len = sizeof(sin); sin.sin_port = port; sin.sin_addr = *addr_p; sin6_p = malloc(sizeof *sin6_p, M_SONAME, M_WAITOK); in6_sin_2_v4mapsin6(&sin, sin6_p); return (struct sockaddr *)sin6_p; } int in6_getsockaddr(struct socket *so, struct sockaddr **nam) { register struct inpcb *inp; struct in6_addr addr; in_port_t port; inp = sotoinpcb(so); KASSERT(inp != NULL, ("in6_getsockaddr: inp == NULL")); INP_RLOCK(inp); port = inp->inp_lport; addr = inp->in6p_laddr; INP_RUNLOCK(inp); *nam = in6_sockaddr(port, &addr); return 0; } int in6_getpeeraddr(struct socket *so, struct sockaddr **nam) { struct inpcb *inp; struct in6_addr addr; in_port_t port; inp = sotoinpcb(so); KASSERT(inp != NULL, ("in6_getpeeraddr: inp == NULL")); INP_RLOCK(inp); port = inp->inp_fport; addr = inp->in6p_faddr; INP_RUNLOCK(inp); *nam = in6_sockaddr(port, &addr); return 0; } int in6_mapped_sockaddr(struct socket *so, struct sockaddr **nam) { struct inpcb *inp; int error; inp = sotoinpcb(so); KASSERT(inp != NULL, ("in6_mapped_sockaddr: inp == NULL")); #ifdef INET if ((inp->inp_vflag & (INP_IPV4 | INP_IPV6)) == INP_IPV4) { error = in_getsockaddr(so, nam); if (error == 0) in6_sin_2_v4mapsin6_in_sock(nam); } else #endif { /* scope issues will be handled in in6_getsockaddr(). */ error = in6_getsockaddr(so, nam); } return error; } int in6_mapped_peeraddr(struct socket *so, struct sockaddr **nam) { struct inpcb *inp; int error; inp = sotoinpcb(so); KASSERT(inp != NULL, ("in6_mapped_peeraddr: inp == NULL")); #ifdef INET if ((inp->inp_vflag & (INP_IPV4 | INP_IPV6)) == INP_IPV4) { error = in_getpeeraddr(so, nam); if (error == 0) in6_sin_2_v4mapsin6_in_sock(nam); } else #endif /* scope issues will be handled in in6_getpeeraddr(). */ error = in6_getpeeraddr(so, nam); return error; } /* * Pass some notification to all connections of a protocol * associated with address dst. The local address and/or port numbers * may be specified to limit the search. The "usual action" will be * taken, depending on the ctlinput cmd. The caller must filter any * cmds that are uninteresting (e.g., no error in the map). * Call the protocol specific routine (if any) to report * any errors for each matching socket. */ void in6_pcbnotify(struct inpcbinfo *pcbinfo, struct sockaddr *dst, u_int fport_arg, const struct sockaddr *src, u_int lport_arg, int cmd, void *cmdarg, struct inpcb *(*notify)(struct inpcb *, int)) { struct inpcb *inp, *inp_temp; struct sockaddr_in6 sa6_src, *sa6_dst; u_short fport = fport_arg, lport = lport_arg; u_int32_t flowinfo; int errno; if ((unsigned)cmd >= PRC_NCMDS || dst->sa_family != AF_INET6) return; sa6_dst = (struct sockaddr_in6 *)dst; if (IN6_IS_ADDR_UNSPECIFIED(&sa6_dst->sin6_addr)) return; /* * note that src can be NULL when we get notify by local fragmentation. */ sa6_src = (src == NULL) ? sa6_any : *(const struct sockaddr_in6 *)src; flowinfo = sa6_src.sin6_flowinfo; /* * Redirects go to all references to the destination, * and use in6_rtchange to invalidate the route cache. * Dead host indications: also use in6_rtchange to invalidate * the cache, and deliver the error to all the sockets. * Otherwise, if we have knowledge of the local port and address, * deliver only to that socket. */ if (PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) { fport = 0; lport = 0; bzero((caddr_t)&sa6_src.sin6_addr, sizeof(sa6_src.sin6_addr)); if (cmd != PRC_HOSTDEAD) notify = in6_rtchange; } errno = inet6ctlerrmap[cmd]; INP_INFO_WLOCK(pcbinfo); LIST_FOREACH_SAFE(inp, pcbinfo->ipi_listhead, inp_list, inp_temp) { INP_WLOCK(inp); if ((inp->inp_vflag & INP_IPV6) == 0) { INP_WUNLOCK(inp); continue; } /* * If the error designates a new path MTU for a destination * and the application (associated with this socket) wanted to * know the value, notify. Note that we notify for all * disconnected sockets if the corresponding application * wanted. This is because some UDP applications keep sending * sockets disconnected. * XXX: should we avoid to notify the value to TCP sockets? */ if (cmd == PRC_MSGSIZE && (inp->inp_flags & IN6P_MTU) != 0 && (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) || IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &sa6_dst->sin6_addr))) { ip6_notify_pmtu(inp, (struct sockaddr_in6 *)dst, (u_int32_t *)cmdarg); } /* * Detect if we should notify the error. If no source and * destination ports are specifed, but non-zero flowinfo and * local address match, notify the error. This is the case * when the error is delivered with an encrypted buffer * by ESP. Otherwise, just compare addresses and ports * as usual. */ if (lport == 0 && fport == 0 && flowinfo && inp->inp_socket != NULL && flowinfo == (inp->inp_flow & IPV6_FLOWLABEL_MASK) && IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &sa6_src.sin6_addr)) goto do_notify; else if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &sa6_dst->sin6_addr) || inp->inp_socket == 0 || (lport && inp->inp_lport != lport) || (!IN6_IS_ADDR_UNSPECIFIED(&sa6_src.sin6_addr) && !IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &sa6_src.sin6_addr)) || (fport && inp->inp_fport != fport)) { INP_WUNLOCK(inp); continue; } do_notify: if (notify) { if ((*notify)(inp, errno)) INP_WUNLOCK(inp); } else INP_WUNLOCK(inp); } INP_INFO_WUNLOCK(pcbinfo); } /* - * Lookup a PCB based on the local address and port. + * Lookup a PCB based on the local address and port. Caller must hold the + * hash lock. No inpcb locks or references are acquired. */ struct inpcb * in6_pcblookup_local(struct inpcbinfo *pcbinfo, struct in6_addr *laddr, u_short lport, int lookupflags, struct ucred *cred) { register struct inpcb *inp; int matchwild = 3, wildcard; KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0, ("%s: invalid lookup flags %d", __func__, lookupflags)); - INP_INFO_WLOCK_ASSERT(pcbinfo); + INP_HASH_WLOCK_ASSERT(pcbinfo); if ((lookupflags & INPLOOKUP_WILDCARD) == 0) { struct inpcbhead *head; /* * Look for an unconnected (wildcard foreign addr) PCB that * matches the local address and port we're looking for. */ head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport, 0, pcbinfo->ipi_hashmask)]; LIST_FOREACH(inp, head, inp_hash) { /* XXX inp locking */ if ((inp->inp_vflag & INP_IPV6) == 0) continue; if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) && IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) && inp->inp_lport == lport) { /* Found. */ if (cred == NULL || prison_equal_ip6(cred->cr_prison, inp->inp_cred->cr_prison)) return (inp); } } /* * Not found. */ return (NULL); } else { struct inpcbporthead *porthash; struct inpcbport *phd; struct inpcb *match = NULL; /* * Best fit PCB lookup. * * First see if this local port is in use by looking on the * port hash list. */ porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport, pcbinfo->ipi_porthashmask)]; LIST_FOREACH(phd, porthash, phd_hash) { if (phd->phd_port == lport) break; } if (phd != NULL) { /* * Port is in use by one or more PCBs. Look for best * fit. */ LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) { wildcard = 0; if (cred != NULL && !prison_equal_ip6(cred->cr_prison, inp->inp_cred->cr_prison)) continue; /* XXX inp locking */ if ((inp->inp_vflag & INP_IPV6) == 0) continue; if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) wildcard++; if (!IN6_IS_ADDR_UNSPECIFIED( &inp->in6p_laddr)) { if (IN6_IS_ADDR_UNSPECIFIED(laddr)) wildcard++; else if (!IN6_ARE_ADDR_EQUAL( &inp->in6p_laddr, laddr)) continue; } else { if (!IN6_IS_ADDR_UNSPECIFIED(laddr)) wildcard++; } if (wildcard < matchwild) { match = inp; matchwild = wildcard; if (matchwild == 0) break; } } } return (match); } } void in6_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp) { struct inpcb *in6p; struct ip6_moptions *im6o; int i, gap; INP_INFO_RLOCK(pcbinfo); LIST_FOREACH(in6p, pcbinfo->ipi_listhead, inp_list) { INP_WLOCK(in6p); im6o = in6p->in6p_moptions; if ((in6p->inp_vflag & INP_IPV6) && im6o != NULL) { /* * Unselect the outgoing ifp for multicast if it * is being detached. */ if (im6o->im6o_multicast_ifp == ifp) im6o->im6o_multicast_ifp = NULL; /* * Drop multicast group membership if we joined * through the interface being detached. */ gap = 0; for (i = 0; i < im6o->im6o_num_memberships; i++) { if (im6o->im6o_membership[i]->in6m_ifp == ifp) { in6_mc_leave(im6o->im6o_membership[i], NULL); gap++; } else if (gap != 0) { im6o->im6o_membership[i - gap] = im6o->im6o_membership[i]; } } im6o->im6o_num_memberships -= gap; } INP_WUNLOCK(in6p); } INP_INFO_RUNLOCK(pcbinfo); } /* * Check for alternatives when higher level complains * about service problems. For now, invalidate cached * routing information. If the route was created dynamically * (by a redirect), time to try a default gateway again. */ void in6_losing(struct inpcb *in6p) { /* * We don't store route pointers in the routing table anymore */ return; } /* * After a routing change, flush old routing * and allocate a (hopefully) better one. */ struct inpcb * in6_rtchange(struct inpcb *inp, int errno) { /* * We don't store route pointers in the routing table anymore */ return inp; } /* * Lookup PCB in hash list. */ struct inpcb * -in6_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in6_addr *faddr, - u_int fport_arg, struct in6_addr *laddr, u_int lport_arg, int lookupflags, - struct ifnet *ifp) +in6_pcblookup_hash_locked(struct inpcbinfo *pcbinfo, struct in6_addr *faddr, + u_int fport_arg, struct in6_addr *laddr, u_int lport_arg, + int lookupflags, struct ifnet *ifp) { struct inpcbhead *head; struct inpcb *inp, *tmpinp; u_short fport = fport_arg, lport = lport_arg; int faith; KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0, ("%s: invalid lookup flags %d", __func__, lookupflags)); - INP_INFO_LOCK_ASSERT(pcbinfo); + INP_HASH_LOCK_ASSERT(pcbinfo); if (faithprefix_p != NULL) faith = (*faithprefix_p)(laddr); else faith = 0; /* * First look for an exact match. */ tmpinp = NULL; head = &pcbinfo->ipi_hashbase[ INP_PCBHASH(faddr->s6_addr32[3] /* XXX */, lport, fport, pcbinfo->ipi_hashmask)]; LIST_FOREACH(inp, head, inp_hash) { /* XXX inp locking */ if ((inp->inp_vflag & INP_IPV6) == 0) continue; if (IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, faddr) && IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) && inp->inp_fport == fport && inp->inp_lport == lport) { /* * XXX We should be able to directly return * the inp here, without any checks. * Well unless both bound with SO_REUSEPORT? */ if (prison_flag(inp->inp_cred, PR_IP6)) return (inp); if (tmpinp == NULL) tmpinp = inp; } } if (tmpinp != NULL) return (tmpinp); /* * Then look for a wildcard match, if requested. */ if ((lookupflags & INPLOOKUP_WILDCARD) != 0) { struct inpcb *local_wild = NULL, *local_exact = NULL; struct inpcb *jail_wild = NULL; int injail; /* * Order of socket selection - we always prefer jails. * 1. jailed, non-wild. * 2. jailed, wild. * 3. non-jailed, non-wild. * 4. non-jailed, wild. */ head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport, 0, pcbinfo->ipi_hashmask)]; LIST_FOREACH(inp, head, inp_hash) { /* XXX inp locking */ if ((inp->inp_vflag & INP_IPV6) == 0) continue; if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) || inp->inp_lport != lport) { continue; } /* XXX inp locking */ if (faith && (inp->inp_flags & INP_FAITH) == 0) continue; injail = prison_flag(inp->inp_cred, PR_IP6); if (injail) { if (prison_check_ip6(inp->inp_cred, laddr) != 0) continue; } else { if (local_exact != NULL) continue; } if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr)) { if (injail) return (inp); else local_exact = inp; } else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) { if (injail) jail_wild = inp; else local_wild = inp; } } /* LIST_FOREACH */ if (jail_wild != NULL) return (jail_wild); if (local_exact != NULL) return (local_exact); if (local_wild != NULL) return (local_wild); } /* if ((lookupflags & INPLOOKUP_WILDCARD) != 0) */ /* * Not found. */ return (NULL); +} + +/* + * Lookup PCB in hash list, using pcbinfo tables. This variation locks the + * hash list lock, and will return the inpcb locked (i.e., requires + * INPLOOKUP_LOCKPCB). + */ +static struct inpcb * +in6_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in6_addr *faddr, + u_int fport, struct in6_addr *laddr, u_int lport, int lookupflags, + struct ifnet *ifp) +{ + struct inpcb *inp; + + INP_HASH_RLOCK(pcbinfo); + inp = in6_pcblookup_hash_locked(pcbinfo, faddr, fport, laddr, lport, + (lookupflags & ~(INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)), ifp); + if (inp != NULL) { + in_pcbref(inp); + INP_HASH_RUNLOCK(pcbinfo); + if (lookupflags & INPLOOKUP_WLOCKPCB) { + INP_WLOCK(inp); + if (in_pcbrele_wlocked(inp)) + return (NULL); + } else if (lookupflags & INPLOOKUP_RLOCKPCB) { + INP_RLOCK(inp); + if (in_pcbrele_rlocked(inp)) + return (NULL); + } else + panic("%s: locking bug", __func__); + } else + INP_HASH_RUNLOCK(pcbinfo); + return (inp); +} + +/* + * Public inpcb lookup routines, accepting a 4-tuple. + */ +struct inpcb * +in6_pcblookup(struct inpcbinfo *pcbinfo, struct in6_addr *faddr, u_int fport, + struct in6_addr *laddr, u_int lport, int lookupflags, struct ifnet *ifp) +{ + + KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0, + ("%s: invalid lookup flags %d", __func__, lookupflags)); + KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0, + ("%s: LOCKPCB not set", __func__)); + + return (in6_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport, + lookupflags, ifp)); } void init_sin6(struct sockaddr_in6 *sin6, struct mbuf *m) { struct ip6_hdr *ip; ip = mtod(m, struct ip6_hdr *); bzero(sin6, sizeof(*sin6)); sin6->sin6_len = sizeof(*sin6); sin6->sin6_family = AF_INET6; sin6->sin6_addr = ip->ip6_src; (void)sa6_recoverscope(sin6); /* XXX: should catch errors... */ return; } Index: head/sys/netinet6/in6_pcb.h =================================================================== --- head/sys/netinet6/in6_pcb.h (revision 222487) +++ head/sys/netinet6/in6_pcb.h (revision 222488) @@ -1,104 +1,108 @@ /*- * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. Neither the name of the project nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * $KAME: in6_pcb.h,v 1.13 2001/02/06 09:16:53 itojun Exp $ */ /*- * Copyright (c) 1982, 1986, 1990, 1993 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)in_pcb.h 8.1 (Berkeley) 6/10/93 * $FreeBSD$ */ #ifndef _NETINET6_IN6_PCB_H_ #define _NETINET6_IN6_PCB_H_ #ifdef _KERNEL #define satosin6(sa) ((struct sockaddr_in6 *)(sa)) #define sin6tosa(sin6) ((struct sockaddr *)(sin6)) #define ifatoia6(ifa) ((struct in6_ifaddr *)(ifa)) void in6_pcbpurgeif0 __P((struct inpcbinfo *, struct ifnet *)); void in6_losing __P((struct inpcb *)); int in6_pcbbind __P((struct inpcb *, struct sockaddr *, struct ucred *)); int in6_pcbconnect __P((struct inpcb *, struct sockaddr *, struct ucred *)); void in6_pcbdisconnect __P((struct inpcb *)); int in6_pcbladdr(struct inpcb *, struct sockaddr *, struct in6_addr *); struct inpcb * in6_pcblookup_local __P((struct inpcbinfo *, struct in6_addr *, u_short, int, struct ucred *)); struct inpcb * - in6_pcblookup_hash __P((struct inpcbinfo *, - struct in6_addr *, u_int, struct in6_addr *, - u_int, int, struct ifnet *)); + in6_pcblookup __P((struct inpcbinfo *, struct in6_addr *, + u_int, struct in6_addr *, u_int, int, + struct ifnet *)); +struct inpcb * + in6_pcblookup_hash_locked __P((struct inpcbinfo *, struct in6_addr *, + u_int, struct in6_addr *, u_int, int, + struct ifnet *)); void in6_pcbnotify __P((struct inpcbinfo *, struct sockaddr *, u_int, const struct sockaddr *, u_int, int, void *, struct inpcb *(*)(struct inpcb *, int))); struct inpcb * in6_rtchange __P((struct inpcb *, int)); struct sockaddr * in6_sockaddr __P((in_port_t port, struct in6_addr *addr_p)); struct sockaddr * in6_v4mapsin6_sockaddr __P((in_port_t port, struct in_addr *addr_p)); int in6_getpeeraddr __P((struct socket *so, struct sockaddr **nam)); int in6_getsockaddr __P((struct socket *so, struct sockaddr **nam)); int in6_mapped_sockaddr __P((struct socket *so, struct sockaddr **nam)); int in6_mapped_peeraddr __P((struct socket *so, struct sockaddr **nam)); int in6_selecthlim __P((struct in6pcb *, struct ifnet *)); int in6_pcbsetport __P((struct in6_addr *, struct inpcb *, struct ucred *)); void init_sin6 __P((struct sockaddr_in6 *sin6, struct mbuf *m)); #endif /* _KERNEL */ #endif /* !_NETINET6_IN6_PCB_H_ */ Index: head/sys/netinet6/in6_src.c =================================================================== --- head/sys/netinet6/in6_src.c (revision 222487) +++ head/sys/netinet6/in6_src.c (revision 222488) @@ -1,1139 +1,1139 @@ /*- * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. Neither the name of the project nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * $KAME: in6_src.c,v 1.132 2003/08/26 04:42:27 keiichi Exp $ */ /*- * Copyright (c) 1982, 1986, 1991, 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. * * @(#)in_pcb.c 8.2 (Berkeley) 1/4/94 */ #include __FBSDID("$FreeBSD$"); #include "opt_inet.h" #include "opt_inet6.h" #include "opt_mpath.