X-Git-Url: https://git.saurik.com/apple/xnu.git/blobdiff_plain/8f6c56a50524aa785f7e596d52dddfb331e18961..c3c9b80d004dbbfdf763edeb97968c6997e3b45b:/bsd/netinet/ip_icmp.c diff --git a/bsd/netinet/ip_icmp.c b/bsd/netinet/ip_icmp.c index edefeff81..0fb3d6f75 100644 --- a/bsd/netinet/ip_icmp.c +++ b/bsd/netinet/ip_icmp.c @@ -1,8 +1,8 @@ /* - * Copyright (c) 2000 Apple Computer, Inc. All rights reserved. + * Copyright (c) 2000-2020 Apple Inc. All rights reserved. * * @APPLE_OSREFERENCE_LICENSE_HEADER_START@ - * + * * This file contains Original Code and/or Modifications of Original Code * as defined in and that are subject to the Apple Public Source License * Version 2.0 (the 'License'). You may not use this file except in @@ -11,10 +11,10 @@ * unlawful or unlicensed copies of an Apple operating system, or to * circumvent, violate, or enable the circumvention or violation of, any * terms of an Apple operating system software license agreement. - * + * * Please obtain a copy of the License at * http://www.opensource.apple.com/apsl/ and read it before using this file. - * + * * The Original Code and all software distributed under the License are * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES, @@ -22,7 +22,7 @@ * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT. * Please see the License for the specific language governing rights and * limitations under the License. - * + * * @APPLE_OSREFERENCE_LICENSE_HEADER_END@ */ /* @@ -59,18 +59,28 @@ * * @(#)ip_icmp.c 8.2 (Berkeley) 1/4/94 */ +/* + * NOTICE: This file was modified by SPARTA, Inc. in 2005 to introduce + * support for mandatory and extensible security protections. This notice + * is included in support of clause 2.2 (b) of the Apple Public License, + * Version 2.0. + */ #include #include #include +#include #include #include #include #include #include +#include + #include #include +#include #define _IP_VHL #include @@ -92,17 +102,10 @@ #include #endif -#if defined(NFAITH) && NFAITH > 0 -#include "faith.h" -#include -#endif +#if NECP +#include +#endif /* NECP */ - /* XXX This one should go in sys/mbuf.h. It is used to avoid that - * a firewall-generated packet loops forever through the firewall. - */ -#ifndef M_SKIP_FIREWALL -#define M_SKIP_FIREWALL 0x4000 -#endif /* * ICMP routines: error generation, receive packet processing, and @@ -110,62 +113,77 @@ * host table maintenance routines. */ -static struct icmpstat icmpstat; -SYSCTL_STRUCT(_net_inet_icmp, ICMPCTL_STATS, stats, CTLFLAG_RD, - &icmpstat, icmpstat, ""); +struct icmpstat icmpstat; +SYSCTL_STRUCT(_net_inet_icmp, ICMPCTL_STATS, stats, + CTLFLAG_RD | CTLFLAG_LOCKED, + &icmpstat, icmpstat, ""); + +static int icmpmaskrepl = 0; +SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, + CTLFLAG_RW | CTLFLAG_LOCKED, + &icmpmaskrepl, 0, ""); -static int icmpmaskrepl = 0; -SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_RW, - &icmpmaskrepl, 0, ""); +static int icmptimestamp = 0; +SYSCTL_INT(_net_inet_icmp, ICMPCTL_TIMESTAMP, timestamp, + CTLFLAG_RW | CTLFLAG_LOCKED, + &icmptimestamp, 0, ""); -static int icmptimestamp = 0; -SYSCTL_INT(_net_inet_icmp, ICMPCTL_TIMESTAMP, timestamp, CTLFLAG_RW, - &icmptimestamp, 0, ""); +static int drop_redirect = 1; +SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, + CTLFLAG_RW | CTLFLAG_LOCKED, + &drop_redirect, 0, ""); -static int drop_redirect = 0; -SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_RW, - &drop_redirect, 0, ""); +static int log_redirect = 0; +SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, + CTLFLAG_RW | CTLFLAG_LOCKED, + &log_redirect, 0, ""); -static int log_redirect = 0; -SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_RW, - &log_redirect, 0, ""); +const static int icmp_datalen = 8; + +#if ICMP_BANDLIM +/* Default values in case CONFIG_ICMP_BANDLIM is not defined in the MASTER file */ +#ifndef CONFIG_ICMP_BANDLIM +#if XNU_TARGET_OS_OSX +#define CONFIG_ICMP_BANDLIM 250 +#else /* !XNU_TARGET_OS_OSX */ +#define CONFIG_ICMP_BANDLIM 50 +#endif /* !XNU_TARGET_OS_OSX */ +#endif /* CONFIG_ICMP_BANDLIM */ -#if ICMP_BANDLIM - -/* +/* * ICMP error-response bandwidth limiting sysctl. If not enabled, sysctl * variable content is -1 and read-only. - */ - -static int icmplim = 250; -SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RW, - &icmplim, 0, ""); -#else + */ +static int icmplim = CONFIG_ICMP_BANDLIM; +SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RW | CTLFLAG_LOCKED, + &icmplim, 0, ""); +#else /* ICMP_BANDLIM */ static int icmplim = -1; -SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RD, - &icmplim, 0, ""); - -#endif +SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RD | CTLFLAG_LOCKED, + &icmplim, 0, ""); +#endif /* ICMP_BANDLIM */ + +static int icmplim_random_incr = CONFIG_ICMP_BANDLIM; +SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM_INCR, icmplim_random_incr, CTLFLAG_RW | CTLFLAG_LOCKED, + &icmplim_random_incr, 0, ""); /* * ICMP broadcast echo sysctl */ -static int icmpbmcastecho = 1; -SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_RW, - &icmpbmcastecho, 0, ""); +static int icmpbmcastecho = 1; +SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_RW | CTLFLAG_LOCKED, + &icmpbmcastecho, 0, ""); - -#if ICMPPRINTFS -int icmpprintfs = 0; +#if (DEBUG | DEVELOPMENT) +static int icmpprintfs = 0; +SYSCTL_INT(_net_inet_icmp, OID_AUTO, verbose, CTLFLAG_RW | CTLFLAG_LOCKED, + &icmpprintfs, 0, ""); #endif -static void icmp_reflect(struct mbuf *); -static void icmp_send(struct mbuf *, struct mbuf *); -static int ip_next_mtu(int, int); - -extern struct protosw inetsw[]; +static void icmp_reflect(struct mbuf *); +static void icmp_send(struct mbuf *, struct mbuf *); /* * Generate an error packet of type error @@ -176,145 +194,264 @@ icmp_error( struct