static int ip_fw_chk(struct ip **pip, int hlen, struct ifnet *oif, int ignport, struct mbuf **m) { struct ip_fw_chain *chain; struct ip_fw *rule = NULL; struct ip *ip = *pip; struct ifnet *const rif = (*m)->m_pkthdr.rcvif; u_short offset = (ip->ip_off & IP_OFFMASK); u_short src_port, dst_port; /* * Go down the chain, looking for enlightment */ for (chain=ip_fw_chain.lh_first; chain; chain = chain->chain.le_next) { register struct ip_fw *const f = chain->rule; /* Check direction inbound */ if (!oif && !(f->fw_flg & IP_FW_F_IN)) continue; /* Check direction outbound */ if (oif && !(f->fw_flg & IP_FW_F_OUT)) continue; /* Fragments */ if ((f->fw_flg & IP_FW_F_FRAG) && !(ip->ip_off & IP_OFFMASK)) continue; /* If src-addr doesn't match, not this rule. */ if (((f->fw_flg & IP_FW_F_INVSRC) != 0) ^ ((ip->ip_src.s_addr & f->fw_smsk.s_addr) != f->fw_src.s_addr)) continue; /* If dest-addr doesn't match, not this rule. */ if (((f->fw_flg & IP_FW_F_INVDST) != 0) ^ ((ip->ip_dst.s_addr & f->fw_dmsk.s_addr) != f->fw_dst.s_addr)) continue; /* Interface check */ if ((f->fw_flg & IF_FW_F_VIAHACK) == IF_FW_F_VIAHACK) { struct ifnet *const iface = oif ? oif : rif; /* Backwards compatibility hack for "via" */ if (!iface || !iface_match(iface, &f->fw_in_if, f->fw_flg & IP_FW_F_OIFNAME)) continue; } else { /* Check receive interface */ if ((f->fw_flg & IP_FW_F_IIFACE) && (!rif || !iface_match(rif, &f->fw_in_if, f->fw_flg & IP_FW_F_IIFNAME))) continue; /* Check outgoing interface */ if ((f->fw_flg & IP_FW_F_OIFACE) && (!oif || !iface_match(oif, &f->fw_out_if, f->fw_flg & IP_FW_F_OIFNAME))) continue; } /* Check IP options */ if (f->fw_ipopt != f->fw_ipnopt && !ipopts_match(ip, f)) continue; /* Check protocol; if wildcard, match */ if (f->fw_prot == IPPROTO_IP) goto got_match; /* If different, don't match */ if (ip->ip_p != f->fw_prot) continue; #define PULLUP_TO(len) do { \ if ((*m)->m_len < (len) \ && (*m = m_pullup(*m, (len))) == 0) { \ goto bogusfrag; \ } \ *pip = ip = mtod(*m, struct ip *); \ offset = (ip->ip_off & IP_OFFMASK); \ } while (0) /* Protocol specific checks */ switch (ip->ip_p) { case IPPROTO_TCP: { struct tcphdr *tcp; if (offset == 1) /* cf. RFC 1858 */ goto bogusfrag; if (offset != 0) { /* * TCP flags and ports aren't available in this * packet -- if this rule specified either one, * we consider the rule a non-match. */ if (f->fw_nports != 0 || f->fw_tcpf != f->fw_tcpnf) continue; break; } PULLUP_TO(hlen + 14); tcp = (struct tcphdr *) ((u_long *)ip + ip->ip_hl); if (f->fw_tcpf != f->fw_tcpnf && !tcpflg_match(tcp, f)) continue; src_port = ntohs(tcp->th_sport); dst_port = ntohs(tcp->th_dport); goto check_ports; } case IPPROTO_UDP: { struct udphdr *udp; if (offset != 0) { /* * Port specification is unavailable -- if this * rule specifies a port, we consider the rule * a non-match. */ if (f->fw_nports != 0) continue; break; } PULLUP_TO(hlen + 4); udp = (struct udphdr *) ((u_long *)ip + ip->ip_hl); src_port = ntohs(udp->uh_sport); dst_port = ntohs(udp->uh_dport); check_ports: if (!port_match(&f->fw_pts[0], IP_FW_GETNSRCP(f), src_port, f->fw_flg & IP_FW_F_SRNG)) continue; if (!port_match(&f->fw_pts[IP_FW_GETNSRCP(f)], IP_FW_GETNDSTP(f), dst_port, f->fw_flg & IP_FW_F_DRNG)) continue; break; } case IPPROTO_ICMP: { struct icmp *icmp; if (offset != 0) /* Type isn't valid */ break; PULLUP_TO(hlen + 2); icmp = (struct icmp *) ((u_long *)ip + ip->ip_hl); if (!icmptype_match(icmp, f)) continue; break; } #undef PULLUP_TO bogusfrag: if (fw_verbose) ipfw_report(NULL, ip, rif, oif); goto dropit; } got_match: /* Ignore divert/tee rule if socket port is "ignport" */ switch (f->fw_flg & IP_FW_F_COMMAND) { case IP_FW_F_DIVERT: case IP_FW_F_TEE: if (f->fw_divert_port == ignport) continue; /* ignore this rule */ break; } /* Update statistics */ f->fw_pcnt += 1; f->fw_bcnt += ip->ip_len; f->timestamp = rtems_bsdnet_seconds_since_boot(); /* Log to console if desired */ if ((f->fw_flg & IP_FW_F_PRN) && fw_verbose) ipfw_report(f, ip, rif, oif); /* Take appropriate action */ switch (f->fw_flg & IP_FW_F_COMMAND) { case IP_FW_F_ACCEPT: return(0); case IP_FW_F_COUNT: continue; case IP_FW_F_DIVERT: return(f->fw_divert_port); case IP_FW_F_TEE: /* * XXX someday tee packet here, but beware that you * can't use m_copym() or m_copypacket() because * the divert input routine modifies the mbuf * (and these routines only increment reference * counts in the case of mbuf clusters), so need * to write custom routine. */ continue; case IP_FW_F_SKIPTO: #ifdef DIAGNOSTIC while (chain->chain.le_next && chain->chain.le_next->rule->fw_number < f->fw_skipto_rule) #else while (chain->chain.le_next->rule->fw_number < f->fw_skipto_rule) #endif chain = chain->chain.le_next; continue; } /* Deny/reject this packet using this rule */ rule = f; break; } #ifdef DIAGNOSTIC /* Rule 65535 should always be there and should always match */ if (!chain) panic("ip_fw: chain"); #endif /* * At this point, we're going to drop the packet. * Send a reject notice if all of the following are true: * * - The packet matched a reject rule * - The packet is not an ICMP packet * - The packet is not a multicast or broadcast packet */ if ((rule->fw_flg & IP_FW_F_COMMAND) == IP_FW_F_REJECT && ip->ip_p != IPPROTO_ICMP && !((*m)->m_flags & (M_BCAST|M_MCAST)) && !IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { switch (rule->fw_reject_code) { case IP_FW_REJECT_RST: { struct tcphdr *const tcp = (struct tcphdr *) ((u_long *)ip + ip->ip_hl); struct tcpiphdr ti, *const tip = (struct tcpiphdr *) ip; if (offset != 0 || (tcp->th_flags & TH_RST)) break; ti.ti_i = *((struct ipovly *) ip); ti.ti_t = *tcp; bcopy(&ti, ip, sizeof(ti)); NTOHL(tip->ti_seq); NTOHL(tip->ti_ack); tip->ti_len = ip->ip_len - hlen - (tip->ti_off << 2); if (tcp->th_flags & TH_ACK) { tcp_respond(NULL, tip, *m, (tcp_seq)0, ntohl(tcp->th_ack), TH_RST); } else { if (tcp->th_flags & TH_SYN) tip->ti_len++; tcp_respond(NULL, tip, *m, tip->ti_seq + tip->ti_len, (tcp_seq)0, TH_RST|TH_ACK); } *m = NULL; break; } default: /* Send an ICMP unreachable using code */ icmp_error(*m, ICMP_UNREACH, rule->fw_reject_code, 0L, 0); *m = NULL; break; } } dropit: /* * Finally, drop the packet. */ if (*m) { m_freem(*m); *m = NULL; } return(0); }