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef RADIX_MPATH #include #endif #include #include #include #include #include #include #include #include #include #include #include #include #include #include static struct mtx addrsel_lock; #define ADDRSEL_LOCK_INIT() mtx_init(&addrsel_lock, "addrsel_lock", NULL, MTX_DEF) #define ADDRSEL_LOCK() mtx_lock(&addrsel_lock) #define ADDRSEL_UNLOCK() mtx_unlock(&addrsel_lock) #define ADDRSEL_LOCK_ASSERT() mtx_assert(&addrsel_lock, MA_OWNED) static struct sx addrsel_sxlock; #define ADDRSEL_SXLOCK_INIT() sx_init(&addrsel_sxlock, "addrsel_sxlock") #define ADDRSEL_SLOCK() sx_slock(&addrsel_sxlock) #define ADDRSEL_SUNLOCK() sx_sunlock(&addrsel_sxlock) #define ADDRSEL_XLOCK() sx_xlock(&addrsel_sxlock) #define ADDRSEL_XUNLOCK() sx_xunlock(&addrsel_sxlock) #define ADDR_LABEL_NOTAPP (-1) static VNET_DEFINE(struct in6_addrpolicy, defaultaddrpolicy); #define V_defaultaddrpolicy VNET(defaultaddrpolicy) VNET_DEFINE(int, ip6_prefer_tempaddr) = 0; static int selectroute __P((struct sockaddr_in6 *, struct ip6_pktopts *, struct ip6_moptions *, struct route_in6 *, struct ifnet **, struct rtentry **, int)); static int in6_selectif __P((struct sockaddr_in6 *, struct ip6_pktopts *, struct ip6_moptions *, struct route_in6 *ro, struct ifnet **)); static struct in6_addrpolicy *lookup_addrsel_policy(struct sockaddr_in6 *); static void init_policy_queue(void); static int add_addrsel_policyent(struct in6_addrpolicy *); static int delete_addrsel_policyent(struct in6_addrpolicy *); static int walk_addrsel_policy __P((int (*)(struct in6_addrpolicy *, void *), void *)); static int dump_addrsel_policyent(struct in6_addrpolicy *, void *); static struct in6_addrpolicy *match_addrsel_policy(struct sockaddr_in6 *); /* * Return an IPv6 address, which is the most appropriate for a given * destination and user specified options. * If necessary, this function lookups the routing table and returns * an entry to the caller for later use. */ #define REPLACE(r) do {\ if ((r) < sizeof(V_ip6stat.ip6s_sources_rule) / \ sizeof(V_ip6stat.ip6s_sources_rule[0])) /* check for safety */ \ V_ip6stat.ip6s_sources_rule[(r)]++; \ /* { \ char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \ printf("in6_selectsrc: replace %s with %s by %d\n", ia_best ? ip6_sprintf(ip6buf, &ia_best->ia_addr.sin6_addr) : "none", ip6_sprintf(ip6b, &ia->ia_addr.sin6_addr), (r)); \ } */ \ goto replace; \ } while(0) #define NEXT(r) do {\ if ((r) < sizeof(V_ip6stat.ip6s_sources_rule) / \ sizeof(V_ip6stat.ip6s_sources_rule[0])) /* check for safety */ \ V_ip6stat.ip6s_sources_rule[(r)]++; \ /* { \ char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \ printf("in6_selectsrc: keep %s against %s by %d\n", ia_best ? ip6_sprintf(ip6buf, &ia_best->ia_addr.sin6_addr) : "none", ip6_sprintf(ip6b, &ia->ia_addr.sin6_addr), (r)); \ } */ \ goto next; /* XXX: we can't use 'continue' here */ \ } while(0) #define BREAK(r) do { \ if ((r) < sizeof(V_ip6stat.ip6s_sources_rule) / \ sizeof(V_ip6stat.ip6s_sources_rule[0])) /* check for safety */ \ V_ip6stat.ip6s_sources_rule[(r)]++; \ goto out; /* XXX: we can't use 'break' here */ \ } while(0) int in6_selectsrc(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, struct inpcb *inp, struct route_in6 *ro, struct ucred *cred, struct ifnet **ifpp, struct in6_addr *srcp) { struct in6_addr dst, tmp; struct ifnet *ifp = NULL; struct in6_ifaddr *ia = NULL, *ia_best = NULL; struct in6_pktinfo *pi = NULL; int dst_scope = -1, best_scope = -1, best_matchlen = -1; struct in6_addrpolicy *dst_policy = NULL, *best_policy = NULL; u_int32_t odstzone; int prefer_tempaddr; int error; struct ip6_moptions *mopts; KASSERT(srcp != NULL, ("%s: srcp is NULL", __func__)); dst = dstsock->sin6_addr; /* make a copy for local operation */ if (ifpp) *ifpp = NULL; if (inp != NULL) { INP_LOCK_ASSERT(inp); mopts = inp->in6p_moptions; } else { mopts = NULL; } /* * If the source address is explicitly specified by the caller, * check if the requested source address is indeed a unicast address * assigned to the node, and can be used as the packet's source * address. If everything is okay, use the address as source. */ if (opts && (pi = opts->ip6po_pktinfo) && !IN6_IS_ADDR_UNSPECIFIED(&pi->ipi6_addr)) { struct sockaddr_in6 srcsock; struct in6_ifaddr *ia6; /* get the outgoing interface */ if ((error = in6_selectif(dstsock, opts, mopts, ro, &ifp)) != 0) return (error); /* * determine the appropriate zone id of the source based on * the zone of the destination and the outgoing interface. * If the specified address is ambiguous wrt the scope zone, * the interface must be specified; otherwise, ifa_ifwithaddr() * will fail matching the address. */ bzero(&srcsock, sizeof(srcsock)); srcsock.sin6_family = AF_INET6; srcsock.sin6_len = sizeof(srcsock); srcsock.sin6_addr = pi->ipi6_addr; if (ifp) { error = in6_setscope(&srcsock.sin6_addr, ifp, NULL); if (error) return (error); } if (cred != NULL && (error = prison_local_ip6(cred, &srcsock.sin6_addr, (inp != NULL && (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0) return (error); ia6 = (struct in6_ifaddr *)ifa_ifwithaddr( (struct sockaddr *)&srcsock); if (ia6 == NULL || (ia6->ia6_flags & (IN6_IFF_ANYCAST | IN6_IFF_NOTREADY))) { if (ia6 != NULL) ifa_free(&ia6->ia_ifa); return (EADDRNOTAVAIL); } pi->ipi6_addr = srcsock.sin6_addr; /* XXX: this overrides pi */ if (ifpp) *ifpp = ifp; bcopy(&ia6->ia_addr.sin6_addr, srcp, sizeof(*srcp)); ifa_free(&ia6->ia_ifa); return (0); } /* * Otherwise, if the socket has already bound the source, just use it. */ if (inp != NULL && !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) { if (cred != NULL && (error = prison_local_ip6(cred, &inp->in6p_laddr, ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0) return (error); bcopy(&inp->in6p_laddr, srcp, sizeof(*srcp)); return (0); } /* * Bypass source address selection and use the primary jail IP * if requested. */ if (cred != NULL && !prison_saddrsel_ip6(cred, srcp)) return (0); /* * If the address is not specified, choose the best one based on * the outgoing interface and the destination address. */ /* get the outgoing interface */ if ((error = in6_selectif(dstsock, opts, mopts, ro, &ifp)) != 0) return (error); #ifdef DIAGNOSTIC if (ifp == NULL) /* this should not happen */ panic("in6_selectsrc: NULL ifp"); #endif error = in6_setscope(&dst, ifp, &odstzone); if (error) return (error); IN6_IFADDR_RLOCK(); TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) { int new_scope = -1, new_matchlen = -1; struct in6_addrpolicy *new_policy = NULL; u_int32_t srczone, osrczone, dstzone; struct in6_addr src; struct ifnet *ifp1 = ia->ia_ifp; /* * We'll never take an address that breaks the scope zone * of the destination. We also skip an address if its zone * does not contain the outgoing interface. * XXX: we should probably use sin6_scope_id here. */ if (in6_setscope(&dst, ifp1, &dstzone) || odstzone != dstzone) { continue; } src = ia->ia_addr.sin6_addr; if (in6_setscope(&src, ifp, &osrczone) || in6_setscope(&src, ifp1, &srczone) || osrczone != srczone) { continue; } /* avoid unusable addresses */ if ((ia->ia6_flags & (IN6_IFF_NOTREADY | IN6_IFF_ANYCAST | IN6_IFF_DETACHED))) { continue; } if (!V_ip6_use_deprecated && IFA6_IS_DEPRECATED(ia)) continue; /* If jailed only take addresses of the jail into account. */ if (cred != NULL && prison_check_ip6(cred, &ia->ia_addr.sin6_addr) != 0) continue; /* Rule 1: Prefer same address */ if (IN6_ARE_ADDR_EQUAL(&dst, &ia->ia_addr.sin6_addr)) { ia_best = ia; BREAK(1); /* there should be no better candidate */ } if (ia_best == NULL) REPLACE(0); /* Rule 2: Prefer appropriate scope */ if (dst_scope < 0) dst_scope = in6_addrscope(&dst); new_scope = in6_addrscope(&ia->ia_addr.sin6_addr); if (IN6_ARE_SCOPE_CMP(best_scope, new_scope) < 0) { if (IN6_ARE_SCOPE_CMP(best_scope, dst_scope) < 0) REPLACE(2); NEXT(2); } else if (IN6_ARE_SCOPE_CMP(new_scope, best_scope) < 0) { if (IN6_ARE_SCOPE_CMP(new_scope, dst_scope) < 0) NEXT(2); REPLACE(2); } /* * Rule 3: Avoid deprecated addresses. Note that the case of * !ip6_use_deprecated is already rejected above. */ if (!IFA6_IS_DEPRECATED(ia_best) && IFA6_IS_DEPRECATED(ia)) NEXT(3); if (IFA6_IS_DEPRECATED(ia_best) && !IFA6_IS_DEPRECATED(ia)) REPLACE(3); /* Rule 4: Prefer home addresses */ /* * XXX: This is a TODO. We should probably merge the MIP6 * case above. */ /* Rule 5: Prefer outgoing interface */ if (ia_best->ia_ifp == ifp && ia->ia_ifp != ifp) NEXT(5); if (ia_best->ia_ifp != ifp && ia->ia_ifp == ifp) REPLACE(5); /* * Rule 6: Prefer matching label * Note that best_policy should be non-NULL here. */ if (dst_policy == NULL) dst_policy = lookup_addrsel_policy(dstsock); if (dst_policy->label != ADDR_LABEL_NOTAPP) { new_policy = lookup_addrsel_policy(&ia->ia_addr); if (dst_policy->label == best_policy->label && dst_policy->label != new_policy->label) NEXT(6); if (dst_policy->label != best_policy->label && dst_policy->label == new_policy->label) REPLACE(6); } /* * Rule 7: Prefer public addresses. * We allow users to reverse the logic by configuring * a sysctl variable, so that privacy conscious users can * always prefer temporary addresses. */ if (opts == NULL || opts->ip6po_prefer_tempaddr == IP6PO_TEMPADDR_SYSTEM) { prefer_tempaddr = V_ip6_prefer_tempaddr; } else if (opts->ip6po_prefer_tempaddr == IP6PO_TEMPADDR_NOTPREFER) { prefer_tempaddr = 0; } else prefer_tempaddr = 1; if (!(ia_best->ia6_flags & IN6_IFF_TEMPORARY) && (ia->ia6_flags & IN6_IFF_TEMPORARY)) { if (prefer_tempaddr) REPLACE(7); else NEXT(7); } if ((ia_best->ia6_flags & IN6_IFF_TEMPORARY) && !(ia->ia6_flags & IN6_IFF_TEMPORARY)) { if (prefer_tempaddr) NEXT(7); else REPLACE(7); } /* * Rule 8: prefer addresses on alive interfaces. * This is a KAME specific rule. */ if ((ia_best->ia_ifp->if_flags & IFF_UP) && !(ia->ia_ifp->if_flags & IFF_UP)) NEXT(8); if (!(ia_best->ia_ifp->if_flags & IFF_UP) && (ia->ia_ifp->if_flags & IFF_UP)) REPLACE(8); /* * Rule 14: Use longest matching prefix. * Note: in the address selection draft, this rule is * documented as "Rule 8". However, since it is also * documented that this rule can be overridden, we assign * a large number so that it is easy to assign smaller numbers * to more preferred rules. */ new_matchlen = in6_matchlen(&ia->ia_addr.sin6_addr, &dst); if (best_matchlen < new_matchlen) REPLACE(14); if (new_matchlen < best_matchlen) NEXT(14); /* Rule 15 is reserved. */ /* * Last resort: just keep the current candidate. * Or, do we need more rules? */ continue; replace: ia_best = ia; best_scope = (new_scope >= 0 ? new_scope : in6_addrscope(&ia_best->ia_addr.sin6_addr)); best_policy = (new_policy ? new_policy : lookup_addrsel_policy(&ia_best->ia_addr)); best_matchlen = (new_matchlen >= 0 ? new_matchlen : in6_matchlen(&ia_best->ia_addr.sin6_addr, &dst)); next: continue; out: break; } if ((ia = ia_best) == NULL) { IN6_IFADDR_RUNLOCK(); return (EADDRNOTAVAIL); } /* * At this point at least one of the addresses belonged to the jail * but it could still be, that we want to further restrict it, e.g. * theoratically IN6_IS_ADDR_LOOPBACK. * It must not be IN6_IS_ADDR_UNSPECIFIED anymore. * prison_local_ip6() will fix an IN6_IS_ADDR_LOOPBACK but should * let all others previously selected pass. * Use tmp to not change ::1 on lo0 to the primary jail address. */ tmp = ia->ia_addr.sin6_addr; if (cred != NULL && prison_local_ip6(cred, &tmp, (inp != NULL && (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0)) != 0) { IN6_IFADDR_RUNLOCK(); return (EADDRNOTAVAIL); } if (ifpp) *ifpp = ifp; bcopy(&tmp, srcp, sizeof(*srcp)); IN6_IFADDR_RUNLOCK(); return (0); } /* * clone - meaningful only for bsdi and freebsd */ static int selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, struct ip6_moptions *mopts, struct route_in6 *ro, struct ifnet **retifp, struct rtentry **retrt, int norouteok) { int error = 0; struct ifnet *ifp = NULL; struct rtentry *rt = NULL; struct sockaddr_in6 *sin6_next; struct in6_pktinfo *pi = NULL; struct in6_addr *dst = &dstsock->sin6_addr; #if 0 char ip6buf[INET6_ADDRSTRLEN]; if (dstsock->sin6_addr.s6_addr32[0] == 0 && dstsock->sin6_addr.s6_addr32[1] == 0 && !IN6_IS_ADDR_LOOPBACK(&dstsock->sin6_addr)) { printf("in6_selectroute: strange destination %s\n", ip6_sprintf(ip6buf, &dstsock->sin6_addr)); } else { printf("in6_selectroute: destination = %s%%%d\n", ip6_sprintf(ip6buf, &dstsock->sin6_addr), dstsock->sin6_scope_id); /* for debug */ } #endif /* If the caller specify the outgoing interface explicitly, use it. */ if (opts && (pi = opts->ip6po_pktinfo) != NULL && pi->ipi6_ifindex) { /* XXX boundary check is assumed to be already done. */ ifp = ifnet_byindex(pi->ipi6_ifindex); if (ifp != NULL && (norouteok || retrt == NULL || IN6_IS_ADDR_MULTICAST(dst))) { /* * we do not have to check or get the route for * multicast. */ goto done; } else goto getroute; } /* * If the destination address is a multicast address and the outgoing * interface for the address is specified by the caller, use it. */ if (IN6_IS_ADDR_MULTICAST(dst) && mopts != NULL && (ifp = mopts->im6o_multicast_ifp) != NULL) { goto done; /* we do not need a route for multicast. */ } getroute: /* * If the next hop address for the packet is specified by the caller, * use it as the gateway. */ if (opts && opts->ip6po_nexthop) { struct route_in6 *ron; struct llentry *la; sin6_next = satosin6(opts->ip6po_nexthop); /* at this moment, we only support AF_INET6 next hops */ if (sin6_next->sin6_family != AF_INET6) { error = EAFNOSUPPORT; /* or should we proceed? */ goto done; } /* * If the next hop is an IPv6 address, then the node identified * by that address must be a neighbor of the sending host. */ ron = &opts->ip6po_nextroute; /* * XXX what do we do here? * PLZ to be fixing */ if (ron->ro_rt == NULL) { rtalloc((struct route *)ron); /* multi path case? */ if (ron->ro_rt == NULL) { if (ron->ro_rt) { RTFREE(ron->ro_rt); ron->ro_rt = NULL; } error = EHOSTUNREACH; goto done; } } rt = ron->ro_rt; ifp = rt->rt_ifp; IF_AFDATA_LOCK(ifp); la = lla_lookup(LLTABLE6(ifp), 0, (struct sockaddr *)&sin6_next->sin6_addr); IF_AFDATA_UNLOCK(ifp); if (la != NULL) LLE_RUNLOCK(la); else { error = EHOSTUNREACH; goto done; } #if 0 if ((ron->ro_rt && (ron->ro_rt->rt_flags & (RTF_UP | RTF_LLINFO)) != (RTF_UP | RTF_LLINFO)) || !IN6_ARE_ADDR_EQUAL(&satosin6(&ron->ro_dst)->sin6_addr, &sin6_next->sin6_addr)) { if (ron->ro_rt) { RTFREE(ron->ro_rt); ron->ro_rt = NULL; } *satosin6(&ron->ro_dst) = *sin6_next; } if (ron->ro_rt == NULL) { rtalloc((struct route *)ron); /* multi path case? */ if (ron->ro_rt == NULL || !(ron->ro_rt->rt_flags & RTF_LLINFO)) { if (ron->ro_rt) { RTFREE(ron->ro_rt); ron->ro_rt = NULL; } error = EHOSTUNREACH; goto done; } } #endif /* * When cloning is required, try to allocate a route to the * destination so that the caller can store path MTU * information. */ goto done; } /* * Use a cached route if it exists and is valid, else try to allocate * a new one. Note that we should check the address family of the * cached destination, in case of sharing the cache with IPv4. */ if (ro) { if (ro->ro_rt && (!(ro->ro_rt->rt_flags & RTF_UP) || ((struct sockaddr *)(&ro->ro_dst))->sa_family != AF_INET6 || !IN6_ARE_ADDR_EQUAL(&satosin6(&ro->ro_dst)->sin6_addr, dst))) { RTFREE(ro->ro_rt); ro->ro_rt = (struct rtentry *)NULL; } if (ro->ro_rt == (struct rtentry *)NULL) { struct sockaddr_in6 *sa6; /* No route yet, so try to acquire one */ bzero(&ro->ro_dst, sizeof(struct sockaddr_in6)); sa6 = (struct sockaddr_in6 *)&ro->ro_dst; *sa6 = *dstsock; sa6->sin6_scope_id = 0; #ifdef RADIX_MPATH rtalloc_mpath((struct route *)ro, ntohl(sa6->sin6_addr.s6_addr32[3])); #else ro->ro_rt = rtalloc1(&((struct route *)ro) ->ro_dst, 0, 0UL); if (ro->ro_rt) RT_UNLOCK(ro->ro_rt); #endif } /* * do not care about the result if we have the nexthop * explicitly specified. */ if (opts && opts->ip6po_nexthop) goto done; if (ro->ro_rt) { ifp = ro->ro_rt->rt_ifp; if (ifp == NULL) { /* can this really happen? */ RTFREE(ro->ro_rt); ro->ro_rt = NULL; } } if (ro->ro_rt == NULL) error = EHOSTUNREACH; rt = ro->ro_rt; /* * Check if the outgoing interface conflicts with * the interface specified by ipi6_ifindex (if specified). * Note that loopback interface is always okay. * (this may happen when we are sending a packet to one of * our own addresses.) */ if (ifp && opts && opts->ip6po_pktinfo && opts->ip6po_pktinfo->ipi6_ifindex) { if (!