X-Git-Url: https://git.saurik.com/apple/xnu.git/blobdiff_plain/43866e378188c25dd1e2208016ab3cbeb086ae6c..743b15655a24ee3fe9f458f383003e011db0558f:/bsd/netinet/tcp_input.c diff --git a/bsd/netinet/tcp_input.c b/bsd/netinet/tcp_input.c index 5ecafc97c..624b3451f 100644 --- a/bsd/netinet/tcp_input.c +++ b/bsd/netinet/tcp_input.c @@ -3,22 +3,19 @@ * * @APPLE_LICENSE_HEADER_START@ * - * Copyright (c) 1999-2003 Apple Computer, Inc. All Rights Reserved. + * 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. 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 + * This 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, QUIET ENJOYMENT OR NON-INFRINGEMENT. - * Please see the License for the specific language governing rights and - * limitations under the License. + * FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT. Please see the + * License for the specific language governing rights and limitations + * under the License. * * @APPLE_LICENSE_HEADER_END@ */ @@ -132,6 +129,7 @@ extern int apple_hwcksum_rx; #if IPSEC extern int ipsec_bypass; +extern lck_mtx_t *sadb_mutex; #endif struct tcpstat tcpstat; @@ -146,7 +144,7 @@ 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, "Delay ACK to try and piggyback it onto a data packet"); @@ -157,11 +155,30 @@ SYSCTL_INT(_net_inet_tcp, OID_AUTO, tcp_lq_overflow, CTLFLAG_RW, "Listen Queue Overflow"); #if TCP_DROP_SYNFIN -static int drop_synfin = 0; +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 +SYSCTL_NODE(_net_inet_tcp, OID_AUTO, reass, CTLFLAG_RW, 0, + "TCP Segment Reassembly Queue"); + +__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"); + +__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"); + + __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"); @@ -172,14 +189,14 @@ struct inpcbhead tcb; #define tcb6 tcb /* for KAME src sync over BSD*'s */ struct inpcbinfo tcbinfo; -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 *, int)); -static int tcp_reass __P((struct tcpcb *, struct tcphdr *, int *, - struct mbuf *)); -static void tcp_xmit_timer __P((struct tcpcb *, int)); -static int tcp_newreno __P((struct tcpcb *, struct tcphdr *)); +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); +static int tcp_newreno __P((struct tcpcb *, struct tcphdr *)); /* Neighbor Discovery, Neighbor Unreachability Detection Upper layer hint. */ #if INET6 @@ -196,16 +213,43 @@ do { \ 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 ; \ +} + /* - * Indicate whether this ack should be delayed. We can delay the ack if - * - delayed acks are enabled and - * - there is no delayed ack timer in progress and + * 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. + * 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. + * */ #define DELAY_ACK(tp) \ - (tcp_delack_enabled && !callout_pending(tp->tt_delack) && \ - (tp->t_flags & TF_RXWIN0SENT) == 0) + (((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 int @@ -221,6 +265,7 @@ tcp_reass(tp, th, tlenp, m) struct tseg_qent *te; struct socket *so = tp->t_inpcb->inp_socket; int flags; + int dowakeup = 0; /* * Call with th==0 after become established to @@ -229,6 +274,21 @@ tcp_reass(tp, th, tlenp, m) if (th == 0) goto present; + /* + * 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. + */ + 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); @@ -237,6 +297,7 @@ tcp_reass(tp, th, tlenp, m) m_freem(m); return (0); } + tcp_reass_qsize++; /* * Find a segment which begins after this one does. @@ -262,6 +323,7 @@ tcp_reass(tp, th, tlenp, m) tcpstat.tcps_rcvdupbyte += *tlenp; m_freem(m); FREE(te, M_TSEGQ); + tcp_reass_qsize--; /* * Try to present any queued data * at the left