[apple/xnu.git] / bsd / netinet / tcp_subr.c

/*
 * Copyright (c) 2000-2007 Apple Inc. All rights reserved.
 *
 * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
 * 
 * This file contains Original Code and/or Modifications of Original Code
 * as defined in and that are subject to the Apple Public Source License
 * Version 2.0 (the 'License'). You may not use this file except in
 * 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.
 * 
 * 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, QUIET ENJOYMENT OR NON-INFRINGEMENT.
 * Please see the License for the specific language governing rights and
 * limitations under the License.
 * 
 * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
 */
/*
 * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995
 *      The Regents of the University of California.  All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 * 3. All advertising materials mentioning features or use of this software
 *    must display the following acknowledgement:
 *      This product includes software developed by the University of
 *      California, Berkeley and its contributors.
 * 4. Neither the name of the University nor the names of its contributors
 *    may be used to endorse or promote products derived from this software
 *    without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 * SUCH DAMAGE.
 *
 *      @(#)tcp_subr.c  8.2 (Berkeley) 5/24/95
 * $FreeBSD: src/sys/netinet/tcp_subr.c,v 1.73.2.22 2001/08/22 00:59:12 silby Exp $
 */
/*
 * NOTICE: This file was modified by SPARTA, Inc. in 2005 to introduce
 * support for mandatory and extensible security protections.  This notice
 * is included in support of clause 2.2 (b) of the Apple Public License,
 * Version 2.0.
 */

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/callout.h>
#include <sys/kernel.h>
#include <sys/sysctl.h>
#include <sys/malloc.h>
#include <sys/mbuf.h>
#include <sys/domain.h>
#include <sys/proc.h>
#include <sys/kauth.h>
#include <sys/socket.h>
#include <sys/socketvar.h>
#include <sys/protosw.h>
#include <sys/random.h>
#include <sys/syslog.h>
#include <kern/locks.h>
#include <kern/zalloc.h>

#include <net/route.h>
#include <net/if.h>

#define _IP_VHL
#include <netinet/in.h>
#include <netinet/in_systm.h>
#include <netinet/ip.h>
#if INET6
#include <netinet/ip6.h>
#endif
#include <netinet/in_pcb.h>
#if INET6
#include <netinet6/in6_pcb.h>
#endif
#include <netinet/in_var.h>
#include <netinet/ip_var.h>
#if INET6
#include <netinet6/ip6_var.h>
#endif
#include <netinet/tcp.h>
#include <netinet/tcp_fsm.h>
#include <netinet/tcp_seq.h>
#include <netinet/tcp_timer.h>
#include <netinet/tcp_var.h>
#if INET6
#include <netinet6/tcp6_var.h>
#endif
#include <netinet/tcpip.h>
#if TCPDEBUG
#include <netinet/tcp_debug.h>
#endif
#include <netinet6/ip6protosw.h>

#if IPSEC
#include <netinet6/ipsec.h>
#if INET6
#include <netinet6/ipsec6.h>
#endif
#endif /*IPSEC*/

#if CONFIG_MACF_NET
#include <security/mac_framework.h>
#endif /* MAC_NET */

#include <libkern/crypto/md5.h>
#include <sys/kdebug.h>

#define DBG_FNC_TCP_CLOSE       NETDBG_CODE(DBG_NETTCP, ((5 << 8) | 2))

extern int tcp_lq_overflow;

/* temporary: for testing */
#if IPSEC
extern int ipsec_bypass;
#endif

int     tcp_mssdflt = TCP_MSS;
SYSCTL_INT(_net_inet_tcp, TCPCTL_MSSDFLT, mssdflt, CTLFLAG_RW, 
    &tcp_mssdflt , 0, "Default TCP Maximum Segment Size");

#if INET6
int     tcp_v6mssdflt = TCP6_MSS;
SYSCTL_INT(_net_inet_tcp, TCPCTL_V6MSSDFLT, v6mssdflt,
        CTLFLAG_RW, &tcp_v6mssdflt , 0,
        "Default TCP Maximum Segment Size for IPv6");
#endif

/*
 * Minimum MSS we accept and use. This prevents DoS attacks where
 * we are forced to a ridiculous low MSS like 20 and send hundreds
 * of packets instead of one. The effect scales with the available
 * bandwidth and quickly saturates the CPU and network interface
 * with packet generation and sending. Set to zero to disable MINMSS
 * checking. This setting prevents us from sending too small packets.
 */
int     tcp_minmss = TCP_MINMSS;
SYSCTL_INT(_net_inet_tcp, OID_AUTO, minmss, CTLFLAG_RW,
    &tcp_minmss , 0, "Minmum TCP Maximum Segment Size");

