X-Git-Url: https://git.saurik.com/apple/xnu.git/blobdiff_plain/1c79356b52d46aa6b508fb032f5ae709b1f2897b..89b3af67bb32e691275bf6fa803d1834b2284115:/bsd/netinet/tcp_input.c diff --git a/bsd/netinet/tcp_input.c b/bsd/netinet/tcp_input.c index cab8916e1..a6ea7bf86 100644 --- a/bsd/netinet/tcp_input.c +++ b/bsd/netinet/tcp_input.c @@ -1,23 +1,29 @@ /* * Copyright (c) 2000 Apple Computer, Inc. All rights reserved. * - * @APPLE_LICENSE_HEADER_START@ + * @APPLE_OSREFERENCE_LICENSE_HEADER_START@ * - * The contents of this file constitute Original Code as defined in and - * are subject to the Apple Public Source License Version 1.1 (the - * "License"). You may not use this file except in compliance with the - * License. Please obtain a copy of the License at - * http://www.apple.com/publicsource and read it before using this file. + * 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 + * compliance with the License. The rights granted to you under the License + * may not be used to create, or enable the creation or redistribution of, + * 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. * - * This Original Code and all software distributed under the License are - * distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, EITHER + * 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, * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT. Please see the - * License for the specific language governing rights and limitations - * under the License. + * 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_LICENSE_HEADER_END@ + * @APPLE_OSREFERENCE_LICENSE_HEADER_END@ */ /* * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1994, 1995 @@ -52,12 +58,9 @@ * SUCH DAMAGE. * * @(#)tcp_input.c 8.12 (Berkeley) 5/24/95 + * $FreeBSD: src/sys/netinet/tcp_input.c,v 1.107.2.16 2001/08/22 00:59:12 silby Exp $ */ -#if ISFB31 -#include "opt_ipfw.h" /* for ipfw_fwd */ -#include "opt_tcpdebug.h" -#endif #include #include @@ -74,16 +77,17 @@ #include /* before tcp_seq.h, for tcp_random18() */ #include +#include #include #include #include #include -#include /* for ICMP_BANDLIM */ +#include /* for ICMP_BANDLIM */ #include -#include -#include /* for ICMP_BANDLIM */ +#include /* for ICMP_BANDLIM */ #include +#include #if INET6 #include #include @@ -96,163 +100,217 @@ #include #include #include +#if INET6 +#include +#endif #include #if TCPDEBUG #include -#if INET6 -union { - struct ip _tcp_si4; - struct ip6_hdr _tcp_si6; -} tcp_saveip; -#else -struct ip tcp_saveip; -#endif /* INET6 */ +u_char tcp_saveipgen[40]; /* the size must be of max ip header, now IPv6 */ struct tcphdr tcp_savetcp; #endif /* TCPDEBUG */ #if IPSEC #include +#if INET6 +#include +#endif #include #endif /*IPSEC*/ #include +#ifndef __APPLE__ +MALLOC_DEFINE(M_TSEGQ, "tseg_qent", "TCP segment queue entry"); +#endif + #define DBG_LAYER_BEG NETDBG_CODE(DBG_NETTCP, 0) #define DBG_LAYER_END NETDBG_CODE(DBG_NETTCP, 2) #define DBG_FNC_TCP_INPUT NETDBG_CODE(DBG_NETTCP, (3 << 8)) #define DBG_FNC_TCP_NEWCONN NETDBG_CODE(DBG_NETTCP, (7 << 8)) static int tcprexmtthresh = 3; -tcp_seq tcp_iss; tcp_cc tcp_ccgen; +extern int apple_hwcksum_rx; + +#if IPSEC +extern int ipsec_bypass; +extern lck_mtx_t *sadb_mutex; +#endif struct tcpstat tcpstat; -SYSCTL_STRUCT(_net_inet_tcp, TCPCTL_STATS, stats, - CTLFLAG_RD, &tcpstat , tcpstat, ""); -int log_in_vain = 0; +static int log_in_vain = 0; SYSCTL_INT(_net_inet_tcp, OID_AUTO, log_in_vain, CTLFLAG_RW, - &log_in_vain, 0, ""); + &log_in_vain, 0, "Log all incoming TCP connections"); + +static int blackhole = 0; +SYSCTL_INT(_net_inet_tcp, OID_AUTO, blackhole, CTLFLAG_RW, + &blackhole, 0, "Do not send RST when dropping refused connections"); -int tcp_delack_enabled = 1; +int tcp_delack_enabled = 3; SYSCTL_INT(_net_inet_tcp, OID_AUTO, delayed_ack, CTLFLAG_RW, - &tcp_delack_enabled, 0, ""); + &tcp_delack_enabled, 0, + "Delay ACK to try and piggyback it onto a data packet"); + +int tcp_lq_overflow = 1; +SYSCTL_INT(_net_inet_tcp, OID_AUTO, tcp_lq_overflow, CTLFLAG_RW, + &tcp_lq_overflow, 0, + "Listen Queue Overflow"); + +#if TCP_DROP_SYNFIN +static int drop_synfin = 1; +SYSCTL_INT(_net_inet_tcp, OID_AUTO, drop_synfin, CTLFLAG_RW, + &drop_synfin, 0, "Drop TCP packets with SYN+FIN set"); +#endif -u_long tcp_now; -struct inpcbhead tcb; -#define tcb6 tcb /* for KAME src sync over BSD*'s */ -struct inpcbinfo tcbinfo; +SYSCTL_NODE(_net_inet_tcp, OID_AUTO, reass, CTLFLAG_RW, 0, + "TCP Segment Reassembly Queue"); -static void tcp_dooptions __P((struct tcpcb *, - u_char *, int, struct tcphdr *, struct tcpopt *)); -static void tcp_pulloutofband __P((struct socket *, - struct tcphdr *, struct mbuf *)); -static void tcp_xmit_timer __P((struct tcpcb *, int)); +__private_extern__ int tcp_reass_maxseg = 0; +SYSCTL_INT(_net_inet_tcp_reass, OID_AUTO, maxsegments, CTLFLAG_RW, + &tcp_reass_maxseg, 0, + "Global maximum number of TCP Segments in Reassembly Queue"); -/* - * Neighbor Discovery, Neighbor Unreachability Detection - * Upper layer hint. - */ -#define ND6_HINT(tp) { \ - if ((tp) && (tp)->t_inpcb && (tp)->t_inpcb->in6p_route.ro_rt) \ - nd6_nud_hint((tp)->t_inpcb->in6p_route.ro_rt, NULL); \ -} +__private_extern__ int tcp_reass_qsize = 0; +SYSCTL_INT(_net_inet_tcp_reass, OID_AUTO, cursegments, CTLFLAG_RD, + &tcp_reass_qsize, 0, + "Global number of TCP Segments currently in Reassembly Queue"); +static int tcp_reass_overflows = 0; +SYSCTL_INT(_net_inet_tcp_reass, OID_AUTO, overflows, CTLFLAG_RD, + &tcp_reass_overflows, 0, + "Global number of TCP Segment Reassembly Queue Overflows"); -extern u_long current_active_connections; -extern u_long last_active_conn_count; -extern u_long *delack_bitmask; +__private_extern__ int slowlink_wsize = 8192; +SYSCTL_INT(_net_inet_tcp, OID_AUTO, slowlink_wsize, CTLFLAG_RW, + &slowlink_wsize, 0, "Maximum advertised window size for slowlink"); +u_long tcp_now; +struct inpcbhead tcb; +#define tcb6 tcb /* for KAME src sync over BSD*'s */ +struct inpcbinfo tcbinfo; +static void tcp_dooptions(struct tcpcb *, + u_char *, int, struct tcphdr *, struct tcpopt *); +static void tcp_pulloutofband(struct socket *, + struct tcphdr *, struct mbuf *, int); +static int tcp_reass(struct tcpcb *, struct tcphdr *, int *, + struct mbuf *); +static void tcp_xmit_timer(struct tcpcb *, int); -/* - * Insert segment ti into reassembly queue of tcp with - * control block tp. Return TH_FIN if reassembly now includes - * a segment with FIN. The macro form does 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). - */ +/* Neighbor Discovery, Neighbor Unreachability Detection Upper layer hint. */ #if INET6 -#define _ONLY_IF_INET6_(x) x +#define ND6_HINT(tp) \ +do { \ + if ((tp) && (tp)->t_inpcb && \ + ((tp)->t_inpcb->inp_vflag & INP_IPV6) != 0 && \ + (tp)->t_inpcb->in6p_route.ro_rt) \ + nd6_nud_hint((tp)->t_inpcb->in6p_route.ro_rt, NULL, 0); \ +} while (0) #else -#define _ONLY_IF_INET6_(x) +#define ND6_HINT(tp) #endif -#define TCP_REASS(tp, th, tilen, m, so, flags, isipv6, needwakeup) { \ - if ((th)->th_seq == (tp)->rcv_nxt && \ - (tp)->segq.lh_first == NULL && \ - (tp)->t_state == TCPS_ESTABLISHED) { \ - if (tcp_delack_enabled) {\ - if (last_active_conn_count > DELACK_BITMASK_THRESH) \ - TCP_DELACK_BITSET(tp->t_inpcb->hash_element); \ - tp->t_flags |= TF_DELACK; \ - } \ - else \ - tp->t_flags |= TF_ACKNOW; \ - (tp)->rcv_nxt += (tilen); \ - flags = (th)->th_flags & TH_FIN; \ - tcpstat.tcps_rcvpack++;\ - tcpstat.tcps_rcvbyte += (tilen);\ - _ONLY_IF_INET6_(ND6_HINT(tp);) \ - sbappend(&(so)->so_rcv, (m)); \ - needwakeup++; \ - } else { \ - (flags) = tcp_reass((tp), (th), (tilen), (m), (isipv6)); \ - tp->t_flags |= TF_ACKNOW; \ - } \ + +extern u_long *delack_bitmask; + +extern void ipfwsyslog( int level, char *format,...); +extern int ChkAddressOK( __uint32_t dstaddr, __uint32_t srcaddr ); +extern int fw_verbose; + +#define log_in_vain_log( a ) { \ + if ( (log_in_vain == 3 ) && (fw_verbose == 2)) { /* Apple logging, log to ipfw.log */ \ + ipfwsyslog a ; \ + } \ + else log a ; \ } /* - * Note: - * in the ip header part of the ipqe_tcp structure only the length is used. + * Indicate whether this ack should be delayed. + * We can delay the ack if: + * - delayed acks are enabled (set to 1) and + * - our last ack wasn't a 0-sized window. We never want to delay + * the ack that opens up a 0-sized window. + * - delayed acks are enabled (set to 2, "more compatible") and + * - our last ack wasn't a 0-sized window. + * - if the peer hasn't sent us a TH_PUSH data packet (this solves 3649245) + * - the peer hasn't sent us a TH_PUSH data packet, if he did, take this as a clue that we + * need to ACK with no delay. This helps higher level protocols who won't send + * us more data even if the window is open because their last "segment" hasn't been ACKed + * - delayed acks are enabled (set to 3, "streaming detection") and + * - if we receive more than 4 full packets per second on this socket, we're streaming acts as "1". + * - if we don't meet that criteria, acts like "2". Allowing faster acking while browsing for example. + * */ -int -tcp_reass(tp, th, tilen, m, isipv6) +#define DELAY_ACK(tp) \ + (((tcp_delack_enabled == 1) && ((tp->t_flags & TF_RXWIN0SENT) == 0)) || \ + (((tcp_delack_enabled == 2) && (tp->t_flags & TF_RXWIN0SENT) == 0) && \ + ((thflags & TH_PUSH) == 0) && ((tp->t_flags & TF_DELACK) == 0)) || \ + (((tcp_delack_enabled == 3) && (tp->t_flags & TF_RXWIN0SENT) == 0) && \ + (((tp->t_rcvtime == 0) && (tp->rcv_byps > (4* tp->t_maxseg))) || (((thflags & TH_PUSH) == 0) && ((tp->t_flags & TF_DELACK) == 0))))) + + +static int tcpdropdropablreq(struct socket *head); +static void tcp_newreno_partial_ack(struct tcpcb *tp, struct tcphdr *th); + + +static int +tcp_reass(tp, th, tlenp, m) register struct tcpcb *tp; register struct tcphdr *th; - u_int16_t tilen; + int *tlenp; struct mbuf *m; -#if INET6 - int isipv6; -#endif { - register struct ipqent *p, *q, *nq, *tiqe; + struct tseg_qent *q; + struct tseg_qent *p = NULL; + struct tseg_qent *nq; + struct tseg_qent *te = NULL; struct socket *so = tp->t_inpcb->inp_socket; int flags; + int dowakeup = 0; /* * Call with th==0 after become established to * force pre-ESTABLISHED data up to user socket. */ - if (th == 0) + if (th == NULL) goto present; -#if 0 /* Not using GETTCP(m) macro */ - m->m_pkthdr.header = ti; -#endif - /* - * Allocate a new queue entry, before we throw away any data. - * If we can't, just drop the packet. XXX + * Limit the number of segments in the reassembly queue to prevent + * holding on to too many segments (and thus running out of mbufs). + * Make sure to let the missing segment through which caused this + * queue. Always keep one global queue entry spare to be able to + * process the missing segment. */ - MALLOC(tiqe, struct ipqent *, sizeof (struct ipqent), M_SONAME, M_NOWAIT); - if (tiqe == NULL) { + if (th->th_seq != tp->rcv_nxt && + tcp_reass_qsize + 1 >= tcp_reass_maxseg) { + tcp_reass_overflows++; + tcpstat.tcps_rcvmemdrop++; + m_freem(m); + return (0); + } + + /* Allocate a new queue entry. If we can't, just drop the pkt. XXX */ + MALLOC(te, struct tseg_qent *, sizeof (struct tseg_qent), M_TSEGQ, + M_NOWAIT); + if (te == NULL) { tcpstat.tcps_rcvmemdrop++; m_freem(m); return (0); } + tcp_reass_qsize++; /* * Find a segment which begins after this one does. */ - for (p = NULL, q = tp->segq.lh_first; q != NULL; - p = q, q = q->ipqe_q.le_next) - if (SEQ_GT(q->ipqe_tcp->ti_seq, th->th_seq)) + LIST_FOREACH(q, &tp->t_segq, tqe_q) { + if (SEQ_GT(q->tqe_th->th_seq, th->th_seq)) break; + p = q; + } /* * If there is a preceding segment, it may provide some of @@ -260,19 +318,16 @@ tcp_reass(tp, th, tilen, m, isipv6) * segment. If it provides all of our data, drop us. */ if (p != NULL) { - register struct tcpiphdr *phdr = p->ipqe_tcp; register int i; - /* conversion to int (in i) handles seq wraparound */ - i = phdr->ti_seq + phdr->ti_len - th->th_seq; + i = p->tqe_th->th_seq + p->tqe_len - th->th_seq; if (i > 0) { - if (i >= tilen) { + if (i >= *tlenp) { tcpstat.tcps_rcvduppack++; - tcpstat.tcps_rcvdupbyte += tilen; + tcpstat.tcps_rcvdupbyte += *tlenp; m_freem(m); - FREE(tiqe, M_SONAME); - -#if 1 /* XXX: NetBSD just return 0 here */ + FREE(te, M_TSEGQ); + tcp_reass_qsize--; /* * Try to present any queued data * at the left window edge to the user. @@ -280,54 +335,47 @@ tcp_reass(tp, th, tilen, m, isipv6) * completes. */ goto present; /* ??? */ -#endif } m_adj(m, i); - tilen -= i; + *tlenp -= i; th->th_seq += i; } } tcpstat.tcps_rcvoopack++; - tcpstat.tcps_rcvoobyte += tilen; + tcpstat.tcps_rcvoobyte += *tlenp; /* * While we overlap succeeding segments trim them or, * if they are completely covered, dequeue them. */ while (q) { - register struct tcpiphdr *qhdr = q->ipqe_tcp; - register int i = (th->th_seq + tilen) - qhdr->ti_seq; - + register int i = (th->th_seq + *tlenp) - q->tqe_th->th_seq; if (i <= 0) break; - if (i < qhdr->ti_len) { - qhdr->ti_seq += i; - qhdr->ti_len -= i; - m_adj(q->ipqe_m, i); + if (i < q->tqe_len) { + q->tqe_th->th_seq += i; + q->tqe_len -= i; + m_adj(q->tqe_m, i); break; } - nq = q->ipqe_q.le_next; - m_freem(q->ipqe_m); - LIST_REMOVE(q, ipqe_q); - FREE(q, M_SONAME); + + nq = LIST_NEXT(q, tqe_q); + LIST_REMOVE(q, tqe_q); + m_freem(q->tqe_m); + FREE(q, M_TSEGQ); + tcp_reass_qsize--; q = nq; } - /* Insert the new fragment queue entry into place. */ - tiqe->ipqe_m = m; - /* - * There is a IP or IPv6 header in the mbuf before th - * so there is space for an ip header (for the length field) - */ -#define thtoti(x) \ - ((struct tcpiphdr *)(((char *)(x)) - (sizeof (struct ip)))) + /* Insert the new segment queue entry into place. */ + te->tqe_m = m; + te->tqe_th = th; + te->tqe_len = *tlenp; - tiqe->ipqe_tcp = thtoti(th); - tiqe->ipqe_tcp->ti_len = tilen; if (p == NULL) { - LIST_INSERT_HEAD(&tp->segq, tiqe, ipqe_q); + LIST_INSERT_HEAD(&tp->t_segq, te, tqe_q); } else { - LIST_INSERT_AFTER(p, tiqe, ipqe_q); + LIST_INSERT_AFTER(p, te, tqe_q); } present: @@ -337,61 +385,89 @@ present: */ if (!TCPS_HAVEESTABLISHED(tp->t_state)) return (0); - q = tp->segq.lh_first; - if (!q || q->ipqe_tcp->ti_seq != tp->rcv_nxt) - return (0); -#if 0 - /* - * XXX from INRIA for NetBSD, but should not happen because - * TCPS_HAVEESTABLISHED(tp->t_state) should be true here. - */ - if (tp->t_state == TCPS_SYN_RECEIVED && q->ipqe_tcp->ti_len) + q = LIST_FIRST(&tp->t_segq); + if (!q || q->tqe_th->th_seq != tp->rcv_nxt) return (0); -#endif do { - tp->rcv_nxt += q->ipqe_tcp->ti_len; - flags = q->ipqe_tcp->ti_flags & TH_FIN; - nq = q->ipqe_q.le_next; - LIST_REMOVE(q, ipqe_q); + tp->rcv_nxt += q->tqe_len; + flags = q->tqe_th->th_flags & TH_FIN; + nq = LIST_NEXT(q, tqe_q); + LIST_REMOVE(q, tqe_q); if (so->so_state & SS_CANTRCVMORE) - m_freem(q->ipqe_m); - else - sbappend(&so->so_rcv, q->ipqe_m); - FREE(q, M_SONAME); + m_freem(q->tqe_m); + else { + if (sbappend(&so->so_rcv, q->tqe_m)) + dowakeup = 1; + } + FREE(q, M_TSEGQ); + tcp_reass_qsize--; q = nq; - } while (q && q->ipqe_tcp->ti_seq == tp->rcv_nxt); + } while (q && q->tqe_th->th_seq == tp->rcv_nxt); + ND6_HINT(tp); + #if INET6 - if (isipv6) - ND6_HINT(tp); + if ((tp->t_inpcb->inp_vflag & INP_IPV6) != 0) { + + KERNEL_DEBUG(DBG_LAYER_BEG, + ((tp->t_inpcb->inp_fport << 16) | tp->t_inpcb->inp_lport), + (((tp->t_inpcb->in6p_laddr.s6_addr16[0] & 0xffff) << 16) | + (tp->t_inpcb->in6p_faddr.s6_addr16[0] & 0xffff)), + 0,0,0); + } + else #endif - - KERNEL_DEBUG(DBG_LAYER_END, ((ti->ti_dport << 16) | ti->ti_sport), - (((ti->ti_src.s_addr & 0xffff) << 16) | (ti->ti_dst.s_addr & 0xffff)), - ti->ti_seq, ti->ti_ack, ti->ti_win); - - sorwakeup(so); + { + KERNEL_DEBUG(DBG_LAYER_BEG, + ((tp->t_inpcb->inp_fport << 16) | tp->t_inpcb->inp_lport), + (((tp->t_inpcb->inp_laddr.s_addr & 0xffff) << 16) | + (tp->t_inpcb->inp_faddr.s_addr & 0xffff)), + 0,0,0); + } + if (dowakeup) + sorwakeup(so); /* done with socket lock held */ return (flags); + } + /* * TCP input routine, follows pages 65-76 of the * protocol specification dated September, 1981 very closely. */ #if INET6 int -tcp6_input(mp, offp, proto) +tcp6_input(mp, offp) struct mbuf **mp; - int *offp, proto; + int *offp; { - tcp_input(*mp, *offp); + register struct mbuf *m = *mp; + struct in6_ifaddr *ia6; + + IP6_EXTHDR_CHECK(m, *offp, sizeof(struct tcphdr), return 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; + + 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 void -tcp_input(m, off) +tcp_input(m, off0) struct mbuf *m; - int off; + int off0; { register struct tcphdr *th; register struct ip *ip = NULL; @@ -399,154 +475,197 @@ tcp_input(m, off) register struct inpcb *inp; u_char *optp = NULL; int optlen = 0; - int len, toff; - int hdroptlen; - u_int16_t tilen; + int len, tlen, off; + int drop_hdrlen; register struct tcpcb *tp = 0; register int thflags; struct socket *so = 0; int todrop, acked, ourfinisacked, needoutput = 0; struct in_addr laddr; -#if 0 +#if INET6 struct in6_addr laddr6; #endif int dropsocket = 0; int iss = 0; + int nosock = 0; u_long tiwin; struct tcpopt to; /* options in this segment */ struct rmxp_tao *taop; /* pointer to our TAO cache entry */ struct rmxp_tao tao_noncached; /* in case there's no cached entry */ - int need_sowwakeup = 0; - int need_sorwakeup = 0; + struct sockaddr_in *next_hop = NULL; #if TCPDEBUG short ostate = 0; #endif + struct m_tag *fwd_tag; + + /* Grab info from PACKET_TAG_IPFORWARD tag prepended to the chain. */ + fwd_tag = m_tag_locate(m, KERNEL_MODULE_TAG_ID, KERNEL_TAG_TYPE_IPFORWARD, NULL); + if (fwd_tag != NULL) { + struct ip_fwd_tag *ipfwd_tag = (struct ip_fwd_tag *)(fwd_tag+1); + + next_hop = ipfwd_tag->next_hop; + m_tag_delete(m, fwd_tag); + } + #if INET6 struct ip6_hdr *ip6 = NULL; - int lgminh; -#else /* INET6 */ -#define lgminh (sizeof (struct tcpiphdr)) + int isipv6; #endif /* INET6 */ - int isipv6 = (mtod(m, struct ip *)->ip_v == 6) ? 1 : 0; - + int rstreason; /* For badport_bandlim accounting purposes */ struct proc *proc0=current_proc(); - + KERNEL_DEBUG(DBG_FNC_TCP_INPUT | DBG_FUNC_START,0,0,0,0,0); +#if INET6 + isipv6 = (mtod(m, struct ip *)->ip_v == 6) ? 1 : 0; +#endif bzero((char *)&to, sizeof(to)); tcpstat.tcps_rcvtotal++; - /* - * Get IP and TCP header together in first mbuf. - * Note: IP leaves IP header in first mbuf. - */ - th = mtod(m, struct tcpiphdr *); - KERNEL_DEBUG(DBG_LAYER_BEG, ((ti->ti_dport << 16) | ti->ti_sport), - (((ti->ti_src.s_addr & 0xffff) << 16) | (ti->ti_dst.s_addr & 0xffff)), - ti->ti_seq, ti->ti_ack, ti->ti_win); -#if INET6 - if (isipv6) { - ip6 = mtod(m, struct ip6_hdr *); - lgminh = sizeof(struct tcpip6hdr); - } else { - lgminh = sizeof(struct tcpiphdr); -#endif /* INET6 */ - ip = mtod(m, struct ip *); - ipov = (struct ipovly *)ip; -#if INET6 - } -#endif /* INET6 */ - -#if INET6 - /* XXX not a good place to put this into... */ - if (isipv6 && - m && (m->m_flags & M_ANYCAST6)) { - icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_ADDR, - (caddr_t)&ip6->ip6_dst - (caddr_t)ip6); - return; - } -#endif /* INET6 */ #if INET6 if (isipv6) { - IP6_EXTHDR_CHECK(m, off, sizeof(struct tcphdr), ); + /* IP6_EXTHDR_CHECK() is already done at tcp6_input() */ ip6 = mtod(m, struct ip6_hdr *); - tilen = ntohs(ip6->ip6_plen) - off + sizeof(*ip6); - - if (in6_cksum(m, IPPROTO_TCP, off, tilen)) { + tlen = sizeof(*ip6) + ntohs(ip6->ip6_plen) - off0; + if (in6_cksum(m, IPPROTO_TCP, off0, tlen)) { tcpstat.tcps_rcvbadsum++; - goto drop; + goto dropnosock; + } + th = (struct tcphdr *)((caddr_t)ip6 + off0); + + KERNEL_DEBUG(DBG_LAYER_BEG, ((th->th_dport << 16) | th->th_sport), + (((ip6->ip6_src.s6_addr16[0]) << 16) | (ip6->ip6_dst.s6_addr16[0])), + th->th_seq, th->th_ack, th->th_win); + /* + * 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 dropnosock; } - th = (struct tcphdr *)((caddr_t)ip6 + off); } else #endif /* INET6 */ { - /* - * Get IP and TCP header together in first mbuf. - * Note: IP leaves IP header in first mbuf. - */ - /* XXX: should we still require this for IPv4? */ - if (off > sizeof (struct ip)) { - ip_stripoptions(m, (struct mbuf *)0); - off = sizeof(struct ip); + /* + * 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_pkthdr.csum_flags & CSUM_TCP_SUM16) + m->m_pkthdr.csum_flags = 0; /* invalidate hwcksuming */ + + } + if (m->m_len < sizeof (struct tcpiphdr)) { + if ((m = m_pullup(m, sizeof (struct tcpiphdr))) == 0) { + tcpstat.tcps_rcvshort++; + return; } - if (m->m_len < lgminh) { - if ((m = m_pullup(m, lgminh)) == 0) { - tcpstat.tcps_rcvshort++; - return; - } + } + ip = mtod(m, struct ip *); + ipov = (struct ipovly *)ip; + th = (struct tcphdr *)((caddr_t)ip + off0); + tlen = ip->ip_len; + + KERNEL_DEBUG(DBG_LAYER_BEG, ((th->th_dport << 16) | th->th_sport), + (((ip->ip_src.s_addr & 0xffff) << 16) | (ip->ip_dst.s_addr & 0xffff)), + th->th_seq, th->th_ack, th->th_win); + + if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) { + if (apple_hwcksum_rx && (m->m_pkthdr.csum_flags & CSUM_TCP_SUM16)) { + u_short pseudo; + char b[9]; + *(uint32_t*)&b[0] = *(uint32_t*)&ipov->ih_x1[0]; + *(uint32_t*)&b[4] = *(uint32_t*)&ipov->ih_x1[4]; + *(uint8_t*)&b[8] = *(uint8_t*)&ipov->ih_x1[8]; + + bzero(ipov->ih_x1, sizeof(ipov->ih_x1)); + ipov->ih_len = (u_short)tlen; + HTONS(ipov->ih_len); + pseudo = in_cksum(m, sizeof (struct ip)); + + *(uint32_t*)&ipov->ih_x1[0] = *(uint32_t*)&b[0]; + *(uint32_t*)&ipov->ih_x1[4] = *(uint32_t*)&b[4]; + *(uint8_t*)&ipov->ih_x1[8] = *(uint8_t*)&b[8]; + + th->th_sum = in_addword(pseudo, (m->m_pkthdr.csum_data & 0xFFFF)); + } else { + 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)); } - ip = mtod(m, struct ip *); - ipov = (struct ipovly *)ip; - + th->th_sum ^= 0xffff; + } else { + char b[9]; /* * Checksum extended TCP header and data. */ - tilen = ip->ip_len; - len = sizeof (struct ip) + tilen; + *(uint32_t*)&b[0] = *(uint32_t*)&ipov->ih_x1[0]; + *(uint32_t*)&b[4] = *(uint32_t*)&ipov->ih_x1[4]; + *(uint8_t*)&b[8] = *(uint8_t*)&ipov->ih_x1[8]; + + len = sizeof (struct ip) + tlen; bzero(ipov->ih_x1, sizeof(ipov->ih_x1)); - ipov->ih_len = (u_short)tilen; + ipov->ih_len = (u_short)tlen; HTONS(ipov->ih_len); - th = (struct tcphdr *)((caddr_t)ip + off); th->th_sum = in_cksum(m, len); - if (th->th_sum) { - tcpstat.tcps_rcvbadsum++; - goto drop; - } + + *(uint32_t*)&ipov->ih_x1[0] = *(uint32_t*)&b[0]; + *(uint32_t*)&ipov->ih_x1[4] = *(uint32_t*)&b[4]; + *(uint8_t*)&ipov->ih_x1[8] = *(uint8_t*)&b[8]; + } + if (th->th_sum) { + tcpstat.tcps_rcvbadsum++; + goto dropnosock; + } +#if INET6 + /* Re-initialization for later version check */ + ip->ip_v = IPVERSION; +#endif } /* * Check that TCP offset makes sense, * pull out TCP options and adjust length. XXX */ - toff = th->th_off << 2; - if (toff < sizeof (struct tcphdr) || toff > tilen) { + off = th->th_off << 2; + if (off < sizeof (struct tcphdr) || off > tlen) { tcpstat.tcps_rcvbadoff++; - goto drop; + goto dropnosock; } - tilen -= toff; - if (toff > sizeof (struct tcphdr)) { + tlen -= off; /* tlen is used instead of ti->ti_len */ + if (off > sizeof (struct tcphdr)) { #if INET6 if (isipv6) { - IP6_EXTHDR_CHECK(m, off, toff, ); + IP6_EXTHDR_CHECK(m, off0, off, return); ip6 = mtod(m, struct ip6_hdr *); - th = (struct tcphdr *)((caddr_t)ip6 + off); + th = (struct tcphdr *)((caddr_t)ip6 + off0); } else #endif /* INET6 */ - { - if (m->m_len < sizeof(struct ip) + toff) { - if ((m = m_pullup(m, sizeof (struct ip) + toff)) == 0) { - tcpstat.tcps_rcvshort++; - return; - } - ip = mtod(m, struct ip *); - ipov = (struct ipovly *)ip; - th = (struct tcphdr *)((caddr_t)ip + off); + { + if (m->m_len < sizeof(struct ip) + off) { + if ((m = m_pullup(m, sizeof (struct ip) + off)) == 0) { + tcpstat.tcps_rcvshort++; + return; } + ip = mtod(m, struct ip *); + ipov = (struct ipovly *)ip; + th = (struct tcphdr *)((caddr_t)ip + off0); } - optlen = toff - sizeof (struct tcphdr); + } + optlen = off - sizeof (struct tcphdr); optp = (u_char *)(th + 1); /* * Do quick retrieval of timestamp options ("options @@ -560,7 +679,7 @@ tcp_input(m, off) optp[TCPOLEN_TSTAMP_APPA] == TCPOPT_EOL)) && *(u_int32_t *)optp == htonl(TCPOPT_TSTAMP_HDR) && (th->th_flags & TH_SYN) == 0) { - to.to_flag |= TOF_TS; + to.to_flags |= TOF_TS; to.to_tsval = ntohl(*(u_int32_t *)(optp + 4)); to.to_tsecr = ntohl(*(u_int32_t *)(optp + 8)); optp = NULL; /* we've parsed the options */ @@ -568,6 +687,18 @@ tcp_input(m, off) } thflags = th->th_flags; +#if TCP_DROP_SYNFIN + /* + * If the drop_synfin option is enabled, drop all packets with + * both the SYN and FIN bits set. This prevents e.g. nmap from + * identifying the TCP/IP stack. + * + * This is a violation of the TCP specification. + */ + if (drop_synfin && (thflags & (TH_SYN|TH_FIN)) == (TH_SYN|TH_FIN)) + goto dropnosock; +#endif + /* * Convert TCP protocol specific fields to host format. */ @@ -577,20 +708,23 @@ tcp_input(m, off) NTOHS(th->th_urp); /* - * Drop TCP, IP headers and TCP options. + * Delay dropping TCP, IP headers, IPv6 ext headers, and TCP options, + * until after ip6_savecontrol() is called and before other functions + * which don't want those proto headers. + * Because ip6_savecontrol() is going to parse the mbuf to + * search for data to be passed up to user-land, it wants mbuf + * parameters to be unchanged. */ - hdroptlen = off+toff; - m->m_data += hdroptlen; - m->m_len -= hdroptlen; + drop_hdrlen = off0 + off; /* * Locate pcb for segment. */ findpcb: #if IPFIREWALL_FORWARD - if (ip_fw_fwd_addr != NULL + if (next_hop != NULL #if INET6 - && isipv6 == NULL + && isipv6 == NULL /* IPv6 support is not yet */ #endif /* INET6 */ ) { /* @@ -603,22 +737,21 @@ findpcb: /* * No, then it's new. Try find the ambushing socket */ - if (!ip_fw_fwd_addr->sin_port) { + if (!next_hop->sin_port) { inp = in_pcblookup_hash(&tcbinfo, ip->ip_src, - th->th_sport, ip_fw_fwd_addr->sin_addr, + th->th_sport, next_hop->sin_addr, th->th_dport, 1, m->m_pkthdr.rcvif); } else { inp = in_pcblookup_hash(&tcbinfo, ip->ip_src, th->th_sport, - ip_fw_fwd_addr->sin_addr, - ntohs(ip_fw_fwd_addr->sin_port), 1, + next_hop->sin_addr, + ntohs(next_hop->sin_port), 1, m->m_pkthdr.rcvif); } } - ip_fw_fwd_addr = NULL; } else #endif /* IPFIREWALL_FORWARD */ - + { #if INET6 if (isipv6) inp = in6_pcblookup_hash(&tcbinfo, &ip6->ip6_src, th->th_sport, @@ -628,25 +761,27 @@ findpcb: #endif /* INET6 */ inp = in_pcblookup_hash(&tcbinfo, ip->ip_src, th->th_sport, ip->ip_dst, th->th_dport, 1, m->m_pkthdr.rcvif); + } #if IPSEC - /* due to difference from other BSD stacks */ - m->m_data -= hdroptlen; - m->m_len += hdroptlen; + if (ipsec_bypass == 0) { + lck_mtx_lock(sadb_mutex); #if INET6 - if (isipv6) { - if (inp != NULL && ipsec6_in_reject_so(m, inp->inp_socket)) { - ipsec6stat.in_polvio++; - goto drop; - } - } else + if (isipv6) { + if (inp != NULL && ipsec6_in_reject_so(m, inp->inp_socket)) { + ipsec6stat.in_polvio++; + lck_mtx_unlock(sadb_mutex); + goto dropnosock; + } + } else #endif /* INET6 */ - if (inp != NULL && ipsec4_in_reject_so(m, inp->inp_socket)) { - ipsecstat.in_polvio++; - goto drop; + if (inp != NULL && ipsec4_in_reject_so(m, inp->inp_socket)) { + ipsecstat.in_polvio++; + lck_mtx_unlock(sadb_mutex); + goto dropnosock; + } + lck_mtx_unlock(sadb_mutex); } - m->m_data += hdroptlen; - m->m_len -= hdroptlen; #endif /*IPSEC*/ /* @@ -656,41 +791,105 @@ findpcb: * but should either do a listen or a connect soon. */ if (inp == NULL) { - if (log_in_vain && thflags & TH_SYN) { + if (log_in_vain) { #if INET6 - char buf[INET6_ADDRSTRLEN]; + char dbuf[MAX_IPv6_STR_LEN], sbuf[MAX_IPv6_STR_LEN]; #else /* INET6 */ - char buf[4*sizeof "123"]; + char dbuf[MAX_IPv4_STR_LEN], sbuf[MAX_IPv4_STR_LEN]; #endif /* INET6 */ #if INET6 if (isipv6) { - strcpy(buf, ip6_sprintf(&ip6->ip6_dst)); - log(LOG_INFO, - "Connection attempt to TCP %s:%d from %s:%d\n", - buf, ntohs(th->th_dport), - ip6_sprintf(&ip6->ip6_src), - ntohs(th->th_sport)); - } else { + inet_ntop(AF_INET6, &ip6->ip6_dst, dbuf, sizeof(dbuf)); + inet_ntop(AF_INET6, &ip6->ip6_src, sbuf, sizeof(sbuf)); + } else #endif - strcpy(buf, inet_ntoa(ip->ip_dst)); - log(LOG_INFO, - "Connection attempt to TCP %s:%d from %s:%d\n", - buf, ntohs(th->th_dport), inet_ntoa(ip->ip_src), - ntohs(th->th_sport)); + { + inet_ntop(AF_INET, &ip->ip_dst, dbuf, sizeof(dbuf)); + inet_ntop(AF_INET, &ip->ip_src, sbuf, sizeof(sbuf)); + } + switch (log_in_vain) { + case 1: + if(thflags & TH_SYN) + log(LOG_INFO, + "Connection attempt to TCP %s:%d from %s:%d\n", + dbuf, ntohs(th->th_dport), + sbuf, + ntohs(th->th_sport)); + break; + case 2: + log(LOG_INFO, + "Connection attempt to TCP %s:%d from %s:%d flags:0x%x\n", + dbuf, ntohs(th->th_dport), sbuf, + ntohs(th->th_sport), thflags); + break; + case 3: + if ((thflags & TH_SYN) && + !(m->m_flags & (M_BCAST | M_MCAST)) && #if INET6 + ((isipv6 && !IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &ip6->ip6_src)) || + (!isipv6 && ip->ip_dst.s_addr != ip->ip_src.s_addr)) +#else + ip->ip_dst.s_addr != ip->ip_src.s_addr +#endif + ) + log_in_vain_log((LOG_INFO, + "Stealth Mode connection attempt to TCP %s:%d from %s:%d\n", + dbuf, ntohs(th->th_dport), + sbuf, + ntohs(th->th_sport))); + break; + default: + break; } -#endif /* INET6 */ } -#if ICMP_BANDLIM - if (badport_bandlim(1) < 0) - goto drop; + if (blackhole) { + if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type != IFT_LOOP) + switch (blackhole) { + case 1: + if (thflags & TH_SYN) + goto dropnosock; + break; + case 2: + goto dropnosock; + default: + goto dropnosock; + } + } + rstreason = BANDLIM_RST_CLOSEDPORT; + goto dropwithresetnosock; + } + so = inp->inp_socket; + if (so == NULL) { + if (in_pcb_checkstate(inp, WNT_RELEASE, 1) == WNT_STOPUSING) + inp = NULL; // pretend we didn't find it +#if TEMPDEBUG + printf("tcp_input: no more socket for inp=%x\n", inp); #endif - goto dropwithreset; + goto dropnosock; } + +#ifdef __APPLE__ + /* + * Bogus state when listening port owned by SharedIP with loopback as the + * only configured interface: BlueBox does not filters loopback + */ + if (so == &tcbinfo.nat_dummy_socket) + goto drop; + +#endif + tcp_lock(so, 1, 2); + if (in_pcb_checkstate(inp, WNT_RELEASE, 1) == WNT_STOPUSING) { + tcp_unlock(so, 1, 2); + inp = NULL; // pretend we didn't find it + goto dropnosock; + } + tp = intotcpcb(inp); - if (tp == 0) + if (tp == 0) { + rstreason = BANDLIM_RST_CLOSEDPORT; goto dropwithreset; + } if (tp->t_state == TCPS_CLOSED) goto drop; @@ -700,34 +899,36 @@ findpcb: else tiwin = th->th_win; - so = inp->inp_socket; if (so->so_options & (SO_DEBUG|SO_ACCEPTCONN)) { #if TCPDEBUG if (so->so_options & SO_DEBUG) { ostate = tp->t_state; #if INET6 if (isipv6) - tcp_saveip._tcp_si6 = *ip6; + bcopy((char *)ip6, (char *)tcp_saveipgen, + sizeof(*ip6)); else - tcp_saveip._tcp_si4 = *ip; -#else /* INET6 */ - tcp_saveip = *ip; #endif /* INET6 */ - + bcopy((char *)ip, (char *)tcp_saveipgen, sizeof(*ip)); tcp_savetcp = *th; } #endif if (so->so_options & SO_ACCEPTCONN) { - register struct tcpcb *tp0 = tp; + register struct tcpcb *tp0 = tp; struct socket *so2; -#if IPSEC struct socket *oso; -#endif + struct sockaddr_storage from; #if INET6 struct inpcb *oinp = sotoinpcb(so); #endif /* INET6 */ + int ogencnt = so->so_gencnt; #if !IPSEC + /* + * Current IPsec implementation makes incorrect IPsec + * cache if this check is done here. + * So delay this until duplicated socket is created. + */ if ((thflags & (TH_RST|TH_ACK|TH_SYN)) != TH_SYN) { /* * Note: dropwithreset makes sure we don't @@ -735,29 +936,101 @@ findpcb: */ if (thflags & TH_ACK) { tcpstat.tcps_badsyn++; + rstreason = BANDLIM_RST_OPENPORT; goto dropwithreset; } goto drop; } #endif KERNEL_DEBUG(DBG_FNC_TCP_NEWCONN | DBG_FUNC_START,0,0,0,0,0); - so2 = sonewconn(so, 0); +#if 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 && !ip6_use_deprecated) { + struct in6_ifaddr *ia6; + if ((ia6 = ip6_getdstifaddr(m)) && + (ia6->ia6_flags & IN6_IFF_DEPRECATED)) { + tp = NULL; + rstreason = BANDLIM_RST_OPENPORT; + goto dropwithreset; + } + } +#endif + if (so->so_filt) { + if (isipv6) { + struct sockaddr_in6 *sin6 = (struct sockaddr_in6*)&from; + + sin6->sin6_len = sizeof(*sin6); + sin6->sin6_family = AF_INET6; + sin6->sin6_port = th->th_sport; + sin6->sin6_flowinfo = 0; + sin6->sin6_addr = ip6->ip6_src; + sin6->sin6_scope_id = 0; + } else { + struct sockaddr_in *sin = (struct sockaddr_in*)&from; + + sin->sin_len = sizeof(*sin); + sin->sin_family = AF_INET; + sin->sin_port = th->th_sport; + sin->sin_addr = ip->ip_src; + } + so2 = sonewconn(so, 0, (struct sockaddr*)&from); + } else { + so2 = sonewconn(so, 0, NULL); + } if (so2 == 0) { tcpstat.tcps_listendrop++; - so2 = sodropablereq(so); - if (so2) { - tcp_drop(sototcpcb(so2), ETIMEDOUT); - so2 = sonewconn(so, 0); + if (tcpdropdropablreq(so)) { + if (so->so_filt) + so2 = sonewconn(so, 0, (struct sockaddr*)&from); + else + so2 = sonewconn(so, 0, NULL); } - if (!so2) + if (!so2) goto drop; } -#if IPSEC + /* + * Make sure listening socket did not get closed during socket allocation, + * not only this is incorrect but it is know to cause panic + */ + if (so->so_gencnt != ogencnt) + goto drop; + oso = so; -#endif + tcp_unlock(so, 0, 0); /* Unlock but keep a reference on listener for now */ + so = so2; + tcp_lock(so, 1, 0); /* * This is ugly, but .... * @@ -775,21 +1048,18 @@ findpcb: if (isipv6) inp->in6p_laddr = ip6->ip6_dst; else { - if (ip6_mapped_addr_on) { - inp->inp_vflag &= ~INP_IPV6; - inp->inp_vflag |= INP_IPV4; - } + inp->inp_vflag &= ~INP_IPV6; + inp->inp_vflag |= INP_IPV4; #endif /* INET6 */ inp->inp_laddr = ip->ip_dst; #if INET6 } #endif /* INET6 */ - inp->inp_lport = th->th_dport; - if (in_pcbinshash(inp) != 0) { + if (in_pcbinshash(inp, 0) != 0) { /* - * Undo the assignments above if we failed to put - * the PCB on the hash lists. + * Undo the assignments above if we failed to + * put the PCB on the hash lists. */ #if INET6 if (isipv6) @@ -798,24 +1068,29 @@ findpcb: #endif /* INET6 */ inp->inp_laddr.s_addr = INADDR_ANY; inp->inp_lport = 0; + tcp_lock(oso, 0, 0); /* release ref on parent */ + tcp_unlock(oso, 1, 0); goto drop; } #if IPSEC /* - * from IPsec perspective, it is important to do it - * after making actual listening socket. - * otherwise, cached security association will bark. + * To avoid creating incorrectly cached IPsec + * association, this is need to be done here. * * Subject: (KAME-snap 748) * From: Wayne Knowles + * ftp://ftp.kame.net/pub/mail-list/snap-users/748 */ if ((thflags & (TH_RST|TH_ACK|TH_SYN)) != TH_SYN) { /* * Note: dropwithreset makes sure we don't * send a RST in response to a RST. */ + tcp_lock(oso, 0, 0); /* release ref on parent */ + tcp_unlock(oso, 1, 0); if (thflags & TH_ACK) { tcpstat.tcps_badsyn++; + rstreason = BANDLIM_RST_OPENPORT; goto dropwithreset; } goto drop; @@ -823,89 +1098,132 @@ findpcb: #endif #if INET6 if (isipv6) { - struct ip6_recvpktopts newopts; - - /* - * Inherit socket options from the listening - * socket. - * Note that in6p_inputopts are not (even - * should not be) copied, since it stores + /* + * Inherit socket options from the listening + * socket. + * Note that in6p_inputopts are not (even + * 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. - */ + * 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); + if (oinp->in6p_outputopts) + inp->in6p_outputopts = + ip6_copypktopts(oinp->in6p_outputopts, + M_NOWAIT); } else #endif /* INET6 */ inp->inp_options = ip_srcroute(); + tcp_lock(oso, 0, 0); #if IPSEC /* copy old policy into new socket's */ - if (ipsec_copy_policy(sotoinpcb(oso)->inp_sp, - inp->inp_sp)) - printf("tcp_input: could not copy policy\n"); + if (sotoinpcb(oso)->inp_sp) + { + int error = 0; + lck_mtx_lock(sadb_mutex); + /* Is it a security hole here to silently fail to copy the policy? */ + if (inp->inp_sp != NULL) + error = ipsec_init_policy(so, &inp->inp_sp); + if (error != 0 || ipsec_copy_policy(sotoinpcb(oso)->inp_sp, inp->inp_sp)) + printf("tcp_input: could not copy policy\n"); + lck_mtx_unlock(sadb_mutex); + } #endif - + tcp_unlock(oso, 1, 0); /* now drop the reference on the listener */ tp = intotcpcb(inp); tp->t_state = TCPS_LISTEN; - tp->t_flags |= tp0->t_flags & (TF_NOPUSH|TF_NOOPT); - + tp->t_flags |= tp0->t_flags & (TF_NOPUSH|TF_NOOPT|TF_NODELAY); + tp->t_inpcb->inp_ip_ttl = tp0->t_inpcb->inp_ip_ttl; /* Compute proper scaling value from buffer space */ while (tp->request_r_scale < TCP_MAX_WINSHIFT && - TCP_MAXWIN << tp->request_r_scale < so->so_rcv.sb_hiwat) + TCP_MAXWIN << tp->request_r_scale < + so->so_rcv.sb_hiwat) tp->request_r_scale++; KERNEL_DEBUG(DBG_FNC_TCP_NEWCONN | DBG_FUNC_END,0,0,0,0,0); } } -#if INET6 - /* save packet options if user wanted */ - if (isipv6 && (inp->in6p_flags & INP_CONTROLOPTS) != 0) { - struct ip6_recvpktopts opts6; - +#if 1 + lck_mtx_assert(((struct inpcb *)so->so_pcb)->inpcb_mtx, LCK_MTX_ASSERT_OWNED); +#endif + /* + * Radar 3529618 + * This is the second part of the MSS DoS prevention code (after + * minmss on the sending side) and it deals with too many too small + * tcp packets in a too short timeframe (1 second). + * + * For every full second we count the number of received packets + * and bytes. If we get a lot of packets per second for this connection + * (tcp_minmssoverload) we take a closer look at it and compute the + * average packet size for the past second. If that is less than + * tcp_minmss we get too many packets with very small payload which + * is not good and burdens our system (and every packet generates + * a wakeup to the process connected to our socket). We can reasonable + * expect this to be small packet DoS attack to exhaust our CPU + * cycles. + * + * Care has to be taken for the minimum packet overload value. This + * value defines the minimum number of packets per second before we + * start to worry. This must not be too low to avoid killing for + * example interactive connections with many small packets like + * telnet or SSH. + * + * + * Account for packet if payload packet, skip over ACK, etc. + * + * The packet per second count is done all the time and is also used + * by "DELAY_ACK" to detect streaming situations. + * + */ + if (tp->t_state == TCPS_ESTABLISHED && tlen > 0) { + if (tp->rcv_reset > tcp_now) { + tp->rcv_pps++; + tp->rcv_byps += tlen + off; /* - * Temporarily re-adjusting the mbuf before ip6_savecontrol(), - * which is necessary for FreeBSD only due to difference from - * other BSD stacks. - * XXX: we'll soon make a more natural fix after getting a - * consensus. + * Setting either tcp_minmssoverload or tcp_minmss to "0" disables + * the check. */ -#ifndef DEFER_MADJ - m->m_data -= hdroptlen; - m->m_len += hdroptlen; + if (tcp_minmss && tcp_minmssoverload && tp->rcv_pps > tcp_minmssoverload) { + if ((tp->rcv_byps / tp->rcv_pps) < tcp_minmss) { + char ipstrbuf[MAX_IPv6_STR_LEN]; + printf("too many small tcp packets from " + "%s:%u, av. %lubyte/packet, " + "dropping connection\n", +#ifdef INET6 + isipv6 ? + inet_ntop(AF_INET6, &inp->in6p_faddr, ipstrbuf, + sizeof(ipstrbuf)) : #endif - ip6_savecontrol(inp, ip6, m, &opts6, &inp->in6p_inputopts); - if (inp->in6p_inputopts) - ip6_update_recvpcbopt(inp->in6p_inputopts, &opts6); - if (opts6.head) { - if (sbappendcontrol(&inp->in6p_socket->so_rcv, - NULL, opts6.head) - == 0) - m_freem(opts6.head); + inet_ntop(AF_INET, &inp->inp_faddr, ipstrbuf, + sizeof(ipstrbuf)), + inp->inp_fport, + tp->rcv_byps / tp->rcv_pps); + tp = tcp_drop(tp, ECONNRESET); +/* tcpstat.tcps_minmssdrops++; */ + goto drop; + } + } + } else { + tp->rcv_reset = tcp_now + PR_SLOWHZ; + tp->rcv_pps = 1; + tp->rcv_byps = tlen + off; } -#ifndef DEFER_MADJ - m->m_data += hdroptlen; /* XXX */ - m->m_len -= hdroptlen; /* XXX */ -#endif } -#endif /* INET6 */ - + /* * Segment received on connection. * Reset idle time and keep-alive timer. */ - tp->t_idle = 0; + tp->t_rcvtime = 0; if (TCPS_HAVEESTABLISHED(tp->t_state)) - tp->t_timer[TCPT_KEEP] = tcp_keepidle; + tp->t_timer[TCPT_KEEP] = TCP_KEEPIDLE(tp); /* * Process options if not in LISTEN state, @@ -913,8 +1231,27 @@ findpcb: */ if (tp->t_state != TCPS_LISTEN && optp) tcp_dooptions(tp, optp, optlen, th, &to); - if (th->th_flags & TH_SYN) - tcp_mss(tp, to.to_maxseg, isipv6); /* sets t_maxseg */ + + + if (tp->t_state == TCPS_SYN_SENT && (thflags & TH_SYN)) { + if (to.to_flags & TOF_SCALE) { + tp->t_flags |= TF_RCVD_SCALE; + tp->requested_s_scale = to.to_requested_s_scale; + } + if (to.to_flags & TOF_TS) { + tp->t_flags |= TF_RCVD_TSTMP; + tp->ts_recent = to.to_tsval; + tp->ts_recent_age = tcp_now; + } + if (to.to_flags & TOF_MSS) + tcp_mss(tp, to.to_mss); + if (tp->sack_enable) { + if (!(to.to_flags & TOF_SACK)) + tp->sack_enable = 0; + else + tp->t_flags |= TF_SACK_PERMIT; + } + } /* * Header prediction: check for the two common cases @@ -936,15 +1273,8 @@ findpcb: if (tp->t_state == TCPS_ESTABLISHED && (thflags & (TH_SYN|TH_FIN|TH_RST|TH_URG|TH_ACK)) == TH_ACK && ((tp->t_flags & (TF_NEEDSYN|TF_NEEDFIN)) == 0) && - ((to.to_flag & TOF_TS) == 0 || + ((to.to_flags & TOF_TS) == 0 || TSTMP_GEQ(to.to_tsval, tp->ts_recent)) && - /* - * Using the CC option is compulsory if once started: - * the segment is OK if no T/TCP was negotiated or - * if the segment has a CC option equal to CCrecv - */ - ((tp->t_flags & (TF_REQ_CC|TF_RCVD_CC)) != (TF_REQ_CC|TF_RCVD_CC) || - ((to.to_flag & TOF_CC) != 0 && to.to_cc == tp->cc_recv)) && th->th_seq == tp->rcv_nxt && tiwin && tiwin == tp->snd_wnd && tp->snd_nxt == tp->snd_max) { @@ -955,38 +1285,61 @@ findpcb: * NOTE that the test is modified according to the latest * proposal of the tcplw@cray.com list (Braden 1993/04/26). */ - if ((to.to_flag & TOF_TS) != 0 && + if ((to.to_flags & TOF_TS) != 0 && SEQ_LEQ(th->th_seq, tp->last_ack_sent)) { tp->ts_recent_age = tcp_now; tp->ts_recent = to.to_tsval; } - if (tilen == 0) { + if (tlen == 0) { if (SEQ_GT(th->th_ack, tp->snd_una) && SEQ_LEQ(th->th_ack, tp->snd_max) && tp->snd_cwnd >= tp->snd_wnd && - tp->t_dupacks < tcprexmtthresh) { + ((!tcp_do_newreno && !tp->sack_enable && + tp->t_dupacks < tcprexmtthresh) || + ((tcp_do_newreno || tp->sack_enable) && + !IN_FASTRECOVERY(tp) && to.to_nsacks == 0 && + TAILQ_EMPTY(&tp->snd_holes)))) { /* * this is a pure ack for outstanding data. */ ++tcpstat.tcps_predack; - if ((to.to_flag & TOF_TS) != 0) + /* + * "bad retransmit" recovery + */ + if (tp->t_rxtshift == 1 && + tcp_now < tp->t_badrxtwin) { + 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); + tp->snd_nxt = tp->snd_max; + tp->t_badrxtwin = 0; + } + if (((to.to_flags & TOF_TS) != 0) && (to.to_tsecr != 0)) /* Makes sure we already have a TS */ tcp_xmit_timer(tp, tcp_now - to.to_tsecr + 1); - else if (tp->t_rtt && + else if (tp->t_rtttime && SEQ_GT(th->th_ack, tp->t_rtseq)) - tcp_xmit_timer(tp, tp->t_rtt); + tcp_xmit_timer(tp, tp->t_rtttime); acked = th->th_ack - tp->snd_una; tcpstat.tcps_rcvackpack++; tcpstat.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; 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); -#if INET6 - /* some progress has been done */ - if (isipv6) - ND6_HINT(tp); -#endif + ND6_HINT(tp); /* some progress has been done */ /* * If all outstanding data are acked, stop @@ -1002,42 +1355,65 @@ findpcb: else if (tp->t_timer[TCPT_PERSIST] == 0) tp->t_timer[TCPT_REXMT] = tp->t_rxtcur; - if (so->so_snd.sb_cc) + sowwakeup(so); /* has to be done with socket lock held */ + if ((so->so_snd.sb_cc) || (tp->t_flags & TF_ACKNOW)) (void) tcp_output(tp); - sowwakeup(so); + tcp_unlock(so, 1, 0); KERNEL_DEBUG(DBG_FNC_TCP_INPUT | DBG_FUNC_END,0,0,0,0,0); return; } } else if (th->th_ack == tp->snd_una && - tp->segq.lh_first == NULL && - tilen <= sbspace(&so->so_rcv)) { + LIST_EMPTY(&tp->t_segq) && + tlen <= sbspace(&so->so_rcv)) { /* * this is a pure, in-sequence data packet * with nothing on the reassembly queue and * we have enough buffer space to take it. */ + /* Clean receiver SACK report if present */ + if (tp->sack_enable && tp->rcv_numsacks) + tcp_clean_sackreport(tp); ++tcpstat.tcps_preddat; - tp->rcv_nxt += tilen; + 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.tcps_rcvpack++; - tcpstat.tcps_rcvbyte += tilen; + tcpstat.tcps_rcvbyte += tlen; + ND6_HINT(tp); /* some progress has been done */ + /* + * Add data to socket buffer. + */ + m_adj(m, drop_hdrlen); /* delayed header drop */ + if (sbappend(&so->so_rcv, m)) + sorwakeup(so); #if INET6 - /* some progress has been done */ - if (isipv6) - ND6_HINT(tp); -#endif - sbappend(&so->so_rcv, m); - KERNEL_DEBUG(DBG_LAYER_END, ((th->th_dport << 16) | th->th_sport), - (((th->th_src.s_addr & 0xffff) << 16) | (th->th_dst.s_addr & 0xffff)), - th->th_seq, th->th_ack, th->th_win); - if (tcp_delack_enabled) { - if (last_active_conn_count > DELACK_BITMASK_THRESH) - TCP_DELACK_BITSET(tp->t_inpcb->hash_element); + if (isipv6) { + KERNEL_DEBUG(DBG_LAYER_END, ((th->th_dport << 16) | th->th_sport), + (((ip6->ip6_src.s6_addr16[0]) << 16) | (ip6->ip6_dst.s6_addr16[0])), + th->th_seq, th->th_ack, th->th_win); + } + else +#endif + { + KERNEL_DEBUG(DBG_LAYER_END, ((th->th_dport << 16) | th->th_sport), + (((ip->ip_src.s_addr & 0xffff) << 16) | (ip->ip_dst.s_addr & 0xffff)), + th->th_seq, th->th_ack, th->th_win); + } + if (DELAY_ACK(tp)) { tp->t_flags |= TF_DELACK; } else { tp->t_flags |= TF_ACKNOW; tcp_output(tp); } - sorwakeup(so); + tcp_unlock(so, 1, 0); KERNEL_DEBUG(DBG_FNC_TCP_INPUT | DBG_FUNC_END,0,0,0,0,0); return; } @@ -1049,11 +1425,18 @@ findpcb: * Receive window is amount of space in rcv queue, * but not less than advertised window. */ +#if 1 + lck_mtx_assert(((struct inpcb *)so->so_pcb)->inpcb_mtx, LCK_MTX_ASSERT_OWNED); +#endif { int win; win = sbspace(&so->so_rcv); if (win < 0) win = 0; + else { /* clip rcv window to 4K for modems */ + if (tp->t_flags & TF_SLOWLINK && slowlink_wsize > 0) + win = min(win, slowlink_wsize); + } tp->rcv_wnd = imax(win, (int)(tp->rcv_adv - tp->rcv_nxt)); } @@ -1075,14 +1458,19 @@ findpcb: */ case TCPS_LISTEN: { register struct sockaddr_in *sin; -#if 0 +#if INET6 register struct sockaddr_in6 *sin6; #endif - if (thflags & TH_RST) +#if 1 + lck_mtx_assert(((struct inpcb *)so->so_pcb)->inpcb_mtx, LCK_MTX_ASSERT_OWNED); +#endif + if (thflags & TH_RST) goto