mbuf *n, int type, int code, - n_long dest, - struct ifnet *destifp) + u_int32_t dest, + u_int32_t nextmtu) { - register struct ip *oip = mtod(n, struct ip *), *nip; - register unsigned oiplen = IP_VHL_HL(oip->ip_vhl) << 2; - register struct icmp *icp; - register struct mbuf *m; - unsigned icmplen; - -#if ICMPPRINTFS - if (icmpprintfs) - printf("icmp_error(%p, %x, %d)\n", oip, type, code); -#endif - if (type != ICMP_REDIRECT) + struct ip *oip = NULL; + struct ip *nip = NULL; + struct icmp *icp = NULL; + struct mbuf *m = NULL; + u_int32_t oiphlen = 0; + u_int32_t icmplen = 0; + u_int32_t icmpelen = 0; + u_int32_t nlen = 0; + + VERIFY((u_int)type <= ICMP_MAXTYPE); + VERIFY(code <= UINT8_MAX); + + /* Expect 32-bit aligned data pointer on strict-align platforms */ + MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(n); + + if (type != ICMP_REDIRECT) { icmpstat.icps_error++; + } + /* + * Don't send error: + * if not the first fragment of message + * if original packet was a multicast or broadcast packet + * if the old packet protocol was ICMP + * error message, only known informational types. + */ + if (n->m_flags & (M_BCAST | M_MCAST)) { + goto freeit; + } + /* - * Don't send error if not the first fragment of message. - * Don't error if the old packet protocol was ICMP - * error message, only known informational types. + * Drop if IP header plus ICMP_MINLEN bytes are not contiguous + * in first mbuf. */ - if (oip->ip_off &~ (IP_MF|IP_DF)) + if (n->m_len < sizeof(struct ip) + ICMP_MINLEN) { + goto freeit; + } + + oip = mtod(n, struct ip *); + oiphlen = IP_VHL_HL(oip->ip_vhl) << 2; + if (n->m_len < oiphlen + ICMP_MINLEN) { + goto freeit; + } + +#if (DEBUG | DEVELOPMENT) + if (icmpprintfs > 1) { + printf("icmp_error(0x%llx, %x, %d)\n", + (uint64_t)VM_KERNEL_ADDRPERM(oip), type, code); + } +#endif + + if (oip->ip_off & ~(IP_MF | IP_DF)) { goto freeit; + } + if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT && - n->m_len >= oiplen + ICMP_MINLEN && - !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiplen))->icmp_type)) { + n->m_len >= oiphlen + ICMP_MINLEN && + !ICMP_INFOTYPE(((struct icmp *)(void *)((caddr_t)oip + oiphlen))-> + icmp_type)) { icmpstat.icps_oldicmp++; goto freeit; } - /* Don't send error in response to a multicast or broadcast packet */ - if (n->m_flags & (M_BCAST|M_MCAST)) - goto freeit; + /* - * First, formulate icmp message + * Calculate the length to quote from original packet and prevent + * the ICMP mbuf from overflowing. + * Unfortunatly this is non-trivial since ip_forward() + * sends us truncated packets. */ - m = m_gethdr(M_DONTWAIT, MT_HEADER); - if (m == NULL) - goto freeit; + nlen = m_length(n); + if (oip->ip_p == IPPROTO_TCP) { + struct tcphdr *th = NULL; + u_int16_t tcphlen = 0; + + /* + * If the packet got truncated and TCP header + * is not contained in the packet, send out + * standard reply with only IP header as payload + */ + if (oiphlen + sizeof(struct tcphdr) > n->m_len && + n->m_next == NULL) { + goto stdreply; + } + + /* + * Otherwise, pull up to get IP and TCP headers + * together + */ + if (n->m_len < (oiphlen + sizeof(struct tcphdr)) && + (n = m_pullup(n, (oiphlen + sizeof(struct tcphdr)))) == NULL) { + goto freeit; + } + + /* + * Reinit pointers derived from mbuf data pointer + * as things might have moved around with m_pullup + */ + oip = mtod(n, struct ip *); + th = (struct tcphdr *)(void *)((caddr_t)oip + oiphlen); + + if (th != ((struct tcphdr *)P2ROUNDDOWN(th, + sizeof(u_int32_t))) || + ((th->th_off << 2) > UINT16_MAX)) { + goto freeit; + } + tcphlen = (uint16_t)(th->th_off << 2); - if (n->m_flags & M_SKIP_FIREWALL) { - /* set M_SKIP_FIREWALL to skip firewall check, since we're called from firewall */ - m->m_flags |= M_SKIP_FIREWALL; + /* Sanity checks */ + if (tcphlen < sizeof(struct tcphdr)) { + goto freeit; + } + if (oip->ip_len < (oiphlen + tcphlen)) { + goto freeit; + } + if ((oiphlen + tcphlen) > n->m_len && n->m_next == NULL) { + goto stdreply; + } + if (n->m_len < (oiphlen + tcphlen) && + (n = m_pullup(n, (oiphlen + tcphlen))) == NULL) { + goto freeit; + } + + /* + * Reinit pointers derived from mbuf data pointer + * as things might have moved around with m_pullup + */ + oip = mtod(n, struct ip *); + th = (struct tcphdr *)(void *)((caddr_t)oip + oiphlen); + + icmpelen = max(tcphlen, min(icmp_datalen, + (oip->ip_len - oiphlen))); + } else { +stdreply: icmpelen = max(ICMP_MINLEN, min(icmp_datalen, + (oip->ip_len - oiphlen))); } - icmplen = min(oiplen + 8, oip->ip_len); + icmplen = min(oiphlen + icmpelen, nlen); if (icmplen < sizeof(struct ip)) { - printf("icmp_error: bad length\n"); - m_free(m); goto freeit; } - m->m_len = icmplen + ICMP_MINLEN; - MH_ALIGN(m, m->m_len); + + /* + * First, formulate icmp message + * Allocate enough space for the IP header, ICMP header + * and the payload (part of the original message to be sent back). + */ + if (MHLEN > (sizeof(struct ip) + ICMP_MINLEN + icmplen)) { + m = m_gethdr(M_DONTWAIT, MT_HEADER); /* MAC-OK */ + } else { + m = m_getcl(M_DONTWAIT, MT_DATA, M_PKTHDR); + } + + if (m == NULL) { + goto freeit; + } + + /* + * Further refine the payload length to the space + * remaining in mbuf after including the IP header and ICMP + * header. + */ + icmplen = min(icmplen, (u_int)M_TRAILINGSPACE(m) - + (u_int)(sizeof(struct ip) - ICMP_MINLEN)); + m_align(m, ICMP_MINLEN + icmplen); + m->m_len = ICMP_MINLEN + icmplen; /* for ICMP header and data */ + icp = mtod(m, struct icmp *); - if ((u_int)type > ICMP_MAXTYPE) - panic("icmp_error"); icmpstat.icps_outhist[type]++; - icp->icmp_type = type; - if (type == ICMP_REDIRECT) + icp->icmp_type = (u_char)type; + if (type == ICMP_REDIRECT) { icp->icmp_gwaddr.s_addr = dest; - else { + } else { icp->icmp_void = 0; /* * The following