(ifp->if_flags & IFF_LOOPBACK) && ifp->if_index != opts->ip6po_pktinfo->ipi6_ifindex) { error = EHOSTUNREACH; goto done; } } } done: if (ifp == NULL && rt == NULL) { /* * This can happen if the caller did not pass a cached route * nor any other hints. We treat this case an error. */ error = EHOSTUNREACH; } if (error == EHOSTUNREACH) V_ip6stat.ip6s_noroute++; if (retifp != NULL) { *retifp = 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); } } if (retrt != NULL) *retrt = rt; /* rt may be NULL */ return (error); } static int in6_selectif(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, struct ip6_moptions *mopts, struct route_in6 *ro, struct ifnet **retifp) { int error; struct route_in6 sro; struct rtentry *rt = NULL; if (ro == NULL) { bzero(&sro, sizeof(sro)); ro = &sro; } if ((error = selectroute(dstsock, opts, mopts, ro, retifp, &rt, 1)) != 0) { if (ro == &sro && rt && rt == sro.ro_rt) RTFREE(rt); return (error); } /* * do not use a rejected or black hole route. * XXX: this check should be done in the L2 output routine. * However, if we skipped this check here, we'd see the following * scenario: * - install a rejected route for a scoped address prefix * (like fe80::/10) * - send a packet to a destination that matches the scoped prefix, * with ambiguity about the scope zone. * - pick the outgoing interface from the route, and disambiguate the * scope zone with the interface. * - ip6_output() would try to get another route with the "new" * destination, which may be valid. * - we'd see no error on output. * Although this may not be very harmful, it should still be confusing. * We thus reject the case here. */ if (rt && (rt->rt_flags & (RTF_REJECT | RTF_BLACKHOLE))) { int flags = (rt->rt_flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH); if (ro == &sro && rt && rt == sro.ro_rt) RTFREE(rt); return (flags); } if (ro == &sro && rt && rt == sro.ro_rt) RTFREE(rt); return (0); } /* * clone - meaningful only for bsdi and freebsd */ int in6_selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, struct ip6_moptions *mopts, struct route_in6 *ro, struct ifnet **retifp, struct rtentry **retrt) { return (selectroute(dstsock, opts, mopts, ro, retifp, retrt, 0)); } /* * Default hop limit selection. The precedence is as follows: * 1. Hoplimit value specified via ioctl. * 2. (If the outgoing interface is detected) the current * hop limit of the interface specified by router advertisement. * 3. The system default hoplimit. */ int in6_selecthlim(struct inpcb *in6p, struct ifnet *ifp) { if (in6p && in6p->in6p_hops >= 0) return (in6p->in6p_hops); else if (ifp) return (ND_IFINFO(ifp)->chlim); else if (in6p && !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) { struct route_in6 ro6; struct ifnet *lifp; bzero(&ro6, sizeof(ro6)); ro6.ro_dst.sin6_family = AF_INET6; ro6.ro_dst.sin6_len = sizeof(struct sockaddr_in6); ro6.ro_dst.sin6_addr = in6p->in6p_faddr; rtalloc((struct route *)&ro6); if (ro6.ro_rt) { lifp = ro6.ro_rt->rt_ifp; RTFREE(ro6.ro_rt); if (lifp) return (ND_IFINFO(lifp)->chlim); } else return (V_ip6_defhlim); } return (V_ip6_defhlim); } /* * XXX: this is borrowed from in6_pcbbind(). If possible, we should * share this function by all *bsd*... */ int in6_pcbsetport(struct in6_addr *laddr, struct inpcb *inp, struct ucred *cred) { struct socket *so = inp->inp_socket; u_int16_t lport = 0; int error, lookupflags = 0; #ifdef INVARIANTS struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; #endif - INP_INFO_WLOCK_ASSERT(pcbinfo); INP_WLOCK_ASSERT(inp); + INP_HASH_WLOCK_ASSERT(pcbinfo); error = prison_local_ip6(cred, laddr, ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0)); if (error) return(error); /* XXX: this is redundant when called from in6_pcbbind */ if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0) lookupflags = INPLOOKUP_WILDCARD; inp->inp_flags |= INP_ANONPORT; error = in_pcb_lport(inp, NULL, &lport, cred, lookupflags); if (error != 0) return (error); inp->inp_lport = lport; if (in_pcbinshash(inp) != 0) { inp->in6p_laddr = in6addr_any; inp->inp_lport = 0; return (EAGAIN); } return (0); } void addrsel_policy_init(void) { init_policy_queue(); /* initialize the "last resort" policy */ bzero(&V_defaultaddrpolicy, sizeof(V_defaultaddrpolicy)); V_defaultaddrpolicy.label = ADDR_LABEL_NOTAPP; if (!IS_DEFAULT_VNET(curvnet)) return; ADDRSEL_LOCK_INIT(); ADDRSEL_SXLOCK_INIT(); } static struct in6_addrpolicy * lookup_addrsel_policy(struct sockaddr_in6 *key) { struct in6_addrpolicy *match = NULL; ADDRSEL_LOCK(); match = match_addrsel_policy(key); if (match == NULL) match = &V_defaultaddrpolicy; else match->use++; ADDRSEL_UNLOCK(); return (match); } /* * Subroutines to manage the address selection policy table via sysctl. */ struct walkarg { struct sysctl_req *w_req; }; static int in6_src_sysctl(SYSCTL_HANDLER_ARGS); SYSCTL_DECL(_net_inet6_ip6); SYSCTL_NODE(_net_inet6_ip6, IPV6CTL_ADDRCTLPOLICY, addrctlpolicy, CTLFLAG_RD, in6_src_sysctl, ""); static int in6_src_sysctl(SYSCTL_HANDLER_ARGS) { struct walkarg w; if (req->newptr) return EPERM; bzero(&w, sizeof(w)); w.w_req = req; return (walk_addrsel_policy(dump_addrsel_policyent, &w)); } int in6_src_ioctl(u_long cmd, caddr_t data) { int i; struct in6_addrpolicy ent0; if (cmd != SIOCAADDRCTL_POLICY && cmd != SIOCDADDRCTL_POLICY) return (EOPNOTSUPP); /* check for safety */ ent0 = *(struct in6_addrpolicy *)data; if (ent0.label == ADDR_LABEL_NOTAPP) return (EINVAL); /* check if the prefix mask is consecutive. */ if (in6_mask2len(&ent0.addrmask.sin6_addr, NULL) < 0) return (EINVAL); /* clear trailing garbages (if any) of the prefix address. */ for (i = 0; i < 4; i++) { ent0.addr.sin6_addr.s6_addr32[i] &= ent0.addrmask.sin6_addr.s6_addr32[i]; } ent0.use = 0; switch (cmd) { case SIOCAADDRCTL_POLICY: return (add_addrsel_policyent(&ent0)); case SIOCDADDRCTL_POLICY: return (delete_addrsel_policyent(&ent0)); } return (0); /* XXX: compromise compilers */ } /* * The followings are implementation of the policy table using a * simple tail queue. * XXX such details should be hidden. * XXX implementation using binary tree should be more efficient. */ struct addrsel_policyent { TAILQ_ENTRY(addrsel_policyent) ape_entry; struct in6_addrpolicy ape_policy; }; TAILQ_HEAD(addrsel_policyhead, addrsel_policyent); static VNET_DEFINE(struct addrsel_policyhead, addrsel_policytab); #define V_addrsel_policytab VNET(addrsel_policytab) static void init_policy_queue(void) { TAILQ_INIT(&V_addrsel_policytab); } static int add_addrsel_policyent(struct in6_addrpolicy *newpolicy) { struct addrsel_policyent *new, *pol; new = malloc(sizeof(*new), M_IFADDR, M_WAITOK); ADDRSEL_XLOCK(); ADDRSEL_LOCK(); /* duplication check */ TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) { if (IN6_ARE_ADDR_EQUAL(&newpolicy->addr.sin6_addr, &pol->ape_policy.addr.sin6_addr) && IN6_ARE_ADDR_EQUAL(&newpolicy->addrmask.sin6_addr, &pol->ape_policy.addrmask.sin6_addr)) { ADDRSEL_UNLOCK(); ADDRSEL_XUNLOCK(); free(new, M_IFADDR); return (EEXIST); /* or override it? */ } } bzero(new, sizeof(*new)); /* XXX: should validate entry */ new->ape_policy = *newpolicy; TAILQ_INSERT_TAIL(&V_addrsel_policytab, new, ape_entry); ADDRSEL_UNLOCK(); ADDRSEL_XUNLOCK(); return (0); } static int delete_addrsel_policyent(struct in6_addrpolicy *key) { struct addrsel_policyent *pol; ADDRSEL_XLOCK(); ADDRSEL_LOCK(); /* search for the entry in the table */ TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) { if (IN6_ARE_ADDR_EQUAL(&key->addr.sin6_addr, &pol->ape_policy.addr.sin6_addr) && IN6_ARE_ADDR_EQUAL(&key->addrmask.sin6_addr, &pol->ape_policy.addrmask.sin6_addr)) { break; } } if (pol == NULL) { ADDRSEL_UNLOCK(); ADDRSEL_XUNLOCK(); return (ESRCH); } TAILQ_REMOVE(&V_addrsel_policytab, pol, ape_entry); ADDRSEL_UNLOCK(); ADDRSEL_XUNLOCK(); return (0); } static int walk_addrsel_policy(int (*callback)(struct in6_addrpolicy *, void *), void *w) { struct addrsel_policyent *pol; int error = 0; ADDRSEL_SLOCK(); TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) { if ((error = (*callback)(&pol->ape_policy, w)) != 0) { ADDRSEL_SUNLOCK(); return (error); } } ADDRSEL_SUNLOCK(); return (error); } static int dump_addrsel_policyent(struct in6_addrpolicy *pol, void *arg) { int error = 0; struct walkarg *w = arg; error = SYSCTL_OUT(w->w_req, pol, sizeof(*pol)); return (error); } static struct in6_addrpolicy * match_addrsel_policy(struct sockaddr_in6 *key) { struct addrsel_policyent *pent; struct in6_addrpolicy *bestpol = NULL, *pol; int matchlen, bestmatchlen = -1; u_char *mp, *ep, *k, *p, m; TAILQ_FOREACH(pent, &V_addrsel_policytab, ape_entry) { matchlen = 0; pol = &pent->ape_policy; mp = (u_char *)&pol->addrmask.sin6_addr; ep = mp + 16; /* XXX: scope field? */ k = (u_char *)&key->sin6_addr; p = (u_char *)&pol->addr.sin6_addr; for (; mp < ep && *mp; mp++, k++, p++) { m = *mp; if ((*k & m) != *p) goto next; /* not match */ if (m == 0xff) /* short cut for a typical case */ matchlen += 8; else { while (m >= 0x80) { matchlen++; m <<= 1; } } } /* matched. check if this is better than the current best. */ if (bestpol == NULL || matchlen > bestmatchlen) { bestpol = pol; bestmatchlen = matchlen; } next: continue; } return (bestpol); } Index: head/sys/netinet6/udp6_usrreq.c =================================================================== --- head/sys/netinet6/udp6_usrreq.c (revision 222487) +++ head/sys/netinet6/udp6_usrreq.c (revision 222488) @@ -1,1106 +1,1116 @@ /*- * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. + * Copyright (c) 2010-2011 Juniper Networks, Inc. * All rights reserved. * + * Portions of this software were developed by Robert N. M. Watson under + * contract to Juniper Networks, Inc. + * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. Neither the name of the project nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * $KAME: udp6_usrreq.c,v 1.27 2001/05/21 05:45:10 jinmei Exp $ * $KAME: udp6_output.c,v 1.31 2001/05/21 16:39:15 jinmei Exp $ */ /*- * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995 * The Regents of the University of California. * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)udp_usrreq.c 8.6 (Berkeley) 5/23/95 */ #include __FBSDID("$FreeBSD$"); #include "opt_inet.h" #include "opt_inet6.h" #include "opt_ipsec.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef IPSEC #include #include #endif /* IPSEC */ #include /* * UDP protocol implementation. * Per RFC 768, August, 1980. */ extern struct protosw inetsw[]; static void udp6_detach(struct socket *so); static void udp6_append(struct inpcb *inp, struct mbuf *n, int off, struct sockaddr_in6 *fromsa) { struct socket *so; struct mbuf *opts; INP_LOCK_ASSERT(inp); #ifdef IPSEC /* Check AH/ESP integrity. */ if (ipsec6_in_reject(n, inp)) { m_freem(n); V_ipsec6stat.in_polvio++; return; } #endif /* IPSEC */ #ifdef MAC if (mac_inpcb_check_deliver(inp, n) != 0) { m_freem(n); return; } #endif opts = NULL; if (inp->inp_flags & INP_CONTROLOPTS || inp->inp_socket->so_options & SO_TIMESTAMP) ip6_savecontrol(inp, n, &opts); m_adj(n, off + sizeof(struct udphdr)); so = inp->inp_socket; SOCKBUF_LOCK(&so->so_rcv); if (sbappendaddr_locked(&so->so_rcv, (struct sockaddr *)fromsa, n, opts) == 0) { SOCKBUF_UNLOCK(&so->so_rcv); m_freem(n); if (opts) m_freem(opts); UDPSTAT_INC(udps_fullsock); } else sorwakeup_locked(so); } int udp6_input(struct mbuf **mp, int *offp, int proto) { struct mbuf *m = *mp; struct ifnet *ifp; struct ip6_hdr *ip6; struct udphdr *uh; struct inpcb *inp; struct udpcb *up; int off = *offp; int plen, ulen; struct sockaddr_in6 fromsa; ifp = m->m_pkthdr.rcvif; ip6 = mtod(m, struct ip6_hdr *); if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) { /* XXX send icmp6 host/port unreach? */ m_freem(m); return (IPPROTO_DONE); } #ifndef PULLDOWN_TEST IP6_EXTHDR_CHECK(m, off, sizeof(struct udphdr), IPPROTO_DONE); ip6 = mtod(m, struct ip6_hdr *); uh = (struct udphdr *)((caddr_t)ip6 + off); #else IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(*uh)); if (!uh) return (IPPROTO_DONE); #endif UDPSTAT_INC(udps_ipackets); /* * Destination port of 0 is illegal, based on RFC768. */ if (uh->uh_dport == 0) goto badunlocked; plen = ntohs(ip6->ip6_plen) - off + sizeof(*ip6); ulen = ntohs((u_short)uh->uh_ulen); if (plen != ulen) { UDPSTAT_INC(udps_badlen); goto badunlocked; } /* * Checksum extended UDP header and data. */ if (uh->uh_sum == 0) { UDPSTAT_INC(udps_nosum); goto badunlocked; } if (in6_cksum(m, IPPROTO_UDP, off, ulen) != 0) { UDPSTAT_INC(udps_badsum); goto badunlocked; } /* * Construct sockaddr format source address. */ init_sin6(&fromsa, m); fromsa.sin6_port = uh->uh_sport; - INP_INFO_RLOCK(&V_udbinfo); if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { struct inpcb *last; struct ip6_moptions *imo; + INP_INFO_RLOCK(&V_udbinfo); /* * In the event that laddr should be set to the link-local * address (this happens in RIPng), the multicast address * specified in the received packet will not match laddr. To * handle this situation, matching is relaxed if the * receiving interface is the same as one specified in the * socket and if the destination multicast address matches * one of the multicast groups specified in the socket. */ /* * KAME note: traditionally we dropped udpiphdr from mbuf * here. We need udphdr for IPsec processing so we do that * later. */ last = NULL; LIST_FOREACH(inp, &V_udb, inp_list) { if ((inp->inp_vflag & INP_IPV6) == 0) continue; if (inp->inp_lport != uh->uh_dport) continue; if (inp->inp_fport != 0 && inp->inp_fport != uh->uh_sport) continue; if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) { if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &ip6->ip6_dst)) continue; } if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) { if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &ip6->ip6_src) || inp->inp_fport != uh->uh_sport) continue; } /* + * XXXRW: Because we weren't holding either the inpcb + * or the hash lock when we checked for a match + * before, we should probably recheck now that the + * inpcb lock is (supposed to be) held. + */ + + /* * Handle socket delivery policy for any-source * and source-specific multicast. [RFC3678] */ imo = inp->in6p_moptions; if (imo && IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { struct sockaddr_in6 mcaddr; int blocked; INP_RLOCK(inp); bzero(&mcaddr, sizeof(struct sockaddr_in6)); mcaddr.sin6_len = sizeof(struct sockaddr_in6); mcaddr.sin6_family = AF_INET6; mcaddr.sin6_addr = ip6->ip6_dst; blocked = im6o_mc_filter(imo, ifp, (struct sockaddr *)&mcaddr, (struct sockaddr *)&fromsa); if (blocked != MCAST_PASS) { if (blocked == MCAST_NOTGMEMBER) IP6STAT_INC(ip6s_notmember); if (blocked == MCAST_NOTSMEMBER || blocked == MCAST_MUTED) UDPSTAT_INC(udps_filtermcast); INP_RUNLOCK(inp); /* XXX */ continue; } INP_RUNLOCK(inp); } if (last != NULL) { struct mbuf *n; if ((n = m_copy(m, 0, M_COPYALL)) != NULL) { INP_RLOCK(last); up = intoudpcb(last); if (up->u_tun_func == NULL) { udp6_append(last, n, off, &fromsa); } else { /* * Engage the tunneling * protocol we will have to * leave the info_lock up, * since we are hunting * through multiple UDP's. * */ (*up->u_tun_func)(n, off, last); } INP_RUNLOCK(last); } } last = inp; /* * Don't look for additional matches if this one does * not have either the SO_REUSEPORT or SO_REUSEADDR * socket options set. This heuristic avoids * searching through all pcbs in the common case of a * non-shared port. It assumes that an application * will never clear these options after setting them. */ if ((last->inp_socket->so_options & (SO_REUSEPORT|SO_REUSEADDR)) == 0) break; } if (last == NULL) { /* * No matching pcb found; discard datagram. (No need * to send an ICMP Port Unreachable for a broadcast * or multicast datgram.) */ UDPSTAT_INC(udps_noport); UDPSTAT_INC(udps_noportmcast); goto badheadlocked; } INP_RLOCK(last); INP_INFO_RUNLOCK(&V_udbinfo); up = intoudpcb(last); if (up->u_tun_func == NULL) { udp6_append(last, m, off, &fromsa); } else { /* * Engage the tunneling protocol. */ (*up->u_tun_func)(m, off, last); } INP_RUNLOCK(last); return (IPPROTO_DONE); } /* * Locate pcb for datagram. */ - inp = in6_pcblookup_hash(&V_udbinfo, &ip6->ip6_src, uh->uh_sport, - &ip6->ip6_dst, uh->uh_dport, 1, m->m_pkthdr.rcvif); + inp = in6_pcblookup(&V_udbinfo, &ip6->ip6_src, uh->uh_sport, + &ip6->ip6_dst, uh->uh_dport, INPLOOKUP_WILDCARD | + INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif); if (inp == NULL) { if (udp_log_in_vain) { char ip6bufs[INET6_ADDRSTRLEN]; char ip6bufd[INET6_ADDRSTRLEN]; log(LOG_INFO, "Connection attempt to UDP [%s]:%d from [%s]:%d\n", ip6_sprintf(ip6bufd, &ip6->ip6_dst), ntohs(uh->uh_dport), ip6_sprintf(ip6bufs, &ip6->ip6_src), ntohs(uh->uh_sport)); } UDPSTAT_INC(udps_noport); if (m->m_flags & M_MCAST) { printf("UDP6: M_MCAST is set in a unicast packet.