drop; - if (thflags & TH_ACK) + if (thflags & TH_ACK) { + rstreason = BANDLIM_RST_OPENPORT; goto dropwithreset; + } if ((thflags & TH_SYN) == 0) goto drop; if (th->th_dport == th->th_sport) { @@ -1096,29 +1484,30 @@ findpcb: if (ip->ip_dst.s_addr == ip->ip_src.s_addr) goto drop; } - + /* + * RFC1122 4.2.3.10, p. 104: discard bcast/mcast SYN + * in_broadcast() should never return true on a received + * packet with M_BCAST not set. + * + * Packets with a multicast source address should also + * be discarded. + */ + if (m->m_flags & (M_BCAST|M_MCAST)) + goto drop; #if INET6 if (isipv6) { - if (m->m_flags & (M_BCAST|M_MCAST) || - IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) + if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || + IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) goto drop; -#if 1 - /* - * Perhaps this should be a call/macro - * to a function like in6_pcbconnect(), but almost - * all of the checks have been done: we know - * that the association is unique, and the - * local address is always set here. - */ - if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) - inp->in6p_laddr = ip6->ip6_dst; - inp->in6p_faddr = ip6->ip6_src; - inp->inp_fport = th->th_sport; - - /* TODO: flowinfo initialization */ - - in_pcbrehash(inp); -#else + } else +#endif + 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; +#if INET6 + if (isipv6) { MALLOC(sin6, struct sockaddr_in6 *, sizeof *sin6, M_SONAME, M_NOWAIT); if (sin6 == NULL) @@ -1132,26 +1521,20 @@ findpcb: if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) inp->in6p_laddr = ip6->ip6_dst; if (in6_pcbconnect(inp, (struct sockaddr *)sin6, - &proc0)) { + proc0)) { inp->in6p_laddr = laddr6; FREE(sin6, M_SONAME); goto drop; } FREE(sin6, M_SONAME); + } else +#endif + { +#if 1 + lck_mtx_assert(((struct inpcb *)so->so_pcb)->inpcb_mtx, LCK_MTX_ASSERT_OWNED); #endif - } - else { -#endif /* INET6 */ - /* - * RFC1122 4.2.3.10, p. 104: discard bcast/mcast SYN - * in_broadcast() should never return true on a received - * packet with M_BCAST not set. - */ - if (m->m_flags & (M_BCAST|M_MCAST) || - IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) - goto drop; MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME, - M_NOWAIT); + M_NOWAIT); if (sin == NULL) goto drop; sin->sin_family = AF_INET; @@ -1162,121 +1545,38 @@ findpcb: laddr = inp->inp_laddr; if (inp->inp_laddr.s_addr == INADDR_ANY) inp->inp_laddr = ip->ip_dst; - if (in_pcbconnect(inp, (struct sockaddr *)sin, &proc0)) { + if (in_pcbconnect(inp, (struct sockaddr *)sin, proc0)) { inp->inp_laddr = laddr; FREE(sin, M_SONAME); goto drop; } FREE(sin, M_SONAME); -#if INET6 } -#endif /* INET6 */ - tp->t_template = tcp_template(tp); - if (tp->t_template == 0) { - tp = tcp_drop(tp, ENOBUFS); - dropsocket = 0; /* socket is already gone */ - goto drop; - } - if ((taop = tcp_gettaocache(inp)) == NULL) { - taop = &tao_noncached; - bzero(taop, sizeof(*taop)); - } tcp_dooptions(tp, optp, optlen, th, &to); - if (th->th_flags & TH_SYN) - tcp_mss(tp, to.to_maxseg, isipv6); /* sets t_maxseg */ + + if (tp->sack_enable) { + if (!(to.to_flags & TOF_SACK)) + tp->sack_enable = 0; + else + tp->t_flags |= TF_SACK_PERMIT; + } + if (iss) tp->iss = iss; - else - tp->iss = tcp_iss; - tcp_iss += TCP_ISSINCR/4; + else { + tp->iss = tcp_new_isn(tp); + } tp->irs = th->th_seq; tcp_sendseqinit(tp); tcp_rcvseqinit(tp); + tp->snd_recover = tp->snd_una; /* * Initialization of the tcpcb for transaction; * set SND.WND = SEG.WND, * initialize CCsend and CCrecv. */ tp->snd_wnd = tiwin; /* initial send-window */ - tp->cc_send = CC_INC(tcp_ccgen); - tp->cc_recv = to.to_cc; - /* - * Perform TAO test on incoming CC (SEG.CC) option, if any. - * - compare SEG.CC against cached CC from the same host, - * if any. - * - if SEG.CC > chached value, SYN must be new and is accepted - * immediately: save new CC in the cache, mark the socket - * connected, enter ESTABLISHED state, turn on flag to - * send a SYN in the next segment. - * A virtual advertised window is set in rcv_adv to - * initialize SWS prevention. Then enter normal segment - * processing: drop SYN, process data and FIN. - * - otherwise do a normal 3-way handshake. - */ - if ((to.to_flag & TOF_CC) != 0) { - if (((tp->t_flags & TF_NOPUSH) != 0) && - taop->tao_cc != 0 && CC_GT(to.to_cc, taop->tao_cc)) { - - taop->tao_cc = to.to_cc; - if (tp->t_state != TCPS_ESTABLISHED) - current_active_connections++; - - tp->t_state = TCPS_ESTABLISHED; - - /* - * If there is a FIN, or if there is data and the - * connection is local, then delay SYN,ACK(SYN) in - * the hope of piggy-backing it on a response - * segment. Otherwise must send ACK now in case - * the other side is slow starting. - */ - if (tcp_delack_enabled && - ((thflags & TH_FIN) || - (tilen != 0 && -#if INET6 - (isipv6 && in6_localaddr(&inp->in6p_faddr)) - || - (!isipv6 && -#endif /* INET6 */ - in_localaddr(inp->inp_faddr) -#if INET6 - ) -#endif /* INET6 */ - ))) { - if (last_active_conn_count > DELACK_BITMASK_THRESH) - TCP_DELACK_BITSET(tp->t_inpcb->hash_element); - - tp->t_flags |= (TF_DELACK | TF_NEEDSYN); - } - else - tp->t_flags |= (TF_ACKNOW | TF_NEEDSYN); - - /* - * Limit the `virtual advertised window' to TCP_MAXWIN - * here. Even if we requested window scaling, it will - * become effective only later when our SYN is acked. - */ - tp->rcv_adv += min(tp->rcv_wnd, TCP_MAXWIN); - tcpstat.tcps_connects++; - soisconnected(so); - tp->t_timer[TCPT_KEEP] = tcp_keepinit; - dropsocket = 0; /* committed to socket */ - tcpstat.tcps_accepts++; - goto trimthenstep6; - } - /* else do standard 3-way handshake */ - } else { - /* - * No CC option, but maybe CC.NEW: - * invalidate cached value. - */ - taop->tao_cc = 0; - } - /* - * TAO test failed or there was no CC option, - * do a standard 3-way handshake. - */ tp->t_flags |= TF_ACKNOW; tp->t_state = TCPS_SYN_RECEIVED; tp->t_timer[TCPT_KEEP] = tcp_keepinit; @@ -1292,8 +1592,10 @@ findpcb: case TCPS_SYN_RECEIVED: if ((thflags & TH_ACK) && (SEQ_LEQ(th->th_ack, tp->snd_una) || - SEQ_GT(th->th_ack, tp->snd_max))) + SEQ_GT(th->th_ack, tp->snd_max))) { + rstreason = BANDLIM_RST_OPENPORT; goto dropwithreset; + } break; /* @@ -1309,26 +1611,11 @@ findpcb: * continue processing rest of data/controls, beginning with URG */ case TCPS_SYN_SENT: - if ((taop = tcp_gettaocache(inp)) == NULL) { - taop = &tao_noncached; - bzero(taop, sizeof(*taop)); - } - if ((thflags & TH_ACK) && (SEQ_LEQ(th->th_ack, tp->iss) || SEQ_GT(th->th_ack, tp->snd_max))) { - /* - * If we have a cached CCsent for the remote host, - * hence we haven't just crashed and restarted, - * do not send a RST. This may be a retransmission - * from the other side after our earlier ACK was lost. - * Our new SYN, when it arrives, will serve as the - * needed ACK. - */ - if (taop->tao_ccsent != 0) - goto drop; - else - goto dropwithreset; + rstreason = BANDLIM_UNLIMITED; + goto dropwithreset; } if (thflags & TH_RST) { if (thflags & TH_ACK) { @@ -1340,27 +1627,10 @@ findpcb: if ((thflags & TH_SYN) == 0) goto drop; tp->snd_wnd = th->th_win; /* initial send window */ - tp->cc_recv = to.to_cc; /* foreign CC */ tp->irs = th->th_seq; tcp_rcvseqinit(tp); if (thflags & TH_ACK) { - /* - * Our SYN was acked. If segment contains CC.ECHO - * option, check it to make sure this segment really - * matches our SYN. If not, just drop it as old - * duplicate, but send an RST if we're still playing - * by the old rules. If no CC.ECHO option, make sure - * we don't get fooled into using T/TCP. - */ - if (to.to_flag & TOF_CCECHO) { - if (tp->cc_send != to.to_ccecho) - if (taop->tao_ccsent != 0) - goto drop; - else - goto dropwithreset; - } else - tp->t_flags &= ~TF_RCVD_CC; tcpstat.tcps_connects++; soisconnected(so); /* Do window scaling on this connection? */ @@ -1369,38 +1639,32 @@ findpcb: tp->snd_scale = tp->requested_s_scale; tp->rcv_scale = tp->request_r_scale; } - /* Segment is acceptable, update cache if undefined. */ - if (taop->tao_ccsent == 0) - taop->tao_ccsent = to.to_ccecho; - tp->rcv_adv += tp->rcv_wnd; 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 (tcp_delack_enabled && tilen != 0) { - if (last_active_conn_count > DELACK_BITMASK_THRESH) - TCP_DELACK_BITSET(tp->t_inpcb->hash_element); + if (DELAY_ACK(tp) && tlen != 0) { tp->t_flags |= TF_DELACK; } - else + else { tp->t_flags |= TF_ACKNOW; + } /* * Received in SYN_SENT[*] state. * Transitions: * SYN_SENT --> ESTABLISHED * SYN_SENT* --> FIN_WAIT_1 */ + tp->t_starttime = 0; if (tp->t_flags & TF_NEEDFIN) { tp->t_state = TCPS_FIN_WAIT_1; tp->t_flags &= ~TF_NEEDFIN; thflags &= ~TH_SYN; } else { - if (tp->t_state != TCPS_ESTABLISHED) - current_active_connections++; tp->t_state = TCPS_ESTABLISHED; - tp->t_timer[TCPT_KEEP] = tcp_keepidle; + tp->t_timer[TCPT_KEEP] = TCP_KEEPIDLE(tp); } } else { /* @@ -1409,37 +1673,12 @@ findpcb: * 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; - tp->t_timer[TCPT_REXMT] = 0; - if (to.to_flag & TOF_CC) { - if (taop->tao_cc != 0 && - CC_GT(to.to_cc, taop->tao_cc)) { - /* - * update cache and make transition: - * SYN-SENT -> ESTABLISHED* - * SYN-SENT* -> FIN-WAIT-1* - */ - taop->tao_cc = to.to_cc; - if (tp->t_flags & TF_NEEDFIN) { - tp->t_state = TCPS_FIN_WAIT_1; - tp->t_flags &= ~TF_NEEDFIN; - } else { - if (tp->t_state != TCPS_ESTABLISHED) - current_active_connections++; - tp->t_state = TCPS_ESTABLISHED; - tp->t_timer[TCPT_KEEP] = tcp_keepidle; - } - tp->t_flags |= TF_NEEDSYN; - } else - tp->t_state = TCPS_SYN_RECEIVED; - } else { - /* CC.NEW or no option => invalidate cache */ - taop->tao_cc = 0; - tp->t_state = TCPS_SYN_RECEIVED; - } + * SYN-SENT* -> SYN-RECEIVED* + */ + tp->t_flags |= TF_ACKNOW; + tp->t_timer[TCPT_REXMT] = 0; + tp->t_state = TCPS_SYN_RECEIVED; + } trimthenstep6: @@ -1449,10 +1688,10 @@ trimthenstep6: * dropping FIN if necessary. */ th->th_seq++; - if (tilen > tp->rcv_wnd) { - todrop = tilen - tp->rcv_wnd; + if (tlen > tp->rcv_wnd) { + todrop = tlen - tp->rcv_wnd; m_adj(m, -todrop); - tilen = tp->rcv_wnd; + tlen = tp->rcv_wnd; thflags &= ~TH_FIN; tcpstat.tcps_rcvpackafterwin++; tcpstat.tcps_rcvbyteafterwin += todrop; @@ -1471,34 +1710,24 @@ trimthenstep6: goto step6; /* * If the state is LAST_ACK or CLOSING or TIME_WAIT: - * if segment contains a SYN and CC [not CC.NEW] option: - * if state == TIME_WAIT and connection duration > MSL, - * drop packet and send RST; - * - * if SEG.CC > CCrecv then is new SYN, and can implicitly - * ack the FIN (and data) in retransmission queue. - * Complete close and delete TCPCB. Then reprocess - * segment, hoping to find new TCPCB in LISTEN state; + * do normal processing. * - * else must be old SYN; drop it. - * else do normal processing. + * NB: Leftover from RFC1644 T/TCP. Cases to be reused later. */ case TCPS_LAST_ACK: case TCPS_CLOSING: case TCPS_TIME_WAIT: - if ((thflags & TH_SYN) && - (to.to_flag & TOF_CC) && tp->cc_recv != 0) { - if (tp->t_state == TCPS_TIME_WAIT && - tp->t_duration > TCPTV_MSL) - goto dropwithreset; - if (CC_GT(to.to_cc, tp->cc_recv)) { - tp = tcp_close(tp); - goto findpcb; - } - else - goto drop; - } break; /* continue normal processing */ + + /* Received a SYN while connection is already established. + * This is a "half open connection and other anomalies" described + * in RFC793 page 34, send an ACK so the remote reset the connection + * or recovers by adjusting its sequence numberering + */ + case TCPS_ESTABLISHED: + if (thflags & TH_SYN) + goto dropafterack; + break; } /* @@ -1525,12 +1754,16 @@ trimthenstep6: * 