assignments assume an overlay with the * zeroed icmp_void field. */ if (type == ICMP_PARAMPROB) { - icp->icmp_pptr = code; + icp->icmp_pptr = (u_char)code; code = 0; } else if (type == ICMP_UNREACH && - code == ICMP_UNREACH_NEEDFRAG && destifp) { - icp->icmp_nextmtu = htons(destifp->if_mtu); + code == ICMP_UNREACH_NEEDFRAG && nextmtu != 0) { + icp->icmp_nextmtu = htons((uint16_t)nextmtu); } } - icp->icmp_code = code; + icp->icmp_code = (u_char)code; + + /* + * Copy icmplen worth of content from original + * mbuf (n) to the new packet after ICMP header. + */ m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip); nip = &icp->icmp_ip; /* * Convert fields to network representation. */ +#if BYTE_ORDER != BIG_ENDIAN HTONS(nip->ip_len); HTONS(nip->ip_off); - +#endif /* - * Now, copy old ip header (without options) - * in front of icmp message. + * Set up ICMP message mbuf and copy old IP header (without options + * in front of ICMP message. */ - if (m->m_data - sizeof(struct ip) < m->m_pktdat) - panic("icmp len"); m->m_data -= sizeof(struct ip); m->m_len += sizeof(struct ip); m->m_pkthdr.len = m->m_len; m->m_pkthdr.rcvif = n->m_pkthdr.rcvif; - m->m_pkthdr.aux = NULL; /* for IPsec */ nip = mtod(m, struct ip *); bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip)); - nip->ip_len = m->m_len; + nip->ip_len = (uint16_t)m->m_len; nip->ip_vhl = IP_VHL_BORING; nip->ip_p = IPPROTO_ICMP; nip->ip_tos = 0; + nip->ip_off = 0; icmp_reflect(m); - freeit: m_freem(n); } -static struct sockaddr_in icmpsrc = { sizeof (struct sockaddr_in), AF_INET }; -static struct sockaddr_in icmpdst = { sizeof (struct sockaddr_in), AF_INET }; -static struct sockaddr_in icmpgw = { sizeof (struct sockaddr_in), AF_INET }; - /* * Process a received ICMP message. */ void -icmp_input(m, hlen) - register struct mbuf *m; - int hlen; +icmp_input(struct mbuf *m, int hlen) { - register struct icmp *icp; - register struct ip *ip = mtod(m, struct ip *); - int icmplen = ip->ip_len; - register int i; + struct sockaddr_in icmpsrc, icmpdst, icmpgw; + struct icmp *icp; + struct ip *ip = mtod(m, struct ip *); + int icmplen; + int i; struct in_ifaddr *ia; - void (*ctlfunc)(int, struct sockaddr *, void *); + void (*ctlfunc)(int, struct sockaddr *, void *, struct ifnet *); int code; - char ipv4str[MAX_IPv4_STR_LEN]; + boolean_t should_log_redirect = false; + + /* Expect 32-bit aligned data pointer on strict-align platforms */ + MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m); + + icmplen = ip->ip_len; /* * Locate icmp structure in mbuf, and check * that not corrupted and of at least minimum length. */ -#if ICMPPRINTFS - if (icmpprintfs) { - char buf[MAX_IPv4_STR_LEN]; - - printf("icmp_input from %s to %s, len %d\n", - inet_ntop(AF_INET, &ip->ip_src, buf, sizeof(buf)), - inet_ntop(AF_INET, &ip->ip_dst, ipv4str, sizeof(ipv4str)), - icmplen); +#if (DEBUG | DEVELOPMENT) + if (icmpprintfs > 2) { + char src_str[MAX_IPv4_STR_LEN]; + char dst_str[MAX_IPv4_STR_LEN]; + + inet_ntop(AF_INET, &ip->ip_src, src_str, sizeof(src_str)); + inet_ntop(AF_INET, &ip->ip_dst, dst_str, sizeof(dst_str)); + printf("%s: from %s to %s, len %d\n", + __func__, src_str, dst_str, icmplen); } #endif if (icmplen < ICMP_MINLEN) { @@ -322,7 +459,7 @@ icmp_input(m, hlen) goto freeit; } i = hlen + min(icmplen, ICMP_ADVLENMIN); - if (m->m_len < i && (m = m_pullup(m, i)) == 0) { + if (m->m_len < i && (m = m_pullup(m, i)) == 0) { icmpstat.icps_tooshort++; return; } @@ -337,187 +474,188 @@ icmp_input(m, hlen) m->m_len += hlen; m->m_data -= hlen; -#if defined(NFAITH) && 0 < NFAITH - if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) { - /* - * Deliver very specific ICMP type only. - */ - switch (icp->icmp_type) { - case ICMP_UNREACH: - case ICMP_TIMXCEED: - break; - default: - goto freeit; - } - } -#endif - -#if ICMPPRINTFS - if (icmpprintfs) +#if (DEBUG | DEVELOPMENT) + if (icmpprintfs > 2) { printf("icmp_input, type %d code %d\n", icp->icmp_type, icp->icmp_code); + } #endif /* * Message type specific processing. */ - if (icp->icmp_type > ICMP_MAXTYPE) + if (icp->icmp_type > ICMP_MAXTYPE) { goto raw; + } + + /* Initialize */ + bzero(&icmpsrc, sizeof(icmpsrc)); + icmpsrc.sin_len = sizeof(struct sockaddr_in); + icmpsrc.sin_family = AF_INET; + bzero(&icmpdst, sizeof(icmpdst)); + icmpdst.sin_len = sizeof(struct sockaddr_in); + icmpdst.sin_family = AF_INET; + bzero(&icmpgw, sizeof(icmpgw)); + icmpgw.sin_len = sizeof(struct sockaddr_in); + icmpgw.sin_family = AF_INET; + icmpstat.icps_inhist[icp->icmp_type]++; code = icp->icmp_code; switch (icp->icmp_type) { - case ICMP_UNREACH: switch (code) { - case ICMP_UNREACH_NET: - case ICMP_UNREACH_HOST: - case ICMP_UNREACH_SRCFAIL: - case ICMP_UNREACH_NET_UNKNOWN: - case ICMP_UNREACH_HOST_UNKNOWN: - case ICMP_UNREACH_ISOLATED: - case ICMP_UNREACH_TOSNET: - case ICMP_UNREACH_TOSHOST: - case ICMP_UNREACH_HOST_PRECEDENCE: - case ICMP_UNREACH_PRECEDENCE_CUTOFF: - code = PRC_UNREACH_NET; - break; + case ICMP_UNREACH_NET: + case ICMP_UNREACH_HOST: + case ICMP_UNREACH_SRCFAIL: + case ICMP_UNREACH_NET_UNKNOWN: + case ICMP_UNREACH_HOST_UNKNOWN: + case ICMP_UNREACH_ISOLATED: + case ICMP_UNREACH_TOSNET: + case ICMP_UNREACH_TOSHOST: + case ICMP_UNREACH_HOST_PRECEDENCE: + case ICMP_UNREACH_PRECEDENCE_CUTOFF: + code = PRC_UNREACH_NET; + break; - case ICMP_UNREACH_NEEDFRAG: - code = PRC_MSGSIZE; - break; + case ICMP_UNREACH_NEEDFRAG: + code = PRC_MSGSIZE; + break; - /* - * RFC 1122, Sections 3.2.2.1 and 4.2.3.9. - * Treat subcodes 2,3 as immediate RST - */ - case ICMP_UNREACH_PROTOCOL: - case ICMP_UNREACH_PORT: - code = PRC_UNREACH_PORT; - break; + /* + * RFC 1122, Sections 3.2.2.1 and 4.2.3.9. + * Treat subcodes 2,3 as immediate RST + */ + case ICMP_UNREACH_PROTOCOL: + case ICMP_UNREACH_PORT: + code = PRC_UNREACH_PORT; + break; - case