/*
 * Number of TCP segments per second we accept from remote host
 * before we start to calculate average segment size. If average
 * segment size drops below the minimum TCP MSS we assume a DoS
 * attack and reset+drop the connection. Care has to be taken not to
 * set this value too small to not kill interactive type connections
 * (telnet, SSH) which send many small packets.
 */
#ifdef FIX_WORKAROUND_FOR_3894301
__private_extern__ int     tcp_minmssoverload = TCP_MINMSSOVERLOAD;
#else
__private_extern__ int     tcp_minmssoverload = 0;
#endif
SYSCTL_INT(_net_inet_tcp, OID_AUTO, minmssoverload, CTLFLAG_RW,
    &tcp_minmssoverload , 0, "Number of TCP Segments per Second allowed to"
    "be under the MINMSS Size");

static int      tcp_do_rfc1323 = 1;
SYSCTL_INT(_net_inet_tcp, TCPCTL_DO_RFC1323, rfc1323, CTLFLAG_RW, 
    &tcp_do_rfc1323 , 0, "Enable rfc1323 (high performance TCP) extensions");

static int      tcp_do_rfc1644 = 0;
SYSCTL_INT(_net_inet_tcp, TCPCTL_DO_RFC1644, rfc1644, CTLFLAG_RW, 
    &tcp_do_rfc1644 , 0, "Enable rfc1644 (TTCP) extensions");

static int      tcp_tcbhashsize = 0;
SYSCTL_INT(_net_inet_tcp, OID_AUTO, tcbhashsize, CTLFLAG_RD,
     &tcp_tcbhashsize, 0, "Size of TCP control-block hashtable");

static int      do_tcpdrain = 0;
SYSCTL_INT(_net_inet_tcp, OID_AUTO, do_tcpdrain, CTLFLAG_RW, &do_tcpdrain, 0,
     "Enable tcp_drain routine for extra help when low on mbufs");

SYSCTL_INT(_net_inet_tcp, OID_AUTO, pcbcount, CTLFLAG_RD, 
    &tcbinfo.ipi_count, 0, "Number of active PCBs");

static int      icmp_may_rst = 1;
SYSCTL_INT(_net_inet_tcp, OID_AUTO, icmp_may_rst, CTLFLAG_RW, &icmp_may_rst, 0, 
    "Certain ICMP unreachable messages may abort connections in SYN_SENT");

static int      tcp_strict_rfc1948 = 0;
SYSCTL_INT(_net_inet_tcp, OID_AUTO, strict_rfc1948, CTLFLAG_RW,
    &tcp_strict_rfc1948, 0, "Determines if RFC1948 is followed exactly");

static int      tcp_isn_reseed_interval = 0;
SYSCTL_INT(_net_inet_tcp, OID_AUTO, isn_reseed_interval, CTLFLAG_RW,
    &tcp_isn_reseed_interval, 0, "Seconds between reseeding of ISN secret");
static int      tcp_background_io_enabled = 1;
SYSCTL_INT(_net_inet_tcp, OID_AUTO, background_io_enabled, CTLFLAG_RW,
    &tcp_background_io_enabled, 0, "Background IO Enabled");

int     tcp_TCPTV_MIN = 1;
SYSCTL_INT(_net_inet_tcp, OID_AUTO, rtt_min, CTLFLAG_RW,
    &tcp_TCPTV_MIN, 0, "min rtt value allowed");

static void     tcp_cleartaocache(void);
static void     tcp_notify(struct inpcb *, int);
struct zone     *sack_hole_zone;

extern unsigned int total_mb_cnt;
extern unsigned int total_cl_cnt;
extern int sbspace_factor;
extern int tcp_sockthreshold;
extern int slowlink_wsize;      /* window correction for slow links */
extern int path_mtu_discovery;


/*
 * Target size of TCP PCB hash tables. Must be a power of two.
 *
 * Note that this can be overridden by the kernel environment
 * variable net.inet.tcp.tcbhashsize
 */
#ifndef TCBHASHSIZE
#define TCBHASHSIZE     CONFIG_TCBHASHSIZE
#endif