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. - * Also, it does not make sense to allow reset segments with - * sequence numbers greater than last_ack_sent to be processed - * since these sequence numbers are just the acknowledgement - * numbers in our outgoing packets being echoed back at us, - * and these acknowledgement numbers are monotonically - * increasing. + * 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 intial reset * segment sequence numbers will be to the left of last_ack_sent, * but they will eventually catch up. @@ -1551,16 +1784,24 @@ trimthenstep6: * SYN_RECEIVED STATE: * If passive open, return to LISTEN state. * If active open, inform user that connection was refused. - * ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT STATES: + * ESTABLISHED, FIN_WAIT_1, FIN_WAIT_2, CLOSE_WAIT STATES: * Inform user that connection was reset, and close tcb. - * CLOSING, LAST_ACK, TIME_WAIT STATES + * CLOSING, LAST_ACK STATES: * Close the tcb. - * TIME_WAIT state: + * TIME_WAIT STATE: * Drop the segment - see Stevens, vol. 2, p. 964 and * RFC 1337. + * + * Radar 4776325: Allows for the case where we ACKed the FIN but + * there is already a RST in flight from the peer. + * In that case, accept the RST for non-established + * state if it's one off from last_ack_sent. */ if (thflags & TH_RST) { - if (tp->last_ack_sent == th->th_seq) { + if ((SEQ_GEQ(th->th_seq, tp->last_ack_sent) && + SEQ_LT(th->th_seq, tp->last_ack_sent + tp->rcv_wnd)) || + (tp->rcv_wnd == 0 && + (tp->last_ack_sent == th->th_seq)) || ((tp->last_ack_sent - 1) == th->th_seq)) { switch (tp->t_state) { case TCPS_SYN_RECEIVED: @@ -1568,9 +1809,11 @@ trimthenstep6: goto close; case TCPS_ESTABLISHED: + if (tp->last_ack_sent != th->th_seq) { + goto drop; + } case TCPS_FIN_WAIT_1: case TCPS_CLOSE_WAIT: - current_active_connections--; /* Drop through ... */ @@ -1585,7 +1828,6 @@ trimthenstep6: case TCPS_CLOSING: case TCPS_LAST_ACK: - current_active_connections--; tp = tcp_close(tp); break; @@ -1596,11 +1838,14 @@ trimthenstep6: goto drop; } +#if 1 + lck_mtx_assert(((struct inpcb *)so->so_pcb)->inpcb_mtx, LCK_MTX_ASSERT_OWNED); +#endif /* * RFC 1323 PAWS: If we have a timestamp reply on this segment * and it's less than ts_recent, drop it. */ - if ((to.to_flag & TOF_TS) != 0 && tp->ts_recent && + 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. */ @@ -1619,22 +1864,14 @@ trimthenstep6: tp->ts_recent = 0; } else { tcpstat.tcps_rcvduppack++; - tcpstat.tcps_rcvdupbyte += tilen; + tcpstat.tcps_rcvdupbyte += tlen; tcpstat.tcps_pawsdrop++; - goto dropafterack; + if (tlen) + goto dropafterack; + goto drop; } } - /* - * T/TCP mechanism - * If T/TCP was negotiated and the segment doesn't have CC, - * or if its CC is wrong then drop the segment. - * RST segments do not have to comply with this. - */ - if ((tp->t_flags & (TF_REQ_CC|TF_RCVD_CC)) == (TF_REQ_CC|TF_RCVD_CC) && - ((to.to_flag & TOF_CC) == 0 || tp->cc_recv != to.to_cc)) - goto dropafterack; - /* * In the SYN-RECEIVED state, validate that the packet belongs to * this connection before trimming the data to fit the receive @@ -1642,8 +1879,10 @@ trimthenstep6: * 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)) + 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) { @@ -1659,8 +1898,8 @@ trimthenstep6: /* * Following if statement from Stevens, vol. 2, p. 960. */ - if (todrop > tilen - || (todrop == tilen && (thflags & TH_FIN) == 0)) { + 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 @@ -1673,16 +1912,16 @@ trimthenstep6: * But keep on processing for RST or ACK. */ tp->t_flags |= TF_ACKNOW; - todrop = tilen; + todrop = tlen; tcpstat.tcps_rcvduppack++; tcpstat.tcps_rcvdupbyte += todrop; } else { tcpstat.tcps_rcvpartduppack++; tcpstat.tcps_rcvpartdupbyte += todrop; } - m_adj(m, todrop); + drop_hdrlen += todrop; /* drop from the top afterwards */ th->th_seq += todrop; - tilen -= todrop; + tlen -= todrop; if (th->th_urp > todrop) th->th_urp -= todrop; else { @@ -1696,9 +1935,10 @@ trimthenstep6: * user processes are gone, then RST the other end. */ if ((so->so_state & SS_NOFDREF) && - tp->t_state > TCPS_CLOSE_WAIT && tilen) { + tp->t_state > TCPS_CLOSE_WAIT && tlen) { tp = tcp_close(tp); tcpstat.tcps_rcvafterclose++; + rstreason = BANDLIM_UNLIMITED; goto dropwithreset; } @@ -1706,11 +1946,11 @@ trimthenstep6: * If segment ends after window, drop trailing data * (and PUSH and FIN); if nothing left, just ACK. */ - todrop = (th->th_seq+tilen) - (tp->rcv_nxt+tp->rcv_wnd); + todrop = (th->th_seq+tlen) - (tp->rcv_nxt+tp->rcv_wnd); if (todrop > 0) { tcpstat.tcps_rcvpackafterwin++; - if (todrop >= tilen) { - tcpstat.tcps_rcvbyteafterwin += tilen; + if (todrop >= tlen) { + tcpstat.tcps_rcvbyteafterwin += tlen; /* * If a new connection request is received * while in TIME_WAIT, drop the old connection @@ -1720,8 +1960,9 @@ trimthenstep6: if (thflags & TH_SYN && tp->t_state == TCPS_TIME_WAIT && SEQ_GT(th->th_seq, tp->rcv_nxt)) { - iss = tp->rcv_nxt + TCP_ISSINCR; + iss = tcp_new_isn(tp); tp = tcp_close(tp); + tcp_unlock(so, 1, 0); goto findpcb; } /* @@ -1739,18 +1980,32 @@ trimthenstep6: } else tcpstat.tcps_rcvbyteafterwin += todrop; m_adj(m, -todrop); - tilen -= todrop; + tlen -= todrop; thflags &= ~(TH_PUSH|TH_FIN); } /* * If last ACK falls within this segment's sequence numbers, * record its timestamp. - * NOTE that the test is modified according to the latest - * proposal of the tcplw@cray.com list (Braden 1993/04/26). + * 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_flag & TOF_TS) != 0 && - SEQ_LEQ(th->th_seq, tp->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 = tcp_now; tp->ts_recent = to.to_tsval; } @@ -1761,6 +2016,7 @@ trimthenstep6: */ if (thflags & TH_SYN) { tp = tcp_drop(tp, ECONNRESET); + rstreason = BANDLIM_UNLIMITED; postevent(so, 0, EV_RESET); goto dropwithreset; } @@ -1792,7 +2048,6 @@ trimthenstep6: tcpstat.tcps_connects++; soisconnected(so); - current_active_connections++; /* Do window scaling? */ if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) == @@ -1800,36 +2055,28 @@ trimthenstep6: tp->snd_scale = tp->requested_s_scale; tp->rcv_scale = tp->request_r_scale; } - /* - * Upon successful completion of 3-way handshake, - * update cache.CC if it was undefined, pass any queued - * data to the user, and advance state appropriately. - */ - if ((taop = tcp_gettaocache(inp)) != NULL && - taop->tao_cc == 0) - taop->tao_cc = tp->cc_recv; - /* * Make transitions: * SYN-RECEIVED -> ESTABLISHED * SYN-RECEIVED* -> FIN-WAIT-1 */ + tp->t_starttime = 0; if (tp->t_flags & TF_NEEDFIN) { tp->t_state = TCPS_FIN_WAIT_1; tp->t_flags &= ~TF_NEEDFIN; } else { tp->t_state = TCPS_ESTABLISHED; - tp->t_timer[TCPT_KEEP] = tcp_keepidle; + tp->t_timer[TCPT_KEEP] = TCP_KEEPIDLE(tp); } /* * If segment contains data or ACK, will call tcp_reass() * later; if not, do so now to pass queued data to user. */ - if (tilen == 0 && (thflags & TH_FIN) == 0) + if (tlen == 0 && (thflags & TH_FIN) == 0) (void) tcp_reass(tp, (struct tcphdr *)0, 0, - (struct mbuf *)0, isipv6); + (struct mbuf *)0); tp->snd_wl1 = th->th_seq - 1; - /* fall into ... */ + /* FALLTHROUGH */ /* * In ESTABLISHED state: drop duplicate ACKs; ACK out of range @@ -1846,9 +2093,15 @@ trimthenstep6: case TCPS_CLOSING: case TCPS_LAST_ACK: case TCPS_TIME_WAIT: - + if (SEQ_GT(th->th_ack, tp->snd_max)) { + tcpstat.tcps_rcvacktoomuch++; + goto dropafterack; + } + if (tp->sack_enable && + (to.to_nsacks > 0 || !TAILQ_EMPTY(&tp->snd_holes))) + tcp_sack_doack(tp, &to, th->th_ack); if (SEQ_LEQ(th->th_ack, tp->snd_una)) { - if (tilen == 0 && tiwin == tp->snd_wnd) { + if (tlen == 0 && tiwin == tp->snd_wnd) { tcpstat.tcps_rcvdupack++; /* * If we have outstanding data (other than @@ -1877,29 +2130,76 @@ trimthenstep6: if (tp->t_timer[TCPT_REXMT] == 0 || th->th_ack != tp->snd_una) tp->t_dupacks = 0; - else if (++tp->t_dupacks == tcprexmtthresh) { + else if (++tp->t_dupacks > tcprexmtthresh || + ((tcp_do_newreno || tp->sack_enable) && + IN_FASTRECOVERY(tp))) { + if (tp->sack_enable && IN_FASTRECOVERY(tp)) { + 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; - u_int win = - min(tp->snd_wnd, tp->snd_cwnd) / 2 / - tp->t_maxseg; + u_int win; + /* + * 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->sack_enable) { + if (IN_FASTRECOVERY(tp)) { + tp->t_dupacks = 0; + break; + } + } else if (tcp_do_newreno) { + if (SEQ_LEQ(th->th_ack, + tp->snd_recover)) { + tp->t_dupacks = 0; + break; + } + } + win = min(tp->snd_wnd, tp->snd_cwnd) / + 2 / tp->t_maxseg; if (win < 2) win = 2; tp->snd_ssthresh = win * tp->t_maxseg; + ENTER_FASTRECOVERY(tp); + tp->snd_recover = tp->snd_max; tp->t_timer[TCPT_REXMT] = 0; - tp->t_rtt = 0; + tp->t_rtttime = 0; + if (tp->sack_enable) { + tcpstat.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); tp->snd_cwnd = tp->snd_ssthresh + - tp->t_maxseg * tp->t_dupacks; + tp->t_maxseg * tp->t_dupacks; if (SEQ_GT(onxt, tp->snd_nxt)) tp->snd_nxt = onxt; goto drop; - } else if (tp->t_dupacks > tcprexmtthresh) { - tp->snd_cwnd += tp->t_maxseg; - (void) tcp_output(tp); - goto drop; } } else tp->t_dupacks = 0; @@ -1909,16 +2209,41 @@ trimthenstep6: * If the congestion window was inflated to account * for the other side's cached packets, retract it. */ - if (tp->t_dupacks >= tcprexmtthresh && - tp->snd_cwnd > tp->snd_ssthresh) - tp->snd_cwnd = tp->snd_ssthresh; - tp->t_dupacks = 0; - if (SEQ_GT(th->th_ack, tp->snd_max)) { - tcpstat.tcps_rcvacktoomuch++; - goto dropafterack; + if (tcp_do_newreno || tp->sack_enable) { + if (IN_FASTRECOVERY(tp)) { + if (SEQ_LT(th->th_ack, tp->snd_recover)) { + if (tp->sack_enable) + tcp_sack_partialack(tp, th); + else + tcp_newreno_partial_ack(tp, th); + } else { + /* + * Out of fast recovery. + * Window inflation should have left us + * with approximately snd_ssthresh + * outstanding data. + * But in case we would be inclined to + * send a burst, better to do it via + * the slow start mechanism. + */ + if (SEQ_GT(th->th_ack + + tp->snd_ssthresh, + tp->snd_max)) + tp->snd_cwnd = tp->snd_max - + th->th_ack + + tp->t_maxseg; + else + tp->snd_cwnd = tp->snd_ssthresh; + } + } + } else { + if (tp->t_dupacks >= tcprexmtthresh && + tp->snd_cwnd > tp->snd_ssthresh) + tp->snd_cwnd = tp->snd_ssthresh; } + tp->t_dupacks = 0; /* - * If we reach this point, ACK is not a duplicate, + * If we reach this point, ACK is not a duplicate, * i.e., it ACKs something we sent. */ if (tp->t_flags & TF_NEEDSYN) { @@ -1944,6 +2269,23 @@ process_ACK: tcpstat.tcps_rcvackpack++; tcpstat.