window edge to the user. @@ -297,6 +359,7 @@ tcp_reass(tp, th, tlenp, m) LIST_REMOVE(q, tqe_q); m_freem(q->tqe_m); FREE(q, M_TSEGQ); + tcp_reass_qsize--; q = nq; } @@ -328,9 +391,12 @@ present: LIST_REMOVE(q, tqe_q); if (so->so_state & SS_CANTRCVMORE) m_freem(q->tqe_m); - else - sbappend(&so->so_rcv, 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->tqe_th->th_seq == tp->rcv_nxt); ND6_HINT(tp); @@ -352,8 +418,9 @@ present: (((tp->t_inpcb->inp_laddr.s_addr & 0xffff) << 16) | (tp->t_inpcb->inp_faddr.s_addr & 0xffff)), 0,0,0); - } - sorwakeup(so); + } + if (dowakeup) + sorwakeup(so); /* done with socket lock held */ return (flags); } @@ -365,14 +432,14 @@ present: */ #if INET6 int -tcp6_input(mp, offp, proto) +tcp6_input(mp, offp) struct mbuf **mp; - int *offp, proto; + int *offp; { register struct mbuf *m = *mp; struct in6_ifaddr *ia6; - IP6_EXTHDR_CHECK(m, *offp, sizeof(struct tcphdr), IPPROTO_DONE); + IP6_EXTHDR_CHECK(m, *offp, sizeof(struct tcphdr), return IPPROTO_DONE); /* * draft-itojun-ipv6-tcp-to-anycast @@ -416,13 +483,26 @@ tcp_input(m, off0) #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 */ + 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 isipv6; @@ -448,7 +528,7 @@ tcp_input(m, off0) 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); @@ -465,7 +545,7 @@ tcp_input(m, off0) */ if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) { /* XXX stat */ - goto drop; + goto dropnosock; } } else #endif /* INET6 */ @@ -544,7 +624,7 @@ tcp_input(m, off0) } if (th->th_sum) { tcpstat.tcps_rcvbadsum++; - goto drop; + goto dropnosock; } #if INET6 /* Re-initialization for later version check */ @@ -559,13 +639,13 @@ tcp_input(m, off0) off = th->th_off << 2; if (off < sizeof (struct tcphdr) || off > tlen) { tcpstat.tcps_rcvbadoff++; - goto drop; + goto dropnosock; } tlen -= off; /* tlen is used instead of ti->ti_len */ if (off > sizeof (struct tcphdr)) { #if INET6 if (isipv6) { - IP6_EXTHDR_CHECK(m, off0, off, ); + IP6_EXTHDR_CHECK(m, off0, off, return); ip6 = mtod(m, struct ip6_hdr *); th = (struct tcphdr *)((caddr_t)ip6 + off0); } else @@ -612,7 +692,7 @@ tcp_input(m, off0) * This is incompatible with RFC1644 extensions (T/TCP). */ if (drop_synfin && (thflags & (TH_SYN|TH_FIN)) == (TH_SYN|TH_FIN)) - goto drop; + goto dropnosock; #endif /* @@ -638,7 +718,7 @@ tcp_input(m, off0) */ findpcb: #if IPFIREWALL_FORWARD - if (ip_fw_fwd_addr != NULL + if (next_hop != NULL #if INET6 && isipv6 == NULL /* IPv6 support is not yet */ #endif /* INET6 */ @@ -653,19 +733,18 @@ 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 */ { @@ -681,17 +760,23 @@ findpcb: } #if IPSEC + if (ipsec_bypass == 0) { + lck_mtx_lock(sadb_mutex); #if INET6 - if (isipv6) { - if (ipsec_bypass == 0 && 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 (ipsec_bypass == 0 && 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); } #endif /*IPSEC*/ @@ -704,55 +789,88 @@ findpcb: if (inp == NULL) { if (log_in_vain) { #if INET6 - char dbuf[INET6_ADDRSTRLEN], sbuf[INET6_ADDRSTRLEN]; + char dbuf[MAX_IPv6_STR_LEN], sbuf[MAX_IPv6_STR_LEN]; #else /* INET6 */ - char dbuf[4*sizeof "123"], sbuf[4*sizeof "123"]; + char dbuf[MAX_IPv4_STR_LEN], sbuf[MAX_IPv4_STR_LEN]; #endif /* INET6 */ #if INET6 if (isipv6) { - strcpy(dbuf, ip6_sprintf(&ip6->ip6_dst)); - strcpy(sbuf, ip6_sprintf(&ip6->ip6_src)); + inet_ntop(AF_INET6, &ip6->ip6_dst, dbuf, sizeof(dbuf)); + inet_ntop(AF_INET6, &ip6->ip6_src, sbuf, sizeof(sbuf)); } else #endif - { - strcpy(dbuf, inet_ntoa(ip->ip_dst)); - strcpy(sbuf, inet_ntoa(ip->ip_src)); - } + { + 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)); + "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); + "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; } } if (blackhole) { - switch (blackhole) { - case 1: - if (thflags & TH_SYN) - goto drop; - break; - case 2: - goto drop; - default: - goto drop; - } + 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 dropwithreset; + 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 dropnosock; + } + 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) { rstreason = BANDLIM_RST_CLOSEDPORT; @@ -776,7 +894,6 @@ 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) { @@ -792,14 +909,14 @@ findpcb: } #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 /* @@ -864,25 +981,51 @@ findpcb: } } #endif - - so2 = sonewconn(so, 0); + 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) { - if (tcp_lq_overflow) - sototcpcb(so2)->t_flags |= - TF_LQ_OVERFLOW; - 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 .... * @@ -908,7 +1051,7 @@ findpcb: } #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. @@ -920,6 +1063,8 @@ 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 @@ -936,6 +1081,8 @@ findpcb: * 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; @@ -968,22 +1115,26 @@ findpcb: } else #endif /* INET6 */ inp->inp_options = ip_srcroute(); + tcp_lock(oso, 0, 0); #if IPSEC /* copy old policy into new socket's */ 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|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 < @@ -994,13 +1145,80 @@ findpcb: } } +#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; + /* + * Setting either tcp_minmssoverload or tcp_minmss to "0" disables + * the check. + */ + 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 + 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; + } + } + /* * Segment received on connection. * Reset idle time and keep-alive timer. */ 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, @@ -1102,9 +1320,10 @@ 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; } @@ -1125,7 +1344,8 @@ findpcb: * Add data to socket buffer. */ m_adj(m, drop_hdrlen); /* delayed header drop */ - sbappend(&so->so_rcv, m); + if (sbappend(&so->so_rcv, m)) + sorwakeup(so); #if INET6 if (isipv6) { KERNEL_DEBUG(DBG_LAYER_END, ((th->th_dport << 16) | th->th_sport), @@ -1139,14 +1359,13 @@ findpcb: (((ip->ip_src.s_addr & 0xffff) << 16) | (ip->ip_dst.s_addr & 0xffff)), th->th_seq, th->th_ack, th->th_win); } - if (tcp_delack_enabled) { - TCP_DELACK_BITSET(tp->t_inpcb->hash_element); + 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; } @@ -1158,6 +1377,9 @@ 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); @@ -1192,7 +1414,10 @@ findpcb: 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) { rstreason = BANDLIM_RST_OPENPORT; @@ -1257,6 +1482,9 @@ findpcb: } else #endif { +#if 1 + lck_mtx_assert(((struct inpcb *)so->so_pcb)->inpcb_mtx, LCK_MTX_ASSERT_OWNED); +#endif MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME, M_NOWAIT); if (sin == NULL) @@ -1326,7 +1554,7 @@ findpcb: * segment. Otherwise must send ACK now in case * the other side is slow starting. */ - if (tcp_delack_enabled && ((thflags & TH_FIN) || + if (DELAY_ACK(tp) && ((thflags & TH_FIN) || (tlen != 0 && #if INET6 (isipv6 && in6_localaddr(&inp->in6p_faddr)) @@ -1338,11 +1566,11 @@ findpcb: ) #endif /* INET6 */ ))) { - TCP_DELACK_BITSET(tp->t_inpcb->hash_element); tp->t_flags |= (TF_DELACK | TF_NEEDSYN); } - else + else { tp->t_flags |= (TF_ACKNOW | TF_NEEDSYN); + } /* * Limit the `virtual advertised window' to TCP_MAXWIN @@ -1481,12 +1709,12 @@ findpcb: * If there's data, delay ACK; if there's also a FIN * ACKNOW will be turned on later. */ - if (tcp_delack_enabled && tlen != 0) { - 