/*
 * This is the actual shape of what we allocate using the zone
 * allocator.  Doing it this way allows us to protect both structures
 * using the same generation count, and also eliminates the overhead
 * of allocating tcpcbs separately.  By hiding the structure here,
 * we avoid changing most of the rest of the code (although it needs
 * to be changed, eventually, for greater efficiency).
 */
#define ALIGNMENT       32
#define ALIGNM1         (ALIGNMENT - 1)
struct  inp_tp {
        union {
                struct  inpcb inp;
                char    align[(sizeof(struct inpcb) + ALIGNM1) & ~ALIGNM1];
        } inp_tp_u;
        struct  tcpcb tcb;
};
#undef ALIGNMENT
#undef ALIGNM1

static struct tcpcb dummy_tcb;


extern struct   inpcbhead       time_wait_slots[];
extern int              cur_tw_slot;
extern u_long           *delack_bitmask;
extern u_long  route_generation;

int  get_inpcb_str_size(void);
int  get_tcp_str_size(void);


int  get_inpcb_str_size(void)
{
        return sizeof(struct inpcb);
}


int  get_tcp_str_size(void)
{
        return sizeof(struct tcpcb);
}

int     tcp_freeq(struct tcpcb *tp);


/*
 * Tcp initialization
 */
void
tcp_init()
{
        int hashsize = TCBHASHSIZE;
        vm_size_t       str_size;
        int i;
        struct inpcbinfo *pcbinfo;
        
        tcp_ccgen = 1;
        tcp_cleartaocache();

        tcp_keepinit = TCPTV_KEEP_INIT;
        tcp_keepidle = TCPTV_KEEP_IDLE;
        tcp_keepintvl = TCPTV_KEEPINTVL;
        tcp_maxpersistidle = TCPTV_KEEP_IDLE;
        tcp_msl = TCPTV_MSL;
        read_random(&tcp_now, sizeof(tcp_now));
        tcp_now = tcp_now & 0x3fffffff; /* Starts tcp internal 100ms clock at a random value */


        LIST_INIT(&tcb);
        tcbinfo.listhead = &tcb;
        pcbinfo = &tcbinfo;
        if (!powerof2(hashsize)) {
                printf("WARNING: TCB hash size not a power of 2\n");
                hashsize = 512; /* safe default */
        }
        tcp_tcbhashsize = hashsize;
        tcbinfo.hashsize = hashsize;
        tcbinfo.hashbase = hashinit(hashsize, M_PCB, &tcbinfo.hashmask);
        tcbinfo.porthashbase = hashinit(hashsize, M_PCB,
                                        &tcbinfo.porthashmask);
        str_size = (vm_size_t) sizeof(struct inp_tp);
        tcbinfo.ipi_zone = (void *) zinit(str_size, 120000*str_size, 8192, "tcpcb");
        sack_hole_zone = zinit(str_size, 120000*str_size, 8192, "sack_hole zone");
        tcp_reass_maxseg = nmbclusters / 16;

#if INET6
#define TCP_MINPROTOHDR (sizeof(struct ip6_hdr) + sizeof(struct tcphdr))
#else /* INET6 */
#define TCP_MINPROTOHDR (sizeof(struct tcpiphdr))
#endif /* INET6 */
        if (max_protohdr < TCP_MINPROTOHDR)
                max_protohdr = TCP_MINPROTOHDR;
        if (max_linkhdr + TCP_MINPROTOHDR > MHLEN)
                panic("tcp_init");
#undef TCP_MINPROTOHDR
        dummy_tcb.t_state = TCP_NSTATES;
        dummy_tcb.t_flags = 0;
        tcbinfo.dummy_cb = (caddr_t) &dummy_tcb;

        /*
         * allocate lock group attribute and group for tcp pcb mutexes
         */
     pcbinfo->mtx_grp_attr = lck_grp_attr_alloc_init();
        pcbinfo->mtx_grp = lck_grp_alloc_init("tcppcb", pcbinfo->mtx_grp_attr);
                
        /*
         * allocate the lock attribute for tcp pcb mutexes
         */
        pcbinfo->mtx_attr = lck_attr_alloc_init();

        if ((pcbinfo->mtx = lck_rw_alloc_init(pcbinfo->mtx_grp, pcbinfo->mtx_attr)) == NULL) {
                printf("tcp_init: mutex not alloced!\n");
                return; /* pretty much dead if this fails... */
        }


        in_pcb_nat_init(&tcbinfo, AF_INET, IPPROTO_TCP, SOCK_STREAM);

        delack_bitmask = _MALLOC((4 * hashsize)/32, M_PCB, M_WAITOK);
        if (delack_bitmask == 0) 
             panic("Delack Memory");