static int ip_fw_chk(struct ip **pip, int hlen, struct ifnet *oif, u_int16_t *cookie, struct mbuf **m, struct ip_fw_chain **flow_id) { struct ip_fw_chain *chain; struct ip_fw *rule = NULL; struct ip *ip = NULL ; struct ifnet *const rif = (*m)->m_pkthdr.rcvif; u_short offset ; u_short src_port, dst_port; #ifdef IPFW_DIVERT_RESTART u_int16_t skipto = *cookie; #else u_int16_t ignport = ntohs(*cookie); #endif if (pip) { /* normal ip packet */ ip = *pip; offset = (ip->ip_off & IP_OFFMASK); } else { /* bridged or non-ip packet */ struct ether_header *eh = mtod(*m, struct ether_header *); switch (ntohs(eh->ether_type)) { case ETHERTYPE_IP : if ((*m)->m_len<sizeof(struct ether_header) + sizeof(struct ip)) goto non_ip ; ip = (struct ip *)(eh + 1 ); if (ip->ip_v != IPVERSION) goto non_ip ; hlen = ip->ip_hl << 2; if (hlen < sizeof(struct ip)) /* minimum header length */ goto non_ip ; if ((*m)->m_len < 14 + hlen + 14) { printf("-- m_len %d, need more...\n", (*m)->m_len); goto non_ip ; } offset = (ip->ip_off & IP_OFFMASK); break ; default : non_ip: ip = NULL ; break ; } } if (*flow_id) { if (fw_one_pass) return 0 ; /* accept if passed first test */ /* * pkt has already been tagged. Look for the next rule * to restart processing */ if ( (chain = (*flow_id)->rule->next_rule_ptr) == NULL ) chain = (*flow_id)->rule->next_rule_ptr = lookup_next_rule(*flow_id) ; if (!chain) goto dropit; } else { chain=LIST_FIRST(&ip_fw_chain); #ifdef IPFW_DIVERT_RESTART if ( skipto ) { /* * If we've been asked to start at a given rule immediatly, * do so. */ if (skipto >= 65535) goto dropit; while (chain && (chain->rule->fw_number <= skipto)) { chain = LIST_NEXT(chain, chain); } if (! chain) goto dropit; } #endif /* IPFW_DIVERT_RESTART */ } *cookie = 0; for (; chain; chain = LIST_NEXT(chain, chain)) { register struct ip_fw *f; again: f = chain->rule; if (oif) { /* Check direction outbound */ if (!(f->fw_flg & IP_FW_F_OUT)) continue; } else { /* Check direction inbound */ if (!(f->fw_flg & IP_FW_F_IN)) continue; } if (ip == NULL ) { /* * do relevant checks for non-ip packets: * after this, only goto got_match or continue */ struct ether_header *eh = mtod(*m, struct ether_header *); int i, h, l ; #if 0 printf("-- ip_fw: rule %d(%d) for %6D <- %6D type 0x%04x\n", f->fw_number, IP_FW_GETNSRCP(f), eh->ether_dhost, ".", eh->ether_shost, ".", ntohs(eh->ether_type) ); #endif /* * make default rule always match or we have a panic */ if (f->fw_number == 65535) goto got_match ; /* * temporary hack: * udp from 0.0.0.0 means this rule applies. * 1 src port is match ether type * 2 src ports (interval) is match ether type * 3 src ports is match ether address */ if (f->fw_src.s_addr != 0 || f->fw_prot != IPPROTO_UDP) continue ; switch (IP_FW_GETNSRCP(f)) { case 1: /* match one type */ if ( /* ( (f->fw_flg & IP_FW_F_INVSRC) != 0) ^ */ ( f->fw_pts[0] == ntohs(eh->ether_type) ) ) { printf("match!