ICMP_UNREACH_NET_PROHIB: - case ICMP_UNREACH_HOST_PROHIB: - case ICMP_UNREACH_FILTER_PROHIB: - code = PRC_UNREACH_ADMIN_PROHIB; - break; + case ICMP_UNREACH_NET_PROHIB: + case ICMP_UNREACH_HOST_PROHIB: + case ICMP_UNREACH_FILTER_PROHIB: + code = PRC_UNREACH_ADMIN_PROHIB; + break; - default: - goto badcode; + default: + goto badcode; } goto deliver; case ICMP_TIMXCEED: - if (code > 1) + if (code > 1) { goto badcode; + } code += PRC_TIMXCEED_INTRANS; goto deliver; case ICMP_PARAMPROB: - if (code > 1) + if (code > 1) { goto badcode; + } code = PRC_PARAMPROB; goto deliver; case ICMP_SOURCEQUENCH: - if (code) + if (code) { goto badcode; + } code = PRC_QUENCH; - deliver: +deliver: /* * Problem with datagram; advise higher level routines. */ - if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) || - IP_VHL_HL(icp->icmp_ip.ip_vhl) < (sizeof(struct ip) >> 2)) { + if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) + || IP_VHL_HL(icp->icmp_ip.ip_vhl) < + (sizeof(struct ip) >> 2) || + (m = m_pullup(m, hlen + ICMP_ADVLEN(icp))) == NULL) { icmpstat.icps_badlen++; goto freeit; } + + ip = mtod(m, struct ip *); + icp = (struct icmp *)(void *)(mtod(m, uint8_t *) + hlen); + +#if BYTE_ORDER != BIG_ENDIAN NTOHS(icp->icmp_ip.ip_len); +#endif + /* Discard ICMP's in response to multicast packets */ - if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr))) + if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr))) { goto badcode; -#if ICMPPRINTFS - if (icmpprintfs) - printf("deliver to protocol %d\n", icp->icmp_ip.ip_p); + } +#if (DEBUG | DEVELOPMENT) + if (icmpprintfs > 2) { + printf("deliver to protocol %d\n", + icp->icmp_ip.ip_p); + } #endif icmpsrc.sin_addr = icp->icmp_ip.ip_dst; -#if 1 - /* - * MTU discovery: - * If we got a needfrag and there is a host route to the - * original destination, and the MTU is not locked, then - * set the MTU in the route to the suggested new value - * (if given) and then notify as usual. The ULPs will - * notice that the MTU has changed and adapt accordingly. - * If no new MTU was suggested, then we guess a new one - * less than the current value. If the new MTU is - * unreasonably small (defined by sysctl tcp_minmss), then - * we reset the MTU to the interface value and enable the - * lock bit, indicating that we are no longer doing MTU - * discovery. - */ - if (code == PRC_MSGSIZE) { - struct rtentry *rt; - int mtu; - - rt = rtalloc1((struct sockaddr *)&icmpsrc, 0, - RTF_CLONING | RTF_PRCLONING); - if (rt && (rt->rt_flags & RTF_HOST) - && !(rt->rt_rmx.rmx_locks & RTV_MTU)) { - mtu = ntohs(icp->icmp_nextmtu); - if (!mtu) - mtu = ip_next_mtu(rt->rt_rmx.rmx_mtu, - 1); -#if DEBUG_MTUDISC - printf("MTU for %s reduced to %d\n", - inet_ntop(AF_INET, &icmpsrc.sin_addr, ipv4str, - sizeof(ipv4str)), - mtu); -#endif - if (mtu < max(296, (tcp_minmss + sizeof(struct tcpiphdr)))) { - /* rt->rt_rmx.rmx_mtu = - rt->rt_ifp->if_mtu; */ - rt->rt_rmx.rmx_locks |= RTV_MTU; - } else if (rt->rt_rmx.rmx_mtu > mtu) { - rt->rt_rmx.rmx_mtu = mtu; - } - } - if (rt) - rtfree(rt); - } -#endif /* - * XXX if the packet contains [IPv4 AH TCP], we can't make a + * if the packet contains [IPv4 AH TCP], we can't make a * notification to TCP layer. */ ctlfunc = ip_protox[icp->icmp_ip.ip_p]->pr_ctlinput; - if (ctlfunc) + + if (ctlfunc) { + LCK_MTX_ASSERT(inet_domain_mutex, LCK_MTX_ASSERT_OWNED); + + lck_mtx_unlock(inet_domain_mutex); + (*ctlfunc)(code, (struct sockaddr *)&icmpsrc, - (void *)&icp->icmp_ip); + (void *)&icp->icmp_ip, m->m_pkthdr.rcvif); + + lck_mtx_lock(inet_domain_mutex); + } break; - badcode: +badcode: icmpstat.icps_badcode++; break; case ICMP_ECHO: - if (!icmpbmcastecho - && (m->m_flags & (M_MCAST | M_BCAST)) != 0) { + if ((m->m_flags & (M_MCAST | M_BCAST))) { + if (icmpbmcastecho == 0) { + icmpstat.icps_bmcastecho++; + break; + } + } + + /* + * rdar://18644769 + * Do not reply when the destination is link local multicast or broadcast + * and the source is not from a directly connected subnet + */ + if ((IN_LOCAL_GROUP(ntohl(ip->ip_dst.s_addr)) || + in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) && + in_localaddr(ip->ip_src) == 0) { icmpstat.icps_bmcastecho++; +#if (DEBUG | DEVELOPMENT) + if (icmpprintfs > 0) { + char src_str[MAX_IPv4_STR_LEN]; + char dst_str[MAX_IPv4_STR_LEN]; + + inet_ntop(AF_INET, &ip->ip_src, src_str, sizeof(src_str)); + inet_ntop(AF_INET, &ip->ip_dst, dst_str, sizeof(dst_str)); + printf("%s: non local (B|M)CAST %s to %s, len %d\n", + __func__, src_str, dst_str, icmplen); + } +#endif break; } + icp->icmp_type = ICMP_ECHOREPLY; #if ICMP_BANDLIM - if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0) + if (badport_bandlim(BANDLIM_ICMP_ECHO)) { goto freeit; - else + } else #endif - goto reflect; + goto reflect; case ICMP_TSTAMP: - - if (icmptimestamp == 0) + if (icmptimestamp == 0) { break; + } if (!icmpbmcastecho && (m->m_flags & (M_MCAST | M_BCAST)) != 0) { @@ -530,26 +668,26 @@ icmp_input(m, hlen) } icp->icmp_type = ICMP_TSTAMPREPLY; icp->icmp_rtime = iptime(); - icp->icmp_ttime = icp->icmp_rtime; /* bogus, do later! */ + icp->icmp_ttime = icp->icmp_rtime; /* bogus, do later! */ #if ICMP_BANDLIM - if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0) + if (badport_bandlim(BANDLIM_ICMP_TSTAMP)) { goto freeit; - else + } else #endif - goto reflect; + goto reflect; case ICMP_MASKREQ: -#define satosin(sa) ((struct sockaddr_in *)(sa)) - if (icmpmaskrepl == 0) + if (icmpmaskrepl == 0) { break; + } /* * We are not able to respond with all ones broadcast * unless we receive it over a point-to-point interface. */ - if (icmplen < ICMP_MASKLEN) + if (icmplen < ICMP_MASKLEN) { break; + } switch (ip->ip_dst.s_addr) { - case INADDR_BROADCAST: case INADDR_ANY: icmpdst.sin_addr = ip->ip_src; @@ -559,55 +697,53 @@ icmp_input(m, hlen) icmpdst.sin_addr = ip->ip_dst; } ia = (struct in_ifaddr *)ifaof_ifpforaddr( - (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif); - if (ia == 0) + (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif); + if (ia == 0) { break; + } + IFA_LOCK(&ia->ia_ifa); if (ia->ia_ifp == 0) { - ifafree(&ia->ia_ifa); - ia = 0; + IFA_UNLOCK(&ia->ia_ifa); + IFA_REMREF(&ia->ia_ifa); + ia = NULL; break; } icp->icmp_type = ICMP_MASKREPLY; icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr; if (ip->ip_src.s_addr == 0) { - if (ia->ia_ifp->if_flags & IFF_BROADCAST) - ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr; - else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT) - ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr; + if (ia->ia_ifp->if_flags & IFF_BROADCAST) { + ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr; + } else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT) { + ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr; + } } - ifafree(&ia->ia_ifa); + IFA_UNLOCK(&ia->ia_ifa); + IFA_REMREF(&ia->ia_ifa); reflect: - ip->ip_len += hlen; /* since ip_input deducts this */ + ip->ip_len += hlen; /* since ip_input deducts this */ icmpstat.icps_reflect++; icmpstat.icps_outhist[icp->icmp_type]++; icmp_reflect(m); return; case ICMP_REDIRECT: - if (log_redirect) { - u_long src, dst, gw; - - src = ntohl(ip->ip_src.s_addr); - dst = ntohl(icp->icmp_ip.ip_dst.s_addr); - gw = ntohl(icp->icmp_gwaddr.s_addr); - printf("icmp redirect from %d.%d.%d.%d: " - "%d.%d.%d.%d => %d.%d.%d.%d\n", - (int)(src >> 24), (int)((src >> 16) & 0xff), - (int)((src >> 8) & 0xff), (int)(src & 0xff), - (int)(dst >> 24), (int)((dst >> 16) & 0xff), - (int)((dst >> 8) & 0xff), (int)(dst & 0xff), - (int)(gw >> 24), (int)((gw >> 16) & 0xff), - (int)((gw >> 8) & 0xff), (int)(gw & 0xff)); - } - if (drop_redirect) + if (drop_redirect) { break; - if (code > 3) + } + if (code > 3) { goto badcode; + } if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) || IP_VHL_HL(icp->icmp_ip.ip_vhl) < (sizeof(struct ip) >> 2)) { icmpstat.icps_badlen++; break; } + +#if (DEBUG | DEVELOPMENT) + should_log_redirect = log_redirect || (icmpprintfs > 0); +#else + should_log_redirect = log_redirect; +#endif /* * Short circuit routing redirects to force * immediate change in the kernel's routing @@ -617,21 +753,22 @@ reflect: */ icmpgw.sin_addr = ip->ip_src; icmpdst.sin_addr = icp->icmp_gwaddr; -#if ICMPPRINTFS - if (icmpprintfs) { - char buf[MAX_IPv4_STR_LEN]; - - printf("redirect dst %s to %s\n", - inet_ntop(AF_INET, &icp->icmp_ip.ip_dst, buf, sizeof(buf)), - inet_ntop(AF_INET, &icp->icmp_gwaddr, ipv4str, - sizeof(ipv4str))); + + if (should_log_redirect) { + char src_str[MAX_IPv4_STR_LEN]; + char dst_str[MAX_IPv4_STR_LEN]; + char gw_str[MAX_IPv4_STR_LEN]; + + inet_ntop(AF_INET, &ip->ip_src, src_str, sizeof(src_str)); + inet_ntop(AF_INET, &icp->icmp_ip.ip_dst, dst_str, sizeof(dst_str)); + inet_ntop(AF_INET, &icp->icmp_gwaddr, gw_str, sizeof(gw_str)); + printf("%s: redirect dst %s to %s from %s\n", __func__, + dst_str, gw_str, src_str); } -#endif icmpsrc.sin_addr = icp->icmp_ip.ip_dst; - rtredirect((struct sockaddr *)&icmpsrc, - (struct sockaddr *)&icmpdst, - (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST, - (struct sockaddr *)&icmpgw, (struct rtentry **)0); + rtredirect(m->m_pkthdr.rcvif, (struct sockaddr *)&icmpsrc, + (struct sockaddr *)&icmpdst, NULL, RTF_GATEWAY | RTF_HOST, + (struct sockaddr *)&icmpgw, NULL); pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc); #if IPSEC key_sa_routechange((struct sockaddr *)&icmpsrc); @@ -664,20 +801,20 @@ freeit: * Reflect the ip packet back to the source */ static void -icmp_reflect(m) - struct mbuf *m; +icmp_reflect(struct mbuf *m) { - register struct ip *ip = mtod(m, struct ip *); - register struct in_ifaddr *ia; + struct ip *ip = mtod(m, struct ip *); + struct sockaddr_in icmpdst; + struct in_ifaddr *ia; struct in_addr t; - struct mbuf *opts = 0; + struct mbuf *opts = NULL; int optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip); if (!in_canforward(ip->ip_src) && ((ntohl(ip->ip_src.s_addr) & IN_CLASSA_NET) != - (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) { - m_freem(m); /* Bad return address */ - goto done; /* Ip_output() will check for broadcast */ + (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) { + m_freem(m); /* Bad return address */ + goto done; /* Ip_output() will check for broadcast */ } t = ip->ip_dst; ip->ip_dst = ip->ip_src; @@ -687,42 +824,68 @@ icmp_reflect(m) * or anonymous), use the address which corresponds * to the incoming interface. */ - lck_mtx_lock(rt_mtx); + lck_rw_lock_shared(in_ifaddr_rwlock); + TAILQ_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) { + IFA_LOCK(&ia->ia_ifa); + if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) { + IFA_ADDREF_LOCKED(&ia->ia_ifa); + IFA_UNLOCK(&ia->ia_ifa); + goto match; + } + IFA_UNLOCK(&ia->ia_ifa); + } + /* + * Slow path; check for broadcast addresses. Find a source + * IP address to use when replying to the broadcast request; + * let IP handle the source interface selection work. + */ for (ia = in_ifaddrhead.tqh_first; ia; ia = ia->ia_link.tqe_next) { - if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) - break; + IFA_LOCK(&ia->ia_ifa); if (ia->ia_ifp && (ia->ia_ifp->if_flags & IFF_BROADCAST) && - t.s_addr == satosin(&ia->ia_broadaddr)->sin_addr.s_addr) + t.s_addr == satosin(&ia->ia_broadaddr)->sin_addr.s_addr) { + IFA_ADDREF_LOCKED(&ia->ia_ifa); + IFA_UNLOCK(&ia->ia_ifa); break; + } + IFA_UNLOCK(&ia->ia_ifa); } - if (ia) - ifaref(&ia->ia_ifa); +match: + lck_rw_done(in_ifaddr_rwlock); + + /* Initialize */ + bzero(&icmpdst, sizeof(icmpdst)); + icmpdst.sin_len = sizeof(struct sockaddr_in); + icmpdst.sin_family = AF_INET; icmpdst.sin_addr = t; - if ((ia == (struct in_ifaddr *)0) && m->m_pkthdr.rcvif) + if ((ia == (struct in_ifaddr *)0) && m->m_pkthdr.rcvif) { ia = (struct in_ifaddr *)ifaof_ifpforaddr( (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif); + } /* * The following happens if the packet was not addressed to us, * and was received on an interface with no IP address. */ if (ia == (struct in_ifaddr *)0) { + lck_rw_lock_shared(in_ifaddr_rwlock); ia = in_ifaddrhead.tqh_first; if (ia == (struct in_ifaddr *)0) {/* no address yet, bail out */ + lck_rw_done(in_ifaddr_rwlock); m_freem(m); - lck_mtx_unlock(rt_mtx); goto done; } - ifaref(&ia->ia_ifa); + IFA_ADDREF(&ia->ia_ifa); + lck_rw_done(in_ifaddr_rwlock); } - lck_mtx_unlock(rt_mtx); + IFA_LOCK_SPIN(&ia->ia_ifa); t = IA_SIN(ia)->sin_addr; + IFA_UNLOCK(&ia->ia_ifa); ip->ip_src = t; - ip->ip_ttl = ip_defttl; - ifafree(&ia->ia_ifa); + ip->ip_ttl = (u_char)ip_defttl; + IFA_REMREF(&ia->ia_ifa); ia = NULL; if (optlen > 0) { - register u_char *cp; + u_char *cp; int opt, cnt; u_int len; @@ -732,52 +895,57 @@ icmp_reflect(m) */ cp = (u_char *) (ip + 1); if ((opts = ip_srcroute()) == 0 && - (opts = m_gethdr(M_DONTWAIT, MT_HEADER))) { + (opts = m_gethdr(M_DONTWAIT, MT_HEADER))) { /* MAC-OK */ opts->m_len = sizeof(struct in_addr); mtod(opts, struct in_addr *)->s_addr = 0; } if (opts) { -#if ICMPPRINTFS - if (icmpprintfs) - printf("icmp_reflect optlen %d rt %d => ", - optlen, opts->m_len); +#if (DEBUG | DEVELOPMENT) + if (icmpprintfs > 1) { + printf("icmp_reflect optlen %d rt %d => ", + optlen, opts->m_len); + } #endif - for (cnt = optlen; cnt > 0; cnt -= len, cp += len) { - opt = cp[IPOPT_OPTVAL]; - if (opt == IPOPT_EOL) - break; - if (opt == IPOPT_NOP) - len = 1; - else { - if (cnt < IPOPT_OLEN + sizeof(*cp)) - break; - len = cp[IPOPT_OLEN]; - if (len < IPOPT_OLEN + sizeof(*cp) || - len > cnt) - break; - } - /* - * Should check for overflow, but it "can't happen" - */ - if (opt == IPOPT_RR || opt == IPOPT_TS || - opt == IPOPT_SECURITY) { - bcopy((caddr_t)cp, - mtod(opts, caddr_t) + opts->m_len, len); - opts->m_len += len; - } - } - /* Terminate & pad, if necessary */ - cnt = opts->m_len % 4; - if (cnt) { - for (; cnt < 4; cnt++) { - *(mtod(opts, caddr_t) + opts->m_len) = - IPOPT_EOL; - opts->m_len++; - } - } -#if ICMPPRINTFS - if (icmpprintfs) - printf("%d\n", opts->m_len); + for (cnt = optlen; cnt > 0; cnt -= len, cp += len) { + opt = cp[IPOPT_OPTVAL]; + if (opt == IPOPT_EOL) { + break; + } + if (opt == IPOPT_NOP) { + len = 1; + } else { + if (cnt < IPOPT_OLEN + sizeof(*cp)) { + break; + } + len = cp[IPOPT_OLEN]; + if (len < IPOPT_OLEN + sizeof(*cp) || + len > cnt) { + break; + } + } + /* + * Should check for overflow, but it "can't happen" + */ + if (opt == IPOPT_RR || opt == IPOPT_TS || + opt == IPOPT_SECURITY) { + bcopy((caddr_t)cp, + mtod(opts, caddr_t) + opts->m_len, len); + opts->m_len += len; + } + } + /* Terminate & pad, if necessary */ + cnt = opts->m_len % 4; + if (cnt) { + for (; cnt < 4; cnt++) { + *(mtod(opts, caddr_t) + opts->m_len) = + IPOPT_EOL; + opts->m_len++; + } + } +#if (DEBUG | DEVELOPMENT) + if (icmpprintfs > 1) { + printf("%d\n", opts->m_len); + } #endif } /* @@ -787,17 +955,19 @@ icmp_reflect(m) ip->ip_len -= optlen; ip->ip_vhl = IP_VHL_BORING; m->m_len -= optlen; - if (m->m_flags & M_PKTHDR) + if (m->m_flags & M_PKTHDR) { m->m_pkthdr.len -= optlen; + } optlen += sizeof(struct ip); bcopy((caddr_t)ip + optlen, (caddr_t)(ip + 1), - (unsigned)(m->m_len - sizeof(struct ip))); + (unsigned)(m->m_len - sizeof(struct ip))); } - m->m_flags &= ~(M_BCAST|M_MCAST); + m->m_flags &= ~(M_BCAST | M_MCAST); icmp_send(m, opts); done: - if (opts) + if (opts) { (void)m_free(opts); + } } /* @@ -805,15 +975,24 @@ done: * after supplying a checksum. */ static void -icmp_send(m, opts) - register struct mbuf *m; - struct mbuf *opts; +icmp_send(struct mbuf *m, struct mbuf *opts) { - register struct ip *ip = mtod(m, struct ip *); - register int hlen; - register struct icmp *icp; + struct ip *ip = mtod(m, struct ip *); + int hlen; + struct icmp *icp; struct route ro; - char ipv4str[MAX_IPv4_STR_LEN]; + struct ip_out_args ipoa; + + bzero(&ipoa, sizeof(ipoa)); + ipoa.ipoa_boundif = IFSCOPE_NONE; + ipoa.ipoa_flags = IPOAF_SELECT_SRCIF | IPOAF_BOUND_SRCADDR; + ipoa.ipoa_sotc = SO_TC_UNSPEC; + ipoa.ipoa_netsvctype = _NET_SERVICE_TYPE_UNSPEC; + + if (!(m->m_pkthdr.pkt_flags & PKTF_LOOP) && m->m_pkthdr.rcvif != NULL) { + ipoa.ipoa_boundif = m->m_pkthdr.rcvif->if_index; + ipoa.ipoa_flags |= IPOAF_BOUND_IF; + } hlen = IP_VHL_HL(ip->ip_vhl) << 2; m->m_data += hlen; @@ -823,34 +1002,33 @@ icmp_send(m, opts) icp->icmp_cksum = in_cksum(m, ip->ip_len - hlen); m->m_data -= hlen; m->m_len += hlen; - m->m_pkthdr.rcvif = 0; - m->m_pkthdr.aux = NULL; + m->m_pkthdr.rcvif = NULL; m->m_pkthdr.csum_data = 0; m->m_pkthdr.csum_flags = 0; -#if ICMPPRINTFS - if (icmpprintfs) { - char buf[MAX_IPv4_STR_LEN]; - - printf("icmp_send dst %s src %s\n", - inet_ntop(AF_INET, &ip->ip_dst, buf, sizeof(buf)), - inet_ntop(AF_INET, &ip->ip_src, ipv4str, sizeof(ipv4str))); +#if (DEBUG | DEVELOPMENT) + if (icmpprintfs > 2) { + char src_str[MAX_IPv4_STR_LEN]; + char dst_str[MAX_IPv4_STR_LEN]; + + inet_ntop(AF_INET, &ip->ip_src, src_str, sizeof(src_str)); + inet_ntop(AF_INET, &ip->ip_dst, dst_str, sizeof(dst_str)); + printf("%s: dst %s src %s\n", __func__, dst_str, src_str); } #endif bzero(&ro, sizeof ro); - (void) ip_output(m, opts, &ro, 0, NULL); - if (ro.ro_rt) - rtfree(ro.ro_rt); + (void) ip_output(m, opts, &ro, IP_OUTARGS, NULL, &ipoa); + ROUTE_RELEASE(&ro); } -n_time -iptime() +u_int32_t +iptime(void) { struct timeval atv; - u_long t; + u_int32_t t; - microtime(&atv); - t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000; - return (htonl(t)); + getmicrotime(&atv); + t = (atv.tv_sec % (24 * 60 * 60)) * 1000 + atv.tv_usec / 1000; + return htonl(t); } #if 1 @@ -859,10 +1037,8 @@ iptime() * given current value MTU. If DIR is less than zero, a larger plateau * is returned; otherwise, a smaller value is returned. */ -static int -ip_next_mtu(mtu, dir) - int mtu; - int dir; +int +ip_next_mtu(int mtu, int dir) { static int mtutab[] = { 65535, 32000, 17914, 8166, 4352, 2002, 1492, 1006, 508, 296, @@ -871,8 +1047,9 @@ ip_next_mtu(mtu, dir) int i; for (i = 0; i < (sizeof mtutab) / (sizeof mtutab[0]); i++) { - if (mtu >= mtutab[i]) + if (mtu >= mtutab[i]) { break; + } } if (dir < 0) { @@ -884,7 +1061,7 @@ ip_next_mtu(mtu, dir) } else { if (mtutab[i] == 0) { return 0; - } else if(mtu > mtutab[i]) { + } else if (mtu > mtutab[i]) { return mtutab[i]; } else { return mtutab[i + 1]; @@ -897,11 +1074,8 @@ ip_next_mtu(mtu, dir) /* * badport_bandlim() - check for ICMP bandwidth limit - * - * Return 0 if it is ok to send an ICMP error response, -1 if we have - * hit our bandwidth limit and it is not ok. - * - * If icmplim is <= 0, the feature is disabled and 0 is returned. + * Returns false when it is ok to send ICMP error and true to limit sending + * of ICMP error. * * For now we separate the TCP and UDP subsystems w/ different 'which' * values. We may eventually remove this separation (and simplify the @@ -910,63 +1084,75 @@ ip_next_mtu(mtu, dir) * Note that the printing of the error message is delayed so we can * properly print the icmp error rate that the system was trying to do * (i.e. 22000/100 pps, etc...). This can cause long delays in printing - * the 'final' error, but it doesn't make sense to solve the printing + * the 'final' error, but it doesn't make sense to solve the printing * delay with more complex code. */ -int +boolean_t badport_bandlim(int which) { - static struct timeval lticks[BANDLIM_MAX + 1]; + static uint64_t lticks[BANDLIM_MAX + 1]; static int lpackets[BANDLIM_MAX + 1]; - struct timeval time; - int secs; - + uint64_t time; + uint64_t secs; + static boolean_t is_initialized = FALSE; + static int icmplim_random; const char *bandlimittype[] = { "Limiting icmp unreach response", "Limiting icmp ping response", "Limiting icmp tstamp response", "Limiting closed port RST response", "Limiting open port RST response" - }; + }; - /* - * Return ok status if feature disabled or argument out of - * ranage. - */ + /* Return ok status if feature disabled or argument out of range. */ - if (icmplim <= 0 || which > BANDLIM_MAX || which < 0) - return(0); + if (icmplim <= 0 || which > BANDLIM_MAX || which < 0) { + return false; + } - getmicrouptime(&time); + if (is_initialized == FALSE) { + if (icmplim_random_incr > 0 && + icmplim <= INT32_MAX - (icmplim_random_incr + 1)) { + icmplim_random = icmplim + (random() % icmplim_random_incr) + 1; + } + is_initialized = TRUE; + } + + time = net_uptime(); + secs = time - lticks[which]; - secs = time.tv_sec - lticks[which].tv_sec ; - /* * reset stats when cumulative delta exceeds one second. */ - if ((secs > 1) || (secs == 1 && (lticks[which].tv_usec > time.tv_usec))) { - if (lpackets[which] > icmplim) { + if (secs > 1) { + if (lpackets[which] > icmplim_random) { printf("%s from %d to %d packets per second\n", - bandlimittype[which], - lpackets[which], - icmplim - ); + bandlimittype[which], + lpackets[which], + icmplim_random + ); } - lticks[which].tv_sec = time.tv_sec; - lticks[which].tv_usec = time.tv_usec; + lticks[which] = time; lpackets[which] = 0; } /* * bump packet count */ - - if (++lpackets[which] > icmplim) { - return(-1); + if (++lpackets[which] > icmplim_random) { + /* + * After hitting the randomized limit, we further randomize the + * behavior of how we apply rate limitation. + * We rate limit based on probability that increases with the + * increase in lpackets[which] count. + */ + if ((random() % (lpackets[which] - icmplim_random)) != 0) { + return true; + } } - return(0); + return false; } #endif @@ -983,9 +1169,8 @@ badport_bandlim(int which) #include #include -extern struct domain inetdomain; -extern u_long rip_sendspace; -extern u_long rip_recvspace; +extern u_int32_t rip_sendspace; +extern u_int32_t rip_recvspace; extern struct inpcbinfo ripcbinfo; int rip_abort(struct socket *); @@ -1000,37 +1185,46 @@ __private_extern__ int icmp_dgram_attach(struct socket *so, int proto, struct pr __private_extern__ int icmp_dgram_ctloutput(struct socket *so, struct sockopt *sopt); __private_extern__ struct pr_usrreqs icmp_dgram_usrreqs = { - rip_abort, pru_accept_notsupp, icmp_dgram_attach, rip_bind, rip_connect, - pru_connect2_notsupp, in_control, rip_detach, rip_disconnect, - pru_listen_notsupp, in_setpeeraddr, pru_rcvd_notsupp, - pru_rcvoob_notsupp, icmp_dgram_send, pru_sense_null, rip_shutdown, - in_setsockaddr, sosend, soreceive, pru_sopoll_notsupp + .pru_abort = rip_abort, + .pru_attach = icmp_dgram_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 = icmp_dgram_send, + .pru_shutdown = rip_shutdown, + .pru_sockaddr = in_getsockaddr, + .pru_sosend = sosend, + .pru_soreceive = soreceive, }; /* Like rip_attach but without root privilege enforcement */ __private_extern__ int -icmp_dgram_attach(struct socket *so, int proto, struct proc *p) +icmp_dgram_attach(struct socket *so, __unused int proto, struct proc *p) { - struct inpcb *inp; - int error, s; - - inp = sotoinpcb(so); - if (inp) - panic("icmp_dgram_attach"); - - error = soreserve(so, rip_sendspace, rip_recvspace); - if (error) - return error; - s = splnet(); - error = in_pcballoc(so, &ripcbinfo, p); - splx(s); - if (error) - return error; - inp = (struct inpcb *)so->so_pcb; - inp->inp_vflag |= INP_IPV4; - inp->inp_ip_p = IPPROTO_ICMP; - inp->inp_ip_ttl = ip_defttl; - return 0; + struct inpcb *inp; + int error; + + inp = sotoinpcb(so); + if (inp) { + panic("icmp_dgram_attach"); + } + + error = soreserve(so, rip_sendspace, rip_recvspace); + if (error) { + return error; + } + error = in_pcballoc(so, &ripcbinfo, p); + if (error) { + return error; + } + inp = (struct inpcb *)so->so_pcb; + inp->inp_vflag |= INP_IPV4; + inp->inp_ip_p = IPPROTO_ICMP; + inp->inp_ip_ttl = (u_char)ip_defttl; + return 0; } /* @@ -1039,60 +1233,86 @@ icmp_dgram_attach(struct socket *so, int proto, struct proc *p) __private_extern__ int icmp_dgram_ctloutput(struct socket *so, struct sockopt *sopt) { - struct inpcb *inp = sotoinpcb(so); - int error, optval; + int error; - if (sopt->sopt_level != IPPROTO_IP) - return (EINVAL); + if (sopt->sopt_level != IPPROTO_IP) { + return EINVAL; + } switch (sopt->sopt_name) { - case IP_OPTIONS: - case IP_HDRINCL: - case IP_TOS: - case IP_TTL: - case IP_RECVOPTS: - case IP_RECVRETOPTS: - case IP_RECVDSTADDR: - case IP_RETOPTS: - case IP_MULTICAST_IF: - case IP_MULTICAST_TTL: - case IP_MULTICAST_LOOP: - case IP_ADD_MEMBERSHIP: - case IP_DROP_MEMBERSHIP: - case IP_MULTICAST_VIF: - case IP_PORTRANGE: - case IP_RECVIF: - case IP_IPSEC_POLICY: -#if defined(NFAITH) && NFAITH > 0 - case IP_FAITH: -#endif - case IP_STRIPHDR: - case IP_RECVTTL: - error = rip_ctloutput(so, sopt); - break; - - default: - error = EINVAL; - break; + case IP_OPTIONS: + case IP_HDRINCL: + case IP_TOS: + case IP_TTL: + case IP_RECVOPTS: + case IP_RECVRETOPTS: + case IP_RECVDSTADDR: + case IP_RETOPTS: + case IP_MULTICAST_IF: + case IP_MULTICAST_IFINDEX: + case IP_MULTICAST_TTL: + case IP_MULTICAST_LOOP: + case IP_ADD_MEMBERSHIP: + case IP_DROP_MEMBERSHIP: + case IP_MULTICAST_VIF: + case IP_PORTRANGE: + case IP_RECVIF: + case IP_IPSEC_POLICY: + case IP_STRIPHDR: + case IP_RECVTTL: + case IP_BOUND_IF: + case IP_DONTFRAG: + case IP_NO_IFT_CELLULAR: + error = rip_ctloutput(so, sopt); + break; + + default: + error = EINVAL; + break; } - return (error); + return error; } __private_extern__ int -icmp_dgram_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, - struct mbuf *control, struct proc *p) +icmp_dgram_send(struct socket *so, int flags, struct mbuf *m, + struct sockaddr *nam, struct mbuf *control, struct proc *p) { struct ip *ip; struct inpcb *inp = sotoinpcb(so); int hlen; struct icmp *icp; - struct in_ifaddr *ia = NULL; + struct in_ifaddr *ia = NULL; int icmplen; + int error = EINVAL; + int inp_flags = inp ? inp->inp_flags : 0; + + if (inp == NULL +#if NECP + || (necp_socket_should_use_flow_divert(inp)) +#endif /* NECP */ + ) { + if (inp != NULL) { + error = EPROTOTYPE; + } + goto bad; + } - if ((inp->inp_flags & INP_HDRINCL) != 0) { +#if CONTENT_FILTER + /* + * If socket is subject to Content Filter, get inp_flags from saved state + */ + if (so->so_cfil_db && nam == NULL) { + cfil_dgram_peek_socket_state(m, &inp_flags); + } +#endif + + if ((inp_flags & INP_HDRINCL) != 0) { + /* Expect 32-bit aligned data ptr on strict-align platforms */ + MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m); /* - * This is not raw IP, we liberal only for fields TOS, id and TTL + * This is not raw IP, we liberal only for fields TOS, + * id and TTL. */ ip = mtod(m, struct ip *); @@ -1102,41 +1322,53 @@ icmp_dgram_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *n goto bad; } /* Only IPv4 */ - if (IP_VHL_V(ip->ip_vhl) != 4) + if (IP_VHL_V(ip->ip_vhl) != 4) { goto bad; - if (hlen < 20 || hlen > 40 || ip->ip_len != m->m_pkthdr.len) + } + if (hlen < 20 || hlen > 40 || ip->ip_len != m->m_pkthdr.len) { goto bad; - /* Bogus fragments can tie up peer resources */ - if (ip->ip_off != 0) + } + /* Bogus fragments can tie up peer resources */ + if ((ip->ip_off & ~IP_DF) != 0) { goto bad; + } /* Allow only ICMP even for user provided IP header */ - if (ip->ip_p != IPPROTO_ICMP) + if (ip->ip_p != IPPROTO_ICMP) { goto bad; - /* To prevent spoofing, specified source address must be one of ours */ + } + /* + * To prevent spoofing, specified source address must + * be one of ours. + */ if (ip->ip_src.s_addr != INADDR_ANY) { socket_unlock(so, 0); - lck_mtx_lock(rt_mtx); + lck_rw_lock_shared(in_ifaddr_rwlock); if (TAILQ_EMPTY(&in_ifaddrhead)) { - lck_mtx_unlock(rt_mtx); + lck_rw_done(in_ifaddr_rwlock); socket_lock(so, 0); goto bad; } - TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) { - if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_src.s_addr) { - lck_mtx_unlock(rt_mtx); + TAILQ_FOREACH(ia, INADDR_HASH(ip->ip_src.s_addr), + ia_hash) { + IFA_LOCK(&ia->ia_ifa); + if (IA_SIN(ia)->sin_addr.s_addr == + ip->ip_src.s_addr) { + IFA_UNLOCK(&ia->ia_ifa); + lck_rw_done(in_ifaddr_rwlock); socket_lock(so, 0); goto ours; } + IFA_UNLOCK(&ia->ia_ifa); } - lck_mtx_unlock(rt_mtx); + lck_rw_done(in_ifaddr_rwlock); socket_lock(so, 0); goto bad; } ours: /* Do not trust we got a valid checksum */ ip->ip_sum = 0; - - icp = (struct icmp *)(((char *)m->m_data) + hlen); + + icp = (struct icmp *)(void *)(((char *)m->m_data) + hlen); icmplen = m->m_pkthdr.len - hlen; } else { if ((icmplen = m->m_pkthdr.len) < ICMP_MINLEN) { @@ -1147,27 +1379,37 @@ ours: /* * Allow only to send request types with code 0 */ - if (icp->icmp_code != 0) + if (icp->icmp_code != 0) { goto bad; + } switch (icp->icmp_type) { - case ICMP_ECHO: - break; - case ICMP_TSTAMP: - if (icmplen != 20) - goto bad; - break; - case ICMP_MASKREQ: - if (icmplen != 12) - goto bad; - break; - default: + case ICMP_ECHO: + break; + case ICMP_TSTAMP: + if (icmplen != 20) { + goto bad; + } + break; + case ICMP_MASKREQ: + if (icmplen != 12) { goto bad; + } + break; + default: + goto bad; } return rip_send(so, flags, m, nam, control, p); bad: - m_freem(m); - return EINVAL; + VERIFY(error != 0); + + if (m != NULL) { + m_freem(m); + } + if (control != NULL) { + m_freem(control); + } + + return error; } #endif /* __APPLE__ */ -