        for (i=0; i < (tcbinfo.hashsize / 32); i++)
                 delack_bitmask[i] = 0;

        for (i=0; i < N_TIME_WAIT_SLOTS; i++) {
             LIST_INIT(&time_wait_slots[i]);
        }

        timeout(tcp_fasttimo, NULL, hz/TCP_RETRANSHZ);
}

/*
 * Fill in the IP and TCP headers for an outgoing packet, given the tcpcb.
 * tcp_template used to store this data in mbufs, but we now recopy it out
 * of the tcpcb each time to conserve mbufs.
 */
void
tcp_fillheaders(tp, ip_ptr, tcp_ptr)
        struct tcpcb *tp;
        void *ip_ptr;
        void *tcp_ptr;
{
        struct inpcb *inp = tp->t_inpcb;
        struct tcphdr *tcp_hdr = (struct tcphdr *)tcp_ptr;

#if INET6
        if ((inp->inp_vflag & INP_IPV6) != 0) {
                struct ip6_hdr *ip6;

                ip6 = (struct ip6_hdr *)ip_ptr;
                ip6->ip6_flow = (ip6->ip6_flow & ~IPV6_FLOWINFO_MASK) |
                        (inp->in6p_flowinfo & IPV6_FLOWINFO_MASK);
                ip6->ip6_vfc = (ip6->ip6_vfc & ~IPV6_VERSION_MASK) |
                        (IPV6_VERSION & IPV6_VERSION_MASK);
                ip6->ip6_nxt = IPPROTO_TCP;
                ip6->ip6_plen = sizeof(struct tcphdr);
                ip6->ip6_src = inp->in6p_laddr;
                ip6->ip6_dst = inp->in6p_faddr;
                tcp_hdr->th_sum = 0;
        } else
#endif
        {
        struct ip *ip = (struct ip *) ip_ptr;

        ip->ip_vhl = IP_VHL_BORING;
        ip->ip_tos = 0;
        ip->ip_len = 0;
        ip->ip_id = 0;
        ip->ip_off = 0;
        ip->ip_ttl = 0;
        ip->ip_sum = 0;
        ip->ip_p = IPPROTO_TCP;
        ip->ip_src = inp->inp_laddr;
        ip->ip_dst = inp->inp_faddr;
        tcp_hdr->th_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr,
                htons(sizeof(struct tcphdr) + IPPROTO_TCP));
        }

        tcp_hdr->th_sport = inp->inp_lport;
        tcp_hdr->th_dport = inp->inp_fport;
        tcp_hdr->th_seq = 0;
        tcp_hdr->th_ack = 0;
        tcp_hdr->th_x2 = 0;
        tcp_hdr->th_off = 5;
        tcp_hdr->th_flags = 0;
        tcp_hdr->th_win = 0;
        tcp_hdr->th_urp = 0;
}

/*
 * Create template to be used to send tcp packets on a connection.
 * Allocates an mbuf and fills in a skeletal tcp/ip header.  The only
 * use for this function is in keepalives, which use tcp_respond.
 */
struct tcptemp *
tcp_maketemplate(tp)
        struct tcpcb *tp;
{
        struct mbuf *m;
        struct tcptemp *n;

        m = m_get(M_DONTWAIT, MT_HEADER);
        if (m == NULL)
                return (0);
        m->m_len = sizeof(struct tcptemp);
        n = mtod(m, struct tcptemp *);

        tcp_fillheaders(tp, (void *)&n->tt_ipgen, (void *)&n->tt_t);
        return (n);
}

/*
 * Send a single message to the TCP at address specified by
 * the given TCP/IP header.  If m == 0, then we make a copy
 * of the tcpiphdr at ti and send directly to the addressed host.
 * This is used to force keep alive messages out using the TCP
 * template for a connection.  If flags are given then we send
 * a message back to the TCP which originated the * segment ti,
 * and discard the mbuf containing it and any other attached mbufs.
 *
 * In any case the ack and sequence number of the transmitted
 * segment are as specified by the parameters.
 *
 * NOTE: If m != NULL, then ti must point to *inside* the mbuf.
 */
void
tcp_respond(
        struct tcpcb *tp,
        void *ipgen,
        register struct tcphdr *th,
        register struct mbuf *m,
        tcp_seq ack,
        tcp_seq seq,
        int flags,
#if CONFIG_FORCE_OUT_IFP
        ifnet_t ifp
#else
        __unused ifnet_t ifp
#endif
        )
{
        register int tlen;
        int win = 0;
        struct route *ro = 0;
        struct route sro;
        struct ip *ip;
        struct tcphdr *nth;
#if INET6
        struct route_in6 *ro6 = 0;
        struct route_in6 sro6;
        struct ip6_hdr *ip6;
        int isipv6;
#endif /* INET6 */
        int ipflags = 0;