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 && tcp_now < tp->t_badrxtwin) { + 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); + tp->snd_nxt = tp->snd_max; + tp->t_badrxtwin = 0; /* XXX probably not required */ + } + /* * If we have a timestamp reply, update smoothed * round trip time. If no timestamp is present but @@ -1952,11 +2294,12 @@ process_ACK: * Since we now have an rtt measurement, cancel the * timer backoff (cf., Phil Karn's retransmit alg.). * Recompute the initial retransmit timer. + * Also makes sure we have a valid time stamp in hand */ - if (to.to_flag & TOF_TS) + if (((to.to_flags & TOF_TS) != 0) && (to.to_tsecr != 0)) tcp_xmit_timer(tp, tcp_now - to.to_tsecr + 1); - else if (tp->t_rtt && SEQ_GT(th->th_ack, tp->t_rtseq)) - tcp_xmit_timer(tp,tp->t_rtt); + else if (tp->t_rtttime && SEQ_GT(th->th_ack, tp->t_rtseq)) + tcp_xmit_timer(tp, tp->t_rtttime); /* * If all outstanding data is acked, stop retransmit @@ -1984,13 +2327,13 @@ process_ACK: * Otherwise open linearly: maxseg per window * (maxseg^2 / cwnd per packet). */ - { - register u_int cw = tp->snd_cwnd; - register u_int incr = tp->t_maxseg; - - if (cw > tp->snd_ssthresh) - incr = incr * incr / cw; - tp->snd_cwnd = min(cw + incr, TCP_MAXWIN<snd_scale); + if ((!tcp_do_newreno && !tp->sack_enable) || + !IN_FASTRECOVERY(tp)) { + register u_int cw = tp->snd_cwnd; + register u_int incr = tp->t_maxseg; + if (cw > tp->snd_ssthresh) + incr = incr * incr / cw; + tp->snd_cwnd = min(cw+incr, TCP_MAXWIN<snd_scale); } if (acked > so->so_snd.sb_cc) { tp->snd_wnd -= so->so_snd.sb_cc; @@ -2001,8 +2344,22 @@ process_ACK: tp->snd_wnd -= acked; ourfinisacked = 0; } - need_sowwakeup++; + sowwakeup(so); + /* detect una wraparound */ + if ((tcp_do_newreno || tp->sack_enable) && + !IN_FASTRECOVERY(tp) && + SEQ_GT(tp->snd_una, tp->snd_recover) && + SEQ_LEQ(th->th_ack, tp->snd_recover)) + tp->snd_recover = th->th_ack - 1; + if ((tcp_do_newreno || tp->sack_enable) && + IN_FASTRECOVERY(tp) && + SEQ_GEQ(th->th_ack, tp->snd_recover)) + EXIT_FASTRECOVERY(tp); tp->snd_una = th->th_ack; + if (tp->sack_enable) { + 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; @@ -2027,8 +2384,8 @@ process_ACK: tp->t_timer[TCPT_2MSL] = tcp_maxidle; } add_to_time_wait(tp); - current_active_connections--; tp->t_state = TCPS_FIN_WAIT_2; + goto drop; } break; @@ -2044,13 +2401,12 @@ process_ACK: tcp_canceltimers(tp); /* Shorten TIME_WAIT [RFC-1644, p.28] */ if (tp->cc_recv != 0 && - tp->t_duration < TCPTV_MSL) + tp->t_starttime < tcp_msl) tp->t_timer[TCPT_2MSL] = tp->t_rxtcur * TCPTV_TWTRUNC; else - tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL; + tp->t_timer[TCPT_2MSL] = 2 * tcp_msl; add_to_time_wait(tp); - current_active_connections--; soisdisconnected(so); } break; @@ -2074,7 +2430,7 @@ process_ACK: * it and restart the finack timer. */ case TCPS_TIME_WAIT: - tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL; + tp->t_timer[TCPT_2MSL] = 2 * tcp_msl; add_to_time_wait(tp); goto dropafterack; } @@ -2090,7 +2446,7 @@ step6: (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 (tilen == 0 && + if (tlen == 0 && tp->snd_wl2 == th->th_ack && tiwin > tp->snd_wnd) tcpstat.tcps_rcvwinupd++; tp->snd_wnd = tiwin; @@ -2148,12 +2504,13 @@ step6: * but if two URG's are pending at once, some out-of-band * data may creep in... ick. */ - if (th->th_urp <= (u_long)tilen + if (th->th_urp <= (u_long)tlen #if SO_OOBINLINE && (so->so_options & SO_OOBINLINE) == 0 #endif ) - tcp_pulloutofband(so, th, m); + tcp_pulloutofband(so, th, m, + drop_hdrlen); /* hdr drop is delayed */ } else /* * If no out of band data is expected, @@ -2172,15 +2529,63 @@ dodata: /* XXX */ * case PRU_RCVD). If a FIN has already been received on this * connection then we just ignore the text. */ - if ((tilen || (thflags&TH_FIN)) && + if ((tlen || (thflags&TH_FIN)) && TCPS_HAVERCVDFIN(tp->t_state) == 0) { - TCP_REASS(tp, th, tilen, m, so, thflags, isipv6, need_sorwakeup); + tcp_seq save_start = th->th_seq; + tcp_seq save_end = th->th_seq + tlen; + 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_ACKNOW) == 0)) { + tp->t_flags |= TF_DELACK; + } + else { + tp->t_flags |= TF_ACKNOW; + } + tp->rcv_nxt += tlen; + thflags = th->th_flags & TH_FIN; + tcpstat.tcps_rcvpack++; + tcpstat.tcps_rcvbyte += tlen; + ND6_HINT(tp); + if (sbappend(&so->so_rcv, m)) + sorwakeup(so); + } else { + thflags = tcp_reass(tp, th, &tlen, m); + tp->t_flags |= TF_ACKNOW; + } + + if (tlen > 0 && tp->sack_enable) + tcp_update_sack_list(tp, save_start, save_end); if (tp->t_flags & TF_DELACK) { - KERNEL_DEBUG(DBG_LAYER_END, ((th->th_dport << 16) | th->th_sport), - (((th->th_src.s_addr & 0xffff) << 16) | (th->th_dst.s_addr & 0xffff)), - th->th_seq, th->th_ack, th->th_win); +#if INET6 + if (isipv6) { + KERNEL_DEBUG(DBG_LAYER_END, ((th->th_dport << 16) | th->th_sport), + (((ip6->ip6_src.s6_addr16[0]) << 16) | (ip6->ip6_dst.s6_addr16[0])), + th->th_seq, th->th_ack, th->th_win); + } + else +#endif + { + KERNEL_DEBUG(DBG_LAYER_END, ((th->th_dport << 16) | th->th_sport), + (((ip->ip_src.s_addr & 0xffff) << 16) | (ip->ip_dst.s_addr & 0xffff)), + th->th_seq, th->th_ack, th->th_win); + } + } /* * Note the amount of data that peer has sent into @@ -2204,18 +2609,18 @@ dodata: /* XXX */ /* * 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 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 (tcp_delack_enabled && (tp->t_flags & TF_NEEDSYN)) { - if (last_active_conn_count > DELACK_BITMASK_THRESH) - TCP_DELACK_BITSET(tp->t_inpcb->hash_element); - + if (DELAY_ACK(tp) && (tp->t_flags & TF_NEEDSYN)) { tp->t_flags |= TF_DELACK; } - else + else { tp->t_flags |= TF_ACKNOW; + } tp->rcv_nxt++; } switch (tp->t_state) { @@ -2225,6 +2630,7 @@ dodata: /* XXX */ * enter the CLOSE_WAIT state. */ case TCPS_SYN_RECEIVED: + tp->t_starttime = 0; case TCPS_ESTABLISHED: tp->t_state = TCPS_CLOSE_WAIT; break; @@ -2247,14 +2653,14 @@ dodata: /* XXX */ tcp_canceltimers(tp); /* Shorten TIME_WAIT [RFC-1644, p.28] */ if (tp->cc_recv != 0 && - tp->t_duration < TCPTV_MSL) { + tp->t_starttime < tcp_msl) { tp->t_timer[TCPT_2MSL] = tp->t_rxtcur * TCPTV_TWTRUNC; /* For transaction client, force ACK now. */ tp->t_flags |= TF_ACKNOW; } else - tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL; + tp->t_timer[TCPT_2MSL] = 2 * tcp_msl; add_to_time_wait(tp); soisdisconnected(so); @@ -2264,25 +2670,15 @@ dodata: /* XXX */ * In TIME_WAIT state restart the 2 MSL time_wait timer. */ case TCPS_TIME_WAIT: - tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL; + tp->t_timer[TCPT_2MSL] = 2 * tcp_msl; add_to_time_wait(tp); break; } } #if TCPDEBUG - if (so->so_options & SO_DEBUG) { -#if INET6 - if (isipv6) - tcp_saveip._tcp_si6.ip6_plen = tilen; - else - tcp_saveip._tcp_si4.ip_len = tilen; -#else /* INET6 */ - tcp_saveip.ip_len = tilen; -#endif /* INET6 */ - - tcp_trace(TA_INPUT, ostate, tp, (void *)&tcp_saveip, + if (so->so_options & SO_DEBUG) + tcp_trace(TA_INPUT, ostate, tp, (void *)tcp_saveipgen, &tcp_savetcp, 0); - } #endif /* @@ -2290,10 +2686,7 @@ dodata: /* XXX */ */ if (needoutput || (tp->t_flags & TF_ACKNOW)) (void) tcp_output(tp); - if (need_sorwakeup) - sorwakeup(so); - if (need_sowwakeup) - sowwakeup(so); + tcp_unlock(so, 1, 0); KERNEL_DEBUG(DBG_FNC_TCP_INPUT | DBG_FUNC_END,0,0,0,0,0); return; @@ -2315,32 +2708,23 @@ dropafterack: */ 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)) ) + SEQ_GT(th->th_ack, tp->snd_max)) ) { + rstreason = BANDLIM_RST_OPENPORT; goto dropwithreset; + } #if TCPDEBUG - if (so->so_options & SO_DEBUG) { -#if INET6 - if (isipv6) - tcp_saveip._tcp_si6.ip6_plen = tilen; - else - tcp_saveip._tcp_si4.ip_len = tilen; -#else /* INET6 */ - tcp_saveip.ip_len = tilen; -#endif /* INET6 */ - tcp_trace(TA_DROP, ostate, tp, (void *)&tcp_saveip, + if (so->so_options & SO_DEBUG) + tcp_trace(TA_DROP, ostate, tp, (void *)tcp_saveipgen, &tcp_savetcp, 0); - } #endif m_freem(m); tp->t_flags |= TF_ACKNOW; (void) tcp_output(tp); - if (need_sorwakeup) - sorwakeup(so); - if (need_sowwakeup) - sowwakeup(so); + tcp_unlock(so, 1, 0); KERNEL_DEBUG(DBG_FNC_TCP_INPUT | DBG_FUNC_END,0,0,0,0,0); return; - +dropwithresetnosock: + nosock = 1; dropwithreset: /* * Generate a RST, dropping incoming segment. @@ -2351,109 +2735,81 @@ dropwithreset: goto drop; #if INET6 if (isipv6) { - if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) - goto drop; /* anycast check is done at the top */ + if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || + IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) + goto drop; } else #endif /* INET6 */ - if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) + 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; + /* IPv6 anycast check is done at tcp6_input() */ + + /* + * Perform bandwidth limiting. + */ +#if ICMP_BANDLIM + if (badport_bandlim(rstreason) < 0) + goto drop; +#endif + #if TCPDEBUG - if (tp == 0 || (tp->t_inpcb->inp_socket->so_options & SO_DEBUG)) { - if (tp == 0) { -#if INET6 - if (isipv6) - tcp_saveip._tcp_si6 = *ip6; - else - tcp_saveip._tcp_si4 = *ip; -#else /* INET6 */ - tcp_saveip = *ip; -#endif /* INET6 */ - } -#if INET6 - if (isipv6) - tcp_saveip._tcp_si6.ip6_plen = tilen; - else - tcp_saveip._tcp_si4.ip_len = tilen; -#else /* INET6 */ - tcp_saveip.ip_len = tilen; -#endif /* INET6 */ - tcp_trace(TA_DROP, ostate, tp, (void *)&tcp_saveip, + if (tp == 0 || (tp->t_inpcb->inp_socket->so_options & SO_DEBUG)) + tcp_trace(TA_DROP, ostate, tp, (void *)tcp_saveipgen, &tcp_savetcp, 0); - } #endif if (thflags & TH_ACK) -#if INET6 - tcp_respond(tp, isipv6 ? (void *)ip6 : (void *)ip, th, m, - (tcp_seq)0, th->th_ack, TH_RST, isipv6); -#else /* INET6 */ - tcp_respond(tp, (void *)ip, th, m, - (tcp_seq)0, th->th_ack, TH_RST, isipv6); -#endif /* INET6 */ + /* mtod() below is safe as long as hdr dropping is delayed */ + tcp_respond(tp, mtod(m, void *), th, m, (tcp_seq)0, th->th_ack, + TH_RST); else { if (thflags & TH_SYN) - tilen++; -#if INET6 - tcp_respond(tp, isipv6 ? (void *)ip6 : (void *)ip, th, m, - th->th_seq+tilen, (tcp_seq)0, TH_RST|TH_ACK, - isipv6); -#else /* INET6 */ - tcp_respond(tp, (void *)ip, th, m, - th->th_seq+tilen, (tcp_seq)0, TH_RST|TH_ACK, - isipv6); -#endif /* INET6 */ + tlen++; + /* mtod() below is safe as long as hdr dropping is delayed */ + tcp_respond(tp, mtod(m, void *), th, m, th->th_seq+tlen, + (tcp_seq)0, TH_RST|TH_ACK); } /* destroy temporarily created socket */ - if (need_sorwakeup) - sorwakeup(so); - if (need_sowwakeup) - sowwakeup(so); - if (dropsocket) - (void) soabort(so); + if (dropsocket) { + (void) soabort(so); + tcp_unlock(so, 1, 0); + } + else + if ((inp != NULL) && (nosock == 0)) + tcp_unlock(so, 1, 0); KERNEL_DEBUG(DBG_FNC_TCP_INPUT | DBG_FUNC_END,0,0,0,0,0); return; - +dropnosock: + nosock = 1; drop: /* * Drop space held by incoming segment and return. */ #if TCPDEBUG - if (tp == 0 || (tp->t_inpcb->inp_socket->so_options & SO_DEBUG)) { - if (tp == 0) { -#if INET6 - if (isipv6) - tcp_saveip._tcp_si6 = *ip6; - else - tcp_saveip._tcp_si4 = *ip; -#else /* INET6 */ - tcp_saveip = *ip; -#endif /* INET6 */ - } -#if INET6 - if (isipv6) - tcp_saveip._tcp_si6.ip6_plen = tilen; - else - tcp_saveip._tcp_si4.ip_len = tilen; -#else /* INET6 */ - tcp_saveip.ip_len = tilen; -#endif /* INET6 */ - tcp_trace(TA_DROP, ostate, tp, (void *)&tcp_saveip, + if (tp == 0 || (tp->t_inpcb->inp_socket->so_options & SO_DEBUG)) + tcp_trace(TA_DROP, ostate, tp, (void *)tcp_saveipgen, &tcp_savetcp, 0); - } #endif m_freem(m); - if (need_sorwakeup) - sorwakeup(so); - if (need_sowwakeup) - sowwakeup(so); /* destroy temporarily created socket */ - if (dropsocket) - (void) soabort(so); + if (dropsocket) { + (void) soabort(so); + tcp_unlock(so, 1, 0); + } + else + if (nosock == 0) + tcp_unlock(so, 1, 0); KERNEL_DEBUG(DBG_FNC_TCP_INPUT | DBG_FUNC_END,0,0,0,0,0); return; } static void tcp_dooptions(tp, cp, cnt, th, to) +/* + * Parse TCP options and place in tcpopt. + */ struct tcpcb *tp; u_char *cp; int cnt; @@ -2470,8 +2826,10 @@ tcp_dooptions(tp, cp, cnt, th, to) if (opt == TCPOPT_NOP) optlen = 1; else { + if (cnt < 2) + break; optlen = cp[1]; - if (optlen <= 0) + if (optlen < 2 || optlen > cnt) break; } switch (opt) { @@ -2485,7 +2843,7 @@ tcp_dooptions(tp, cp, cnt, th, to) if (!