\n"); goto got_match ; } default: break ; } continue; } /* Fragments */ if ((f->fw_flg & IP_FW_F_FRAG) && !(ip->ip_off & IP_OFFMASK)) continue; /* If src-addr doesn't match, not this rule. */ if (((f->fw_flg & IP_FW_F_INVSRC) != 0) ^ ((ip->ip_src.s_addr & f->fw_smsk.s_addr) != f->fw_src.s_addr)) continue; /* If dest-addr doesn't match, not this rule. */ if (((f->fw_flg & IP_FW_F_INVDST) != 0) ^ ((ip->ip_dst.s_addr & f->fw_dmsk.s_addr) != f->fw_dst.s_addr)) continue; /* Interface check */ if ((f->fw_flg & IF_FW_F_VIAHACK) == IF_FW_F_VIAHACK) { struct ifnet *const iface = oif ? oif : rif; /* Backwards compatibility hack for "via" */ if (!iface || !iface_match(iface, &f->fw_in_if, f->fw_flg & IP_FW_F_OIFNAME)) continue; } else { /* Check receive interface */ if ((f->fw_flg & IP_FW_F_IIFACE) && (!rif || !iface_match(rif, &f->fw_in_if, f->fw_flg & IP_FW_F_IIFNAME))) continue; /* Check outgoing interface */ if ((f->fw_flg & IP_FW_F_OIFACE) && (!oif || !iface_match(oif, &f->fw_out_if, f->fw_flg & IP_FW_F_OIFNAME))) continue; } /* Check IP options */ if (f->fw_ipopt != f->fw_ipnopt && !ipopts_match(ip, f)) continue; /* Check protocol; if wildcard, match */ if (f->fw_prot == IPPROTO_IP) goto got_match; /* If different, don't match */ if (ip->ip_p != f->fw_prot) continue; #define PULLUP_TO(len) do { \ if ((*m)->m_len < (len) ) { \ if ( (*m = m_pullup(*m, (len))) == 0) \ goto bogusfrag; \ *pip = ip = mtod(*m, struct ip *); \ offset = (ip->ip_off & IP_OFFMASK); \ } \ } while (0) /* Protocol specific checks */ switch (ip->ip_p) { case IPPROTO_TCP: { struct tcphdr *tcp; if (offset == 1) /* cf. RFC 1858 */ goto bogusfrag; if (offset != 0) { /* * TCP flags and ports aren't available in this * packet -- if this rule specified either one, * we consider the rule a non-match. */ if (f->fw_nports != 0 || f->fw_tcpf != f->fw_tcpnf) continue; break; } PULLUP_TO(hlen + 14); tcp = (struct tcphdr *) ((u_long *)ip + ip->ip_hl); if (f->fw_tcpf != f->fw_tcpnf && !tcpflg_match(tcp, f)) continue; src_port = ntohs(tcp->th_sport); dst_port = ntohs(tcp->th_dport); goto check_ports; } case IPPROTO_UDP: { struct udphdr *udp; if (offset != 0) { /* * Port specification is unavailable -- if this * rule specifies a port, we consider the rule * a non-match. */ if (f->fw_nports != 0) continue; break; } PULLUP_TO(hlen + 4); udp = (struct udphdr *) ((u_long *)ip + ip->ip_hl); src_port = ntohs(udp->uh_sport); dst_port = ntohs(udp->uh_dport); check_ports: if (!port_match(&f->fw_pts[0], IP_FW_GETNSRCP(f), src_port, f->fw_flg & IP_FW_F_SRNG)) continue; if (!port_match(&f->fw_pts[IP_FW_GETNSRCP(f)], IP_FW_GETNDSTP(f), dst_port, f->fw_flg & IP_FW_F_DRNG)) continue; break; } case IPPROTO_ICMP: { struct icmp *icmp; if (offset != 0) /* Type isn't valid */ break; PULLUP_TO(hlen + 2); icmp = (struct icmp *) ((u_long *)ip + ip->ip_hl); if (!icmptype_match(icmp, f)) continue; break; } #undef PULLUP_TO bogusfrag: if (fw_verbose) ipfw_report(NULL, ip, rif, oif); goto dropit; } got_match: *flow_id = chain ; /* XXX set flow id */ #ifndef IPFW_DIVERT_RESTART /* Ignore divert/tee rule if socket port is "ignport" */ switch (f->fw_flg & IP_FW_F_COMMAND) { case IP_FW_F_DIVERT: case IP_FW_F_TEE: if (f->fw_divert_port == ignport) continue; /* ignore this rule */ break; } #endif /* IPFW_DIVERT_RESTART */ /* Update statistics */ f->fw_pcnt += 1; /* * note -- bridged-ip packets still have some fields * in network order, including ip_len */ if (ip) { if (pip) f->fw_bcnt += ip->ip_len; else f->fw_bcnt += ntohs(ip->ip_len); } f->timestamp = time.tv_sec; /* Log to console if desired */ if ((f->fw_flg & IP_FW_F_PRN) && fw_verbose) ipfw_report(f, ip, rif, oif); /* Take appropriate action */ switch (f->fw_flg & IP_FW_F_COMMAND) { case IP_FW_F_ACCEPT: return(0); case IP_FW_F_COUNT: continue; #ifdef IPDIVERT case IP_FW_F_DIVERT: #ifdef IPFW_DIVERT_RESTART *cookie = f->fw_number; #else *cookie = htons(f->fw_divert_port); #endif /* IPFW_DIVERT_RESTART */ return(f->fw_divert_port); #endif case IP_FW_F_TEE: /* * XXX someday tee packet here, but beware that you * can't use m_copym() or m_copypacket() because * the divert input routine modifies the mbuf * (and these routines only increment reference * counts in the case of mbuf clusters), so need * to write custom routine. */ continue; case IP_FW_F_SKIPTO: /* XXX check */ if ( f->next_rule_ptr ) chain = f->next_rule_ptr ; else chain = lookup_next_rule(chain) ; if (!chain) goto dropit; goto again ; #ifdef DUMMYNET case IP_FW_F_PIPE: return(f->fw_pipe_nr | 0x10000 ); #endif } /* Deny/reject this packet using this rule */ rule = f; break; } #ifdef DIAGNOSTIC /* Rule 65535 should always be there and should always match */ if (!chain) panic("ip_fw: chain"); #endif /* * At this point, we're going to drop the packet. * Send a reject notice if all of the following are true: * * - The packet matched a reject rule * - The packet is not an ICMP packet, or is an ICMP query packet * - The packet is not a multicast or broadcast packet */ if ((rule->fw_flg & IP_FW_F_COMMAND) == IP_FW_F_REJECT && ip && (ip->ip_p != IPPROTO_ICMP || is_icmp_query(ip)) && !((*m)->m_flags & (M_BCAST|M_MCAST)) && !IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { switch (rule->fw_reject_code) { case IP_FW_REJECT_RST: { struct tcphdr *const tcp = (struct tcphdr *) ((u_long *)ip + ip->ip_hl); struct tcpiphdr ti, *const tip = (struct tcpiphdr *) ip; if (offset != 0 || (tcp->th_flags & TH_RST)) break; ti.ti_i = *((struct ipovly *) ip); ti.ti_t = *tcp; bcopy(&ti, ip, sizeof(ti)); NTOHL(tip->ti_seq); NTOHL(tip->ti_ack); tip->ti_len = ip->ip_len - hlen - (tip->ti_off << 2); if (tcp->th_flags & TH_ACK) { tcp_respond(NULL, tip, *m, (tcp_seq)0, ntohl(tcp->th_ack), TH_RST); } else { if (tcp->th_flags & TH_SYN) tip->ti_len++; tcp_respond(NULL, tip, *m, tip->ti_seq + tip->ti_len, (tcp_seq)0, TH_RST|TH_ACK); } *m = NULL; break; } default: /* Send an ICMP unreachable using code */ icmp_error(*m, ICMP_UNREACH, rule->fw_reject_code, 0L, 0); *m = NULL; break; } } dropit: /* * Finally, drop the packet. */ if (*m) { m_freem(*m); *m = NULL; } return(0); }