#if INET6
        isipv6 = IP_VHL_V(((struct ip *)ipgen)->ip_vhl) == 6;
        ip6 = ipgen;
#endif /* INET6 */
        ip = ipgen;

        if (tp) {
                if (!(flags & TH_RST)) {
                        win = tcp_sbspace(tp);
                        if (win > (long)TCP_MAXWIN << tp->rcv_scale)
                                win = (long)TCP_MAXWIN << tp->rcv_scale;
                }
#if INET6
                if (isipv6)
                        ro6 = &tp->t_inpcb->in6p_route;
                else
#endif /* INET6 */
                ro = &tp->t_inpcb->inp_route;
        } else {
#if INET6
                if (isipv6) {
                        ro6 = &sro6;
                        bzero(ro6, sizeof *ro6);
                } else
#endif /* INET6 */
                {
                        ro = &sro;
                        bzero(ro, sizeof *ro);
                }
        }
        if (m == 0) {
                m = m_gethdr(M_DONTWAIT, MT_HEADER);    /* MAC-OK */
                if (m == NULL)
                        return;
                tlen = 0;
                m->m_data += max_linkhdr;
#if INET6
                if (isipv6) {
                        bcopy((caddr_t)ip6, mtod(m, caddr_t), 
                              sizeof(struct ip6_hdr));
                        ip6 = mtod(m, struct ip6_hdr *);
                        nth = (struct tcphdr *)(ip6 + 1);
                } else
#endif /* INET6 */
                {
                        bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
                        ip = mtod(m, struct ip *);
                        nth = (struct tcphdr *)(ip + 1);
                }
                bcopy((caddr_t)th, (caddr_t)nth, sizeof(struct tcphdr));
                flags = TH_ACK;
        } else {
                m_freem(m->m_next);
                m->m_next = 0;
                m->m_data = (caddr_t)ipgen;
                /* m_len is set later */
                tlen = 0;
#define xchg(a,b,type) { type t; t=a; a=b; b=t; }
#if INET6
                if (isipv6) {
                        xchg(ip6->ip6_dst, ip6->ip6_src, struct in6_addr);
                        nth = (struct tcphdr *)(ip6 + 1);
                } else
#endif /* INET6 */
              {
                xchg(ip->ip_dst.s_addr, ip->ip_src.s_addr, n_long);
                nth = (struct tcphdr *)(ip + 1);
              }
                if (th != nth) {
                        /*
                         * this is usually a case when an extension header
                         * exists between the IPv6 header and the
                         * TCP header.
                         */
                        nth->th_sport = th->th_sport;
                        nth->th_dport = th->th_dport;
                }
                xchg(nth->th_dport, nth->th_sport, n_short);
#undef xchg
        }
#if INET6
        if (isipv6) {
                ip6->ip6_plen = htons((u_short)(sizeof (struct tcphdr) +
                                                tlen));
                tlen += sizeof (struct ip6_hdr) + sizeof (struct tcphdr);
        } else
#endif
      {
        tlen += sizeof (struct tcpiphdr);
        ip->ip_len = tlen;
        ip->ip_ttl = ip_defttl;
      }
        m->m_len = tlen;
        m->m_pkthdr.len = tlen;
        m->m_pkthdr.rcvif = 0;
#if CONFIG_MACF_NET
        if (tp != NULL && tp->t_inpcb != NULL) {
                /*
                 * Packet is associated with a socket, so allow the
                 * label of the response to reflect the socket label.
                 */
                mac_mbuf_label_associate_inpcb(tp->t_inpcb, m);
        } else {
                /*
                 * Packet is not associated with a socket, so possibly
                 * update the label in place.
                 */
                mac_netinet_tcp_reply(m);
        }
#endif
        