\n"); UDPSTAT_INC(udps_noportmcast); - goto badheadlocked; + goto badunlocked; } - INP_INFO_RUNLOCK(&V_udbinfo); if (V_udp_blackhole) goto badunlocked; if (badport_bandlim(BANDLIM_ICMP6_UNREACH) < 0) goto badunlocked; icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0); return (IPPROTO_DONE); } - INP_RLOCK(inp); - INP_INFO_RUNLOCK(&V_udbinfo); + INP_RLOCK_ASSERT(inp); up = intoudpcb(inp); if (up->u_tun_func == NULL) { udp6_append(inp, m, off, &fromsa); } else { /* * Engage the tunneling protocol. */ (*up->u_tun_func)(m, off, inp); } INP_RUNLOCK(inp); return (IPPROTO_DONE); badheadlocked: INP_INFO_RUNLOCK(&V_udbinfo); badunlocked: if (m) m_freem(m); return (IPPROTO_DONE); } void udp6_ctlinput(int cmd, struct sockaddr *sa, void *d) { struct udphdr uh; struct ip6_hdr *ip6; struct mbuf *m; int off = 0; struct ip6ctlparam *ip6cp = NULL; const struct sockaddr_in6 *sa6_src = NULL; void *cmdarg; struct inpcb *(*notify)(struct inpcb *, int) = udp_notify; struct udp_portonly { u_int16_t uh_sport; u_int16_t uh_dport; } *uhp; if (sa->sa_family != AF_INET6 || sa->sa_len != sizeof(struct sockaddr_in6)) return; if ((unsigned)cmd >= PRC_NCMDS) return; if (PRC_IS_REDIRECT(cmd)) notify = in6_rtchange, d = NULL; else if (cmd == PRC_HOSTDEAD) d = NULL; else if (inet6ctlerrmap[cmd] == 0) return; /* if the parameter is from icmp6, decode it. */ if (d != NULL) { ip6cp = (struct ip6ctlparam *)d; m = ip6cp->ip6c_m; ip6 = ip6cp->ip6c_ip6; off = ip6cp->ip6c_off; cmdarg = ip6cp->ip6c_cmdarg; sa6_src = ip6cp->ip6c_src; } else { m = NULL; ip6 = NULL; cmdarg = NULL; sa6_src = &sa6_any; } if (ip6) { /* * XXX: We assume that when IPV6 is non NULL, * M and OFF are valid. */ /* Check if we can safely examine src and dst ports. */ if (m->m_pkthdr.len < off + sizeof(*uhp)) return; bzero(&uh, sizeof(uh)); m_copydata(m, off, sizeof(*uhp), (caddr_t)&uh); (void) in6_pcbnotify(&V_udbinfo, sa, uh.uh_dport, (struct sockaddr *)ip6cp->ip6c_src, uh.uh_sport, cmd, cmdarg, notify); } else (void) in6_pcbnotify(&V_udbinfo, sa, 0, (const struct sockaddr *)sa6_src, 0, cmd, cmdarg, notify); } static int udp6_getcred(SYSCTL_HANDLER_ARGS) { struct xucred xuc; struct sockaddr_in6 addrs[2]; struct inpcb *inp; int error; error = priv_check(req->td, PRIV_NETINET_GETCRED); if (error) return (error); if (req->newlen != sizeof(addrs)) return (EINVAL); if (req->oldlen != sizeof(struct xucred)) return (EINVAL); error = SYSCTL_IN(req, addrs, sizeof(addrs)); if (error) return (error); if ((error = sa6_embedscope(&addrs[0], V_ip6_use_defzone)) != 0 || (error = sa6_embedscope(&addrs[1], V_ip6_use_defzone)) != 0) { return (error); } - INP_INFO_RLOCK(&V_udbinfo); - inp = in6_pcblookup_hash(&V_udbinfo, &addrs[1].sin6_addr, - addrs[1].sin6_port, &addrs[0].sin6_addr, addrs[0].sin6_port, 1, - NULL); + inp = in6_pcblookup(&V_udbinfo, &addrs[1].sin6_addr, + addrs[1].sin6_port, &addrs[0].sin6_addr, addrs[0].sin6_port, + INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, NULL); if (inp != NULL) { - INP_RLOCK(inp); - INP_INFO_RUNLOCK(&V_udbinfo); + INP_RLOCK_ASSERT(inp); if (inp->inp_socket == NULL) error = ENOENT; if (error == 0) error = cr_canseesocket(req->td->td_ucred, inp->inp_socket); if (error == 0) cru2x(inp->inp_cred, &xuc); INP_RUNLOCK(inp); - } else { - INP_INFO_RUNLOCK(&V_udbinfo); + } else error = ENOENT; - } if (error == 0) error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred)); return (error); } SYSCTL_PROC(_net_inet6_udp6, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW, 0, 0, udp6_getcred, "S,xucred", "Get the xucred of a UDP6 connection"); static int udp6_output(struct inpcb *inp, struct mbuf *m, struct sockaddr *addr6, struct mbuf *control, struct thread *td) { u_int32_t ulen = m->m_pkthdr.len; u_int32_t plen = sizeof(struct udphdr) + ulen; struct ip6_hdr *ip6; struct udphdr *udp6; struct in6_addr *laddr, *faddr, in6a; struct sockaddr_in6 *sin6 = NULL; struct ifnet *oifp = NULL; int scope_ambiguous = 0; u_short fport; int error = 0; struct ip6_pktopts *optp, opt; int af = AF_INET6, hlen = sizeof(struct ip6_hdr); int flags; struct sockaddr_in6 tmp; INP_WLOCK_ASSERT(inp); + INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo); if (addr6) { /* addr6 has been validated in udp6_send(). */ sin6 = (struct sockaddr_in6 *)addr6; /* protect *sin6 from overwrites */ tmp = *sin6; sin6 = &tmp; /* * Application should provide a proper zone ID or the use of * default zone IDs should be enabled. Unfortunately, some * applications do not behave as it should, so we need a * workaround. Even if an appropriate ID is not determined, * we'll see if we can determine the outgoing interface. If we * can, determine the zone ID based on the interface below. */ if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone) scope_ambiguous = 1; if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0) return (error); } if (control) { if ((error = ip6_setpktopts(control, &opt, inp->in6p_outputopts, td->td_ucred, IPPROTO_UDP)) != 0) goto release; optp = &opt; } else optp = inp->in6p_outputopts; if (sin6) { faddr = &sin6->sin6_addr; /* * IPv4 version of udp_output calls in_pcbconnect in this case, * which needs splnet and affects performance. * Since we saw no essential reason for calling in_pcbconnect, * we get rid of such kind of logic, and call in6_selectsrc * and in6_pcbsetport in order to fill in the local address * and the local port. */ if (sin6->sin6_port == 0) { error = EADDRNOTAVAIL; goto release; } if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) { /* how about ::ffff:0.0.0.0 case? */ error = EISCONN; goto release; } fport = sin6->sin6_port; /* allow 0 port */ if (IN6_IS_ADDR_V4MAPPED(faddr)) { if ((inp->inp_flags & IN6P_IPV6_V6ONLY)) { /* * I believe we should explicitly discard the * packet when mapped addresses are disabled, * rather than send the packet as an IPv6 one. * If we chose the latter approach, the packet * might be sent out on the wire based on the * default route, the situation which we'd * probably want to avoid. * (20010421 jinmei@kame.net) */ error = EINVAL; goto release; } if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) && !IN6_IS_ADDR_V4MAPPED(&inp->in6p_laddr)) { /* * when remote addr is an IPv4-mapped address, * local addr should not be an IPv6 address, * since you cannot determine how to map IPv6 * source address to IPv4. */ error = EINVAL; goto release; } af = AF_INET; } if (!IN6_IS_ADDR_V4MAPPED(faddr)) { error = in6_selectsrc(sin6, optp, inp, NULL, td->td_ucred, &oifp, &in6a); if (error) goto release; if (oifp && scope_ambiguous && (error = in6_setscope(&sin6->sin6_addr, oifp, NULL))) { goto release; } laddr = &in6a; } else laddr = &inp->in6p_laddr; /* XXX */ if (laddr == NULL) { if (error == 0) error = EADDRNOTAVAIL; goto release; } if (inp->inp_lport == 0 && (error = in6_pcbsetport(laddr, inp, td->td_ucred)) != 0) { /* Undo an address bind that may have occurred. */ inp->in6p_laddr = in6addr_any; goto release; } } else { if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) { error = ENOTCONN; goto release; } if (IN6_IS_ADDR_V4MAPPED(&inp->in6p_faddr)) { if ((inp->inp_flags & IN6P_IPV6_V6ONLY)) { /* * XXX: this case would happen when the * application sets the V6ONLY flag after * connecting the foreign address. * Such applications should be fixed, * so we bark here. */ log(LOG_INFO, "udp6_output: IPV6_V6ONLY " "option was set for a connected socket\n"); error = EINVAL; goto release; } else af = AF_INET; } laddr = &inp->in6p_laddr; faddr = &inp->in6p_faddr; fport = inp->inp_fport; } if (af == AF_INET) hlen = sizeof(struct ip); /* * Calculate data length and get a mbuf * for UDP and IP6 headers. */ M_PREPEND(m, hlen + sizeof(struct udphdr), M_DONTWAIT); if (m == 0) { error = ENOBUFS; goto release; } /* * Stuff checksum and output datagram. */ udp6 = (struct udphdr *)(mtod(m, caddr_t) + hlen); udp6->uh_sport = inp->inp_lport; /* lport is always set in the PCB */ udp6->uh_dport = fport; if (plen <= 0xffff) udp6->uh_ulen = htons((u_short)plen); else udp6->uh_ulen = 0; udp6->uh_sum = 0; switch (af) { case AF_INET6: ip6 = mtod(m, struct ip6_hdr *); ip6->ip6_flow = inp->inp_flow & IPV6_FLOWINFO_MASK; ip6->ip6_vfc &= ~IPV6_VERSION_MASK; ip6->ip6_vfc |= IPV6_VERSION; #if 0 /* ip6_plen will be filled in ip6_output. */ ip6->ip6_plen = htons((u_short)plen); #endif ip6->ip6_nxt = IPPROTO_UDP; ip6->ip6_hlim = in6_selecthlim(inp, NULL); ip6->ip6_src = *laddr; ip6->ip6_dst = *faddr; if ((udp6->uh_sum = in6_cksum(m, IPPROTO_UDP, sizeof(struct ip6_hdr), plen)) == 0) { udp6->uh_sum = 0xffff; } flags = 0; UDPSTAT_INC(udps_opackets); error = ip6_output(m, optp, NULL, flags, inp->in6p_moptions, NULL, inp); break; case AF_INET: error = EAFNOSUPPORT; goto release; } goto releaseopt; release: m_freem(m); releaseopt: if (control) { ip6_clearpktopts(&opt, -1); m_freem(control); } return (error); } static void udp6_abort(struct socket *so) { struct inpcb *inp; inp = sotoinpcb(so); KASSERT(inp != NULL, ("udp6_abort: inp == NULL")); #ifdef INET if (inp->inp_vflag & INP_IPV4) { struct pr_usrreqs *pru; pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs; (*pru->pru_abort)(so); return; } #endif - INP_INFO_WLOCK(&V_udbinfo); INP_WLOCK(inp); if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) { + INP_HASH_WLOCK(&V_udbinfo); in6_pcbdisconnect(inp); inp->in6p_laddr = in6addr_any; + INP_HASH_WUNLOCK(&V_udbinfo); soisdisconnected(so); } INP_WUNLOCK(inp); - INP_INFO_WUNLOCK(&V_udbinfo); } static int udp6_attach(struct socket *so, int proto, struct thread *td) { struct inpcb *inp; int error; inp = sotoinpcb(so); KASSERT(inp == NULL, ("udp6_attach: inp != NULL")); if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { error = soreserve(so, udp_sendspace, udp_recvspace); if (error) return (error); } INP_INFO_WLOCK(&V_udbinfo); error = in_pcballoc(so, &V_udbinfo); if (error) { INP_INFO_WUNLOCK(&V_udbinfo); return (error); } inp = (struct inpcb *)so->so_pcb; inp->inp_vflag |= INP_IPV6; if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) inp->inp_vflag |= INP_IPV4; inp->in6p_hops = -1; /* use kernel default */ inp->in6p_cksum = -1; /* just to be sure */ /* * XXX: ugly!! * IPv4 TTL initialization is necessary for an IPv6 socket as well, * because the socket may be bound to an IPv6 wildcard address, * which may match an IPv4-mapped IPv6 address. */ inp->inp_ip_ttl = V_ip_defttl; error = udp_newudpcb(inp); if (error) { in_pcbdetach(inp); in_pcbfree(inp); INP_INFO_WUNLOCK(&V_udbinfo); return (error); } INP_WUNLOCK(inp); INP_INFO_WUNLOCK(&V_udbinfo); return (0); } static int udp6_bind(struct socket *so, struct sockaddr *nam, struct thread *td) { struct inpcb *inp; int error; inp = sotoinpcb(so); KASSERT(inp != NULL, ("udp6_bind: inp == NULL")); - INP_INFO_WLOCK(&V_udbinfo); INP_WLOCK(inp); + INP_HASH_WLOCK(&V_udbinfo); inp->inp_vflag &= ~INP_IPV4; inp->inp_vflag |= INP_IPV6; if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) { struct sockaddr_in6 *sin6_p; sin6_p = (struct sockaddr_in6 *)nam; if (IN6_IS_ADDR_UNSPECIFIED(&sin6_p->sin6_addr)) inp->inp_vflag |= INP_IPV4; #ifdef INET else if (IN6_IS_ADDR_V4MAPPED(&sin6_p->sin6_addr)) { struct sockaddr_in sin; in6_sin6_2_sin(&sin, sin6_p); inp->inp_vflag |= INP_IPV4; inp->inp_vflag &= ~INP_IPV6; error = in_pcbbind(inp, (struct sockaddr *)&sin, td->td_ucred); goto out; } #endif } error = in6_pcbbind(inp, nam, td->td_ucred); #ifdef INET out: #endif + INP_HASH_WUNLOCK(&V_udbinfo); INP_WUNLOCK(inp); - INP_INFO_WUNLOCK(&V_udbinfo); return (error); } static void udp6_close(struct socket *so) { struct inpcb *inp; inp = sotoinpcb(so); KASSERT(inp != NULL, ("udp6_close: inp == NULL")); #ifdef INET if (inp->inp_vflag & INP_IPV4) { struct pr_usrreqs *pru; pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs; (*pru->pru_disconnect)(so); return; } #endif - INP_INFO_WLOCK(&V_udbinfo); INP_WLOCK(inp); if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) { + INP_HASH_WLOCK(&V_udbinfo); in6_pcbdisconnect(inp); inp->in6p_laddr = in6addr_any; + INP_HASH_WUNLOCK(&V_udbinfo); soisdisconnected(so); } INP_WUNLOCK(inp); - INP_INFO_WUNLOCK(&V_udbinfo); } static int udp6_connect(struct socket *so, struct sockaddr *nam, struct thread *td) { struct inpcb *inp; struct sockaddr_in6 *sin6; int error; inp = sotoinpcb(so); sin6 = (struct sockaddr_in6 *)nam; KASSERT(inp != NULL, ("udp6_connect: inp == NULL")); - INP_INFO_WLOCK(&V_udbinfo); + /* + * XXXRW: Need to clarify locking of v4/v6 flags. + */ INP_WLOCK(inp); #ifdef INET if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { struct sockaddr_in sin; if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) { error = EINVAL; goto out; } if (inp->inp_faddr.s_addr != INADDR_ANY) { error = EISCONN; goto out; } in6_sin6_2_sin(&sin, sin6); inp->inp_vflag |= INP_IPV4; inp->inp_vflag &= ~INP_IPV6; error = prison_remote_ip4(td->td_ucred, &sin.sin_addr); if (error != 0) goto out; + INP_HASH_WLOCK(&V_udbinfo); error = in_pcbconnect(inp, (struct sockaddr *)&sin, td->td_ucred); + INP_HASH_WUNLOCK(&V_udbinfo); if (error == 0) soisconnected(so); goto out; } #endif if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) { error = EISCONN; goto out; } inp->inp_vflag &= ~INP_IPV4; inp->inp_vflag |= INP_IPV6; error = prison_remote_ip6(td->td_ucred, &sin6->sin6_addr); if (error != 0) goto out; + INP_HASH_WLOCK(&V_udbinfo); error = in6_pcbconnect(inp, nam, td->td_ucred); + INP_HASH_WUNLOCK(&V_udbinfo); if (error == 0) soisconnected(so); out: INP_WUNLOCK(inp); - INP_INFO_WUNLOCK(&V_udbinfo); return (error); } static void udp6_detach(struct socket *so) { struct inpcb *inp; struct udpcb *up; inp = sotoinpcb(so); KASSERT(inp != NULL, ("udp6_detach: inp == NULL")); INP_INFO_WLOCK(&V_udbinfo); INP_WLOCK(inp); up = intoudpcb(inp); KASSERT(up != NULL, ("%s: up == NULL", __func__)); in_pcbdetach(inp); in_pcbfree(inp); INP_INFO_WUNLOCK(&V_udbinfo); udp_discardcb(up); } static int udp6_disconnect(struct socket *so) { struct inpcb *inp; int error; inp = sotoinpcb(so); KASSERT(inp != NULL, ("udp6_disconnect: inp == NULL")); - INP_INFO_WLOCK(&V_udbinfo); - INP_WLOCK(inp); - #ifdef INET if (inp->inp_vflag & INP_IPV4) { struct pr_usrreqs *pru; pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs; - error = (*pru->pru_disconnect)(so); - goto out; + (void)(*pru->pru_disconnect)(so); + return (0); } #endif + INP_WLOCK(inp); + if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) { error = ENOTCONN; goto out; } + INP_HASH_WLOCK(&V_udbinfo); in6_pcbdisconnect(inp); inp->in6p_laddr = in6addr_any; + INP_HASH_WUNLOCK(&V_udbinfo); SOCK_LOCK(so); so->so_state &= ~SS_ISCONNECTED; /* XXX */ SOCK_UNLOCK(so); out: INP_WUNLOCK(inp); - INP_INFO_WUNLOCK(&V_udbinfo); return (0); } static int udp6_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr, struct mbuf *control, struct thread *td) { struct inpcb *inp; int error = 0; inp = sotoinpcb(so); KASSERT(inp != NULL, ("udp6_send: inp == NULL")); - INP_INFO_WLOCK(&V_udbinfo); INP_WLOCK(inp); if (addr) { if (addr->sa_len != sizeof(struct sockaddr_in6)) { error = EINVAL; goto bad; } if (addr->sa_family != AF_INET6) { error = EAFNOSUPPORT; goto bad; } } #ifdef INET if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) { int hasv4addr; struct sockaddr_in6 *sin6 = 0; if (addr == 0) hasv4addr = (inp->inp_vflag & INP_IPV4); else { sin6 = (struct sockaddr_in6 *)addr; hasv4addr = IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) ? 1 : 0; } if (hasv4addr) { struct pr_usrreqs *pru; /* * XXXRW: We release UDP-layer locks before calling * udp_send() in order to avoid recursion. However, * this does mean there is a short window where inp's * fields are unstable. Could this lead to a * potential race in which the factors causing us to * select the UDPv4 output routine are invalidated? */ INP_WUNLOCK(inp); - INP_INFO_WUNLOCK(&V_udbinfo); if (sin6) in6_sin6_2_sin_in_sock(addr); pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs; /* addr will just be freed in sendit(). */ return ((*pru->pru_send)(so, flags, m, addr, control, td)); } } #endif #ifdef MAC mac_inpcb_create_mbuf(inp, m); #endif + INP_HASH_WLOCK(&V_udbinfo); error = udp6_output(inp, m, addr, control, td); + INP_HASH_WUNLOCK(&V_udbinfo); #ifdef INET #endif INP_WUNLOCK(inp); - INP_INFO_WUNLOCK(&V_udbinfo); return (error); bad: INP_WUNLOCK(inp); - INP_INFO_WUNLOCK(&V_udbinfo); m_freem(m); return (error); } struct pr_usrreqs udp6_usrreqs = { .pru_abort = udp6_abort, .pru_attach = udp6_attach, .pru_bind = udp6_bind, .pru_connect = udp6_connect, .pru_control = in6_control, .pru_detach = udp6_detach, .pru_disconnect = udp6_disconnect, .pru_peeraddr = in6_mapped_peeraddr, .pru_send = udp6_send, .pru_shutdown = udp_shutdown, .pru_sockaddr = in6_mapped_sockaddr, .pru_soreceive = soreceive_dgram, .pru_sosend = sosend_dgram, .pru_sosetlabel = in_pcbsosetlabel, .pru_close = udp6_close };