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: @@ -1499,7 +1727,7 @@ findpcb: thflags &= ~TH_SYN; } else { tp->t_state = TCPS_ESTABLISHED; - tp->t_timer[TCPT_KEEP] = tcp_keepidle; + tp->t_timer[TCPT_KEEP] = TCP_KEEPIDLE(tp); } } else { /* @@ -1527,7 +1755,7 @@ findpcb: 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); } tp->t_flags |= TF_NEEDSYN; } else @@ -1592,12 +1820,23 @@ trimthenstep6: } if (CC_GT(to.to_cc, tp->cc_recv)) { tp = tcp_close(tp); + tcp_unlock(so, 1, 50); 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; } /* @@ -1692,6 +1931,9 @@ 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. @@ -1821,6 +2063,7 @@ trimthenstep6: SEQ_GT(th->th_seq, tp->rcv_nxt)) { iss = tcp_new_isn(tp); tp = tcp_close(tp); + tcp_unlock(so, 1, 0); goto findpcb; } /* @@ -1918,7 +2161,7 @@ trimthenstep6: 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() @@ -2151,10 +2394,17 @@ process_ACK: tp->snd_wnd -= acked; ourfinisacked = 0; } + sowwakeup(so); + /* detect una wraparound */ + if (SEQ_GEQ(tp->snd_una, tp->snd_recover) && + SEQ_LT(th->th_ack, tp->snd_recover)) + tp->snd_recover = th->th_ack; + if (SEQ_GT(tp->snd_una, tp->snd_high) && + SEQ_LEQ(th->th_ack, tp->snd_high)) + tp->snd_high = th->th_ack - 1; tp->snd_una = th->th_ack; if (SEQ_LT(tp->snd_nxt, tp->snd_una)) tp->snd_nxt = tp->snd_una; - sowwakeup(so); switch (tp->t_state) { @@ -2178,6 +2428,7 @@ process_ACK: } add_to_time_wait(tp); tp->t_state = TCPS_FIN_WAIT_2; + goto drop; } break; @@ -2337,25 +2588,19 @@ dodata: /* XXX */ if (th->th_seq == tp->rcv_nxt && LIST_EMPTY(&tp->t_segq) && TCPS_HAVEESTABLISHED(tp->t_state)) { -#ifdef __APPLE__ - if (tcp_delack_enabled) { - TCP_DELACK_BITSET(tp->t_inpcb->hash_element); + if (DELAY_ACK(tp) && ((tp->t_flags & TF_ACKNOW) == 0)) { tp->t_flags |= TF_DELACK; } -#else - if (DELAY_ACK(tp)) - callout_reset(tp->tt_delack, tcp_delacktime, - tcp_timer_delack, tp); -#endif - else + 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); - sbappend(&so->so_rcv, m); - sorwakeup(so); + if (sbappend(&so->so_rcv, m)) + sorwakeup(so); } else { thflags = tcp_reass(tp, th, &tlen, m); tp->t_flags |= TF_ACKNOW; @@ -2404,12 +2649,12 @@ dodata: /* XXX */ * 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)) { - 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) { @@ -2475,6 +2720,7 @@ dodata: /* XXX */ */ if (needoutput || (tp->t_flags & TF_ACKNOW)) (void) tcp_output(tp); + tcp_unlock(so, 1, 0); KERNEL_DEBUG(DBG_FNC_TCP_INPUT | DBG_FUNC_END,0,0,0,0,0); return; @@ -2508,9 +2754,11 @@ dropafterack: m_freem(m); tp->t_flags |= TF_ACKNOW; (void) tcp_output(tp); + 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. @@ -2558,11 +2806,17 @@ dropwithreset: (tcp_seq)0, TH_RST|TH_ACK); } /* destroy temporarily created socket */ - 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. @@ -2574,8 +2828,13 @@ drop: #endif m_freem(m); /* 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; } @@ -2905,7 +3164,12 @@ tcp_mss(tp, offer) 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 @@ -2913,6 +3177,7 @@ tcp_mss(tp, offer) * funny things may happen in tcp_output. */ offer = max(offer, 64); + } taop->tao_mssopt = offer; /* @@ -2992,21 +3257,16 @@ tcp_mss(tp, offer) (tp->t_flags & TF_RCVD_CC) == TF_RCVD_CC)) mss -= TCPOLEN_CC_APPA; -#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) @@ -3020,7 +3280,8 @@ tcp_mss(tp, offer) 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) { @@ -3134,6 +3395,7 @@ tcp_newreno(tp, th) * 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)) @@ -3147,3 +3409,93 @@ tcp_newreno(tp, th) } return (0); } + +/* + * 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; + +} + +