(th->th_flags & TH_SYN)) continue; bcopy((char *) cp + 2, (char *) &mss, sizeof(mss)); - to->to_maxseg = ntohs(mss); + NTOHS(mss); break; case TCPOPT_WINDOW: @@ -2500,7 +2858,7 @@ tcp_dooptions(tp, cp, cnt, th, to) case TCPOPT_TIMESTAMP: if (optlen != TCPOLEN_TIMESTAMP) continue; - to->to_flag |= TOF_TS; + to->to_flags |= TOF_TS; bcopy((char *)cp + 2, (char *)&to->to_tsval, sizeof(to->to_tsval)); NTOHL(to->to_tsval); @@ -2518,47 +2876,25 @@ tcp_dooptions(tp, cp, cnt, th, to) tp->ts_recent_age = tcp_now; } break; - case TCPOPT_CC: - if (optlen != TCPOLEN_CC) + case TCPOPT_SACK_PERMITTED: + if (!tcp_do_sack || + optlen != TCPOLEN_SACK_PERMITTED) continue; - to->to_flag |= TOF_CC; - bcopy((char *)cp + 2, - (char *)&to->to_cc, sizeof(to->to_cc)); - NTOHL(to->to_cc); - /* - * A CC or CC.new option received in a SYN makes - * it ok to send CC in subsequent segments. - */ if (th->th_flags & TH_SYN) - tp->t_flags |= TF_RCVD_CC; - break; - case TCPOPT_CCNEW: - if (optlen != TCPOLEN_CC) - continue; - if (!(th->th_flags & TH_SYN)) - continue; - to->to_flag |= TOF_CCNEW; - bcopy((char *)cp + 2, - (char *)&to->to_cc, sizeof(to->to_cc)); - NTOHL(to->to_cc); - /* - * A CC or CC.new option received in a SYN makes - * it ok to send CC in subsequent segments. - */ - tp->t_flags |= TF_RCVD_CC; + to->to_flags |= TOF_SACK; break; - case TCPOPT_CCECHO: - if (optlen != TCPOLEN_CC) + case TCPOPT_SACK: + if (optlen <= 2 || (optlen - 2) % TCPOLEN_SACK != 0) continue; - if (!(th->th_flags & TH_SYN)) - continue; - to->to_flag |= TOF_CCECHO; - bcopy((char *)cp + 2, - (char *)&to->to_ccecho, sizeof(to->to_ccecho)); - NTOHL(to->to_ccecho); + to->to_nsacks = (optlen - 2) / TCPOLEN_SACK; + to->to_sacks = cp + 2; + tcpstat.tcps_sack_rcv_blocks++; + break; } } + if (th->th_flags & TH_SYN) + tcp_mss(tp, mss); /* sets t_maxseg */ } /* @@ -2568,12 +2904,13 @@ tcp_dooptions(tp, cp, cnt, th, to) * sequencing purposes. */ static void -tcp_pulloutofband(so, th, m) +tcp_pulloutofband(so, th, m, off) struct socket *so; struct tcphdr *th; register struct mbuf *m; + int off; /* delayed to be droped hdrlen */ { - int cnt = th->th_urp - 1; + int cnt = off + th->th_urp - 1; while (cnt >= 0) { if (m->m_len > cnt) { @@ -2584,6 +2921,8 @@ tcp_pulloutofband(so, th, m) 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; @@ -2601,7 +2940,7 @@ tcp_pulloutofband(so, th, m) static void tcp_xmit_timer(tp, rtt) register struct tcpcb *tp; - short rtt; + int rtt; { register int delta; @@ -2645,7 +2984,7 @@ tcp_xmit_timer(tp, rtt) tp->t_srtt = rtt << TCP_RTT_SHIFT; tp->t_rttvar = rtt << (TCP_RTTVAR_SHIFT - 1); } - tp->t_rtt = 0; + tp->t_rtttime = 0; tp->t_rxtshift = 0; /* @@ -2701,12 +3040,9 @@ tcp_xmit_timer(tp, rtt) * MSS of our peer. */ void -tcp_mss(tp, offer, isipv6) +tcp_mss(tp, offer) struct tcpcb *tp; int offer; -#if INET6 - int isipv6; -#endif { register struct rtentry *rt; struct ifnet *ifp; @@ -2717,13 +3053,18 @@ tcp_mss(tp, offer, isipv6) struct rmxp_tao *taop; int origoffer = offer; #if INET6 - int lgminh = isipv6 ? sizeof (struct tcpip6hdr) : - sizeof (struct tcpiphdr); -#else /* INET6 */ -#define lgminh (sizeof (struct tcpiphdr)) -#endif /* INET6 */ + int isipv6; + int min_protoh; +#endif inp = tp->t_inpcb; +#if INET6 + isipv6 = ((inp->inp_vflag & INP_IPV6) != 0) ? 1 : 0; + min_protoh = isipv6 ? sizeof (struct ip6_hdr) + sizeof (struct tcphdr) + : sizeof (struct tcpiphdr); +#else +#define min_protoh (sizeof (struct tcpiphdr)) +#endif #if INET6 if (isipv6) rt = tcp_rtlookup6(inp); @@ -2739,6 +3080,16 @@ tcp_mss(tp, offer, isipv6) return; } ifp = rt->rt_ifp; + /* + * Slower link window correction: + * If a value is specificied for slowlink_wsize use it for PPP links + * believed to be on a serial modem (speed <128Kbps). Excludes 9600bps as + * it is the default value adversized by pseudo-devices over ppp. + */ + if (ifp->if_type == IFT_PPP && slowlink_wsize > 0 && + ifp->if_baudrate > 9600 && ifp->if_baudrate <= 128000) { + tp->t_flags |= TF_SLOWLINK; + } so = inp->inp_socket; taop = rmx_taop(rt->rt_rmx); @@ -2758,7 +3109,12 @@ tcp_mss(tp, offer, isipv6) isipv6 ? tcp_v6mssdflt : #endif /* INET6 */ tcp_mssdflt; - else + else { + /* + * Prevent DoS attack with too small MSS. Round up + * to at least minmss. + */ + offer = max(offer, tcp_minmss); /* * Sanity check: make sure that maxopd will be large * enough to allow some data on segments even is the @@ -2766,6 +3122,7 @@ tcp_mss(tp, offer, isipv6) * funny things may happen in tcp_output. */ offer = max(offer, 64); + } taop->tao_mssopt = offer; /* @@ -2792,23 +3149,26 @@ tcp_mss(tp, offer, isipv6) 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); + ((tp->t_srtt >> 2) + tp->t_rttvar) >> 1, + tp->t_rttmin, TCPTV_REXMTMAX); } /* * if there's an mtu associated with the route, use it * else, use the link mtu. */ if (rt->rt_rmx.rmx_mtu) - mss = rt->rt_rmx.rmx_mtu - lgminh; + mss = rt->rt_rmx.rmx_mtu - min_protoh; else + { mss = #if INET6 - isipv6 ? nd_ifinfo[rt->rt_ifp->if_index].linkmtu : + (isipv6 ? nd_ifinfo[rt->rt_ifp->if_index].linkmtu : #endif - ifp->if_mtu - lgminh; - - if (rt->rt_rmx.rmx_mtu == 0) { + ifp->if_mtu +#if INET6 + ) +#endif + - min_protoh; #if INET6 if (isipv6) { if (!in6_localaddr(&inp->in6p_faddr)) @@ -2829,34 +3189,25 @@ tcp_mss(tp, offer, isipv6) tp->t_maxopd = mss; /* - * In case of T/TCP, origoffer==-1 indicates, that no segments - * were received yet. In this case we just guess, otherwise - * we do the same as before T/TCP. + * 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 && + 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 ((tp->t_flags & (TF_REQ_CC|TF_NOOPT)) == TF_REQ_CC && - (origoffer == -1 || - (tp->t_flags & TF_RCVD_CC) == TF_RCVD_CC)) - mss -= TCPOLEN_CC_APPA; + tp->t_maxseg = mss; -#if (MCLBYTES & (MCLBYTES - 1)) == 0 - if (mss > MCLBYTES) - mss &= ~(MCLBYTES-1); -#else - if (mss > MCLBYTES) - mss = mss / MCLBYTES * MCLBYTES; -#endif /* - * If there's a pipesize, change the socket buffer - * to that size. Make the socket buffers an integral + * If there's a pipesize (ie loopback), change the socket + * buffer to that size only if it's bigger than the current + * sockbuf size. Make the socket buffers an integral * number of mss units; if the mss is larger than * the socket buffer, decrease the mss. */ #if RTV_SPIPE - if ((bufsize = rt->rt_rmx.rmx_sendpipe) == 0) + bufsize = rt->rt_rmx.rmx_sendpipe; + if (bufsize < so->so_snd.sb_hiwat) #endif bufsize = so->so_snd.sb_hiwat; if (bufsize < mss) @@ -2870,7 +3221,8 @@ tcp_mss(tp, offer, isipv6) tp->t_maxseg = mss; #if RTV_RPIPE - if ((bufsize = rt->rt_rmx.rmx_recvpipe) == 0) + bufsize = rt->rt_rmx.rmx_recvpipe; + if (bufsize < so->so_rcv.sb_hiwat) #endif bufsize = so->so_rcv.sb_hiwat; if (bufsize > mss) { @@ -2879,17 +3231,24 @@ tcp_mss(tp, offer, isipv6) bufsize = sb_max; (void)sbreserve(&so->so_rcv, bufsize); } + /* - * Don't force slow-start on local network. + * Set the slow-start flight size depending on whether this + * is a local network or not. */ + if ( #if INET6 - if (isipv6) { - if (!in6_localaddr(&inp->in6p_faddr)) - tp->snd_cwnd = mss; - } else -#endif /* INET6 */ - if (!in_localaddr(inp->inp_faddr)) - tp->snd_cwnd = mss; + (isipv6 && in6_localaddr(&inp->in6p_faddr)) || + (!isipv6 && +#endif + in_localaddr(inp->inp_faddr) +#if INET6 + ) +#endif + ) + tp->snd_cwnd = mss * ss_fltsz_local; + else + tp->snd_cwnd = mss * ss_fltsz; if (rt->rt_rmx.rmx_ssthresh) { /* @@ -2907,21 +3266,23 @@ tcp_mss(tp, offer, isipv6) * Determine the MSS option to send on an outgoing SYN. */ int -tcp_mssopt(tp, isipv6) +tcp_mssopt(tp) struct tcpcb *tp; -#if INET6 - int isipv6; -#endif { struct rtentry *rt; int mss; #if INET6 - int lgminh = isipv6 ? sizeof (struct tcpip6hdr) : - sizeof (struct tcpiphdr); -#else /* INET6 */ -#define lgminh (sizeof (struct tcpiphdr)) -#endif /* INET6 */ + int isipv6; + int min_protoh; +#endif +#if INET6 + isipv6 = ((tp->t_inpcb->inp_vflag & INP_IPV6) != 0) ? 1 : 0; + min_protoh = isipv6 ? sizeof (struct ip6_hdr) + sizeof (struct tcphdr) + : sizeof (struct tcpiphdr); +#else +#define min_protoh (sizeof (struct tcpiphdr)) +#endif #if INET6 if (isipv6) rt = tcp_rtlookup6(tp->t_inpcb); @@ -2934,8 +3295,171 @@ tcp_mssopt(tp, isipv6) isipv6 ? tcp_v6mssdflt : #endif /* INET6 */ tcp_mssdflt; + /* + * Slower link window correction: + * If a value is specificied for slowlink_wsize use it for PPP links + * believed to be on a serial modem (speed <128Kbps). Excludes 9600bps as + * it is the default value adversized by pseudo-devices over ppp. + */ + if (rt->rt_ifp->if_type == IFT_PPP && slowlink_wsize > 0 && + rt->rt_ifp->if_baudrate > 9600 && rt->rt_ifp->if_baudrate <= 128000) { + tp->t_flags |= TF_SLOWLINK; + } + + if (rt->rt_rmx.rmx_mtu) + mss = rt->rt_rmx.rmx_mtu; + else { + mss = +#if INET6 + (isipv6 ? nd_ifinfo[rt->rt_ifp->if_index].linkmtu : +#endif + rt->rt_ifp->if_mtu +#if INET6 + ); +#endif + } + return (mss - min_protoh); +} + +/* + * 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(tp, th) + struct tcpcb *tp; + struct tcphdr *th; +{ + tcp_seq onxt = tp->snd_nxt; + u_long ocwnd = tp->snd_cwnd; + tp->t_timer[TCPT_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 + (th->th_ack - tp->snd_una); + 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. + */ + tp->snd_cwnd -= (th->th_ack - tp->snd_una - tp->t_maxseg); +} + +/* + * Drop a random TCP connection that hasn't been serviced yet and + * is eligible for discard. There is a one in qlen chance that + * we will return a null, saying that there are no dropable + * requests. In this case, the protocol specific code should drop + * the new request. This insures fairness. + * + * The listening TCP socket "head" must be locked + */ +static int +tcpdropdropablreq(struct socket *head) +{ + struct socket *so; + unsigned int i, j, qlen; + static int rnd; + static struct timeval old_runtime; + static unsigned int cur_cnt, old_cnt; + struct timeval tv; + struct inpcb *inp = NULL; + struct tcpcb *tp; + + microtime(&tv); + if ((i = (tv.tv_sec - old_runtime.tv_sec)) != 0) { + old_runtime = tv; + old_cnt = cur_cnt / i; + cur_cnt = 0; + } + + so = TAILQ_FIRST(&head->so_incomp); + if (!so) + return 0; + + qlen = head->so_incqlen; + if (++cur_cnt > qlen || old_cnt > qlen) { + rnd = (314159 * rnd + 66329) & 0xffff; + j = ((qlen + 1) * rnd) >> 16; + + while (j-- && so) + so = TAILQ_NEXT(so, so_list); + } + /* Find a connection that is not already closing */ + while (so) { + inp = (struct inpcb *)so->so_pcb; + + if (in_pcb_checkstate(inp, WNT_ACQUIRE, 0) != WNT_STOPUSING) + break; + + so = TAILQ_NEXT(so, so_list); + } + if (!so) + return 0; + + head->so_incqlen--; + head->so_qlen--; + TAILQ_REMOVE(&head->so_incomp, so, so_list); + tcp_unlock(head, 0, 0); + + /* Let's remove this connection from the incomplete list */ + tcp_lock(so, 1, 0); + + if (in_pcb_checkstate(inp, WNT_RELEASE, 1) == WNT_STOPUSING) { + tcp_unlock(so, 1, 0); + return 0; + } + + so->so_head = NULL; + so->so_usecount--; /* No more held by so_head */ + + /* + * We do not want to lose track of the PCB right away in case we receive + * more segments from the peer + */ + tp = sototcpcb(so); + tp->t_flags |= TF_LQ_OVERFLOW; + tp->t_state = TCPS_CLOSED; + (void) tcp_output(tp); + tcpstat.tcps_drops++; + soisdisconnected(so); + tcp_canceltimers(tp); + add_to_time_wait(tp); + + tcp_unlock(so, 1, 0); + tcp_lock(head, 0, 0); + + return 1; + +} + +static int +tcp_getstat SYSCTL_HANDLER_ARGS +{ + + int error; + + if (req->oldptr == 0) { + req->oldlen= (size_t)sizeof(struct tcpstat); + } - mss = rt->rt_ifp->if_mtu - lgminh; + error = SYSCTL_OUT(req, &tcpstat, (size_t) req->oldlen); + + return (error); - return mss; } + +SYSCTL_PROC(_net_inet_tcp, TCPCTL_STATS, stats, CTLFLAG_RD, 0, 0, + tcp_getstat, "S,tcpstat", "TCP statistics (struct tcpstat, netinet/tcp_var.h)"); + +