#if CONFIG_IP_EDGEHOLE
        if (tp && tp->t_inpcb)
                ip_edgehole_mbuf_tag(tp->t_inpcb, m);
#endif
        
        nth->th_seq = htonl(seq);
        nth->th_ack = htonl(ack);
        nth->th_x2 = 0;
        nth->th_off = sizeof (struct tcphdr) >> 2;
        nth->th_flags = flags;
        if (tp)
                nth->th_win = htons((u_short) (win >> tp->rcv_scale));
        else
                nth->th_win = htons((u_short)win);
        nth->th_urp = 0;
#if INET6
        if (isipv6) {
                nth->th_sum = 0;
                nth->th_sum = in6_cksum(m, IPPROTO_TCP,
                                        sizeof(struct ip6_hdr),
                                        tlen - sizeof(struct ip6_hdr));
                ip6->ip6_hlim = in6_selecthlim(tp ? tp->t_inpcb : NULL,
                                               ro6 && ro6->ro_rt ?
                                               ro6->ro_rt->rt_ifp :
                                               NULL);
        } else
#endif /* INET6 */
        {
                nth->th_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr,
                htons((u_short)(tlen - sizeof(struct ip) + ip->ip_p)));
                m->m_pkthdr.csum_flags = CSUM_TCP;
                m->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum);
        }
#if TCPDEBUG
        if (tp == NULL || (tp->t_inpcb->inp_socket->so_options & SO_DEBUG))
                tcp_trace(TA_OUTPUT, 0, tp, mtod(m, void *), th, 0);
#endif
#if IPSEC
        if (ipsec_bypass == 0 && ipsec_setsocket(m, tp ? tp->t_inpcb->inp_socket : NULL) != 0) {
                m_freem(m);
                return;
        }
#endif
#if INET6
        if (isipv6) {
                (void)ip6_output(m, NULL, ro6, ipflags, NULL, NULL, 0);
                if (ro6 == &sro6 && ro6->ro_rt) {
                        rtfree(ro6->ro_rt);
                        ro6->ro_rt = NULL;
                }
        } else
#endif /* INET6 */
        {
#if CONFIG_FORCE_OUT_IFP
                ifp = (tp && tp->t_inpcb) ? tp->t_inpcb->pdp_ifp :
                          (ifp && (ifp->if_flags & IFF_POINTOPOINT) != 0) ? ifp : NULL;
#endif
                (void) ip_output_list(m, 0, NULL, ro, ipflags, NULL, ifp);
                if (ro == &sro && ro->ro_rt) {
                        rtfree(ro->ro_rt);
                        ro->ro_rt = NULL;
                }
        }
}

/*
 * Create a new TCP control block, making an
 * empty reassembly queue and hooking it to the argument
 * protocol control block.  The `inp' parameter must have
 * come from the zone allocator set up in tcp_init().
 */
struct tcpcb *
tcp_newtcpcb(inp)
        struct inpcb *inp;
{
        struct inp_tp *it;
        register struct tcpcb *tp;
        register struct socket *so = inp->inp_socket;   
#if INET6
        int isipv6 = (inp->inp_vflag & INP_IPV6) != 0;
#endif /* INET6 */

        if (so->cached_in_sock_layer == 0) {
             it = (struct inp_tp *)inp;
             tp = &it->tcb;
        }
        else
             tp = (struct tcpcb *) inp->inp_saved_ppcb;

        bzero((char *) tp, sizeof(struct tcpcb));
        LIST_INIT(&tp->t_segq);
        tp->t_maxseg = tp->t_maxopd =
#if INET6
                isipv6 ? tcp_v6mssdflt :
#endif /* INET6 */
                tcp_mssdflt;

        if (tcp_do_rfc1323)
                tp->t_flags = (TF_REQ_SCALE|TF_REQ_TSTMP);
        tp->sack_enable = tcp_do_sack;
        TAILQ_INIT(&tp->snd_holes);
        tp->t_inpcb = inp;      /* XXX */
        /*
         * Init srtt to TCPTV_SRTTBASE (0), so we can tell that we have no
         * rtt estimate.  Set rttvar so that srtt + 4 * rttvar gives
         * reasonable initial retransmit time.
         */
        tp->t_srtt = TCPTV_SRTTBASE;
        tp->t_rttvar = ((TCPTV_RTOBASE - TCPTV_SRTTBASE) << TCP_RTTVAR_SHIFT) / 4;
        tp->t_rttmin = tcp_TCPTV_MIN;
        tp->t_rxtcur = TCPTV_RTOBASE;
        tp->snd_cwnd = TCP_MAXWIN << TCP_MAX_WINSHIFT;
        tp->snd_bwnd = TCP_MAXWIN << TCP_MAX_WINSHIFT;
        tp->snd_ssthresh = TCP_MAXWIN << TCP_MAX_WINSHIFT;
        tp->snd_ssthresh_prev = TCP_MAXWIN << TCP_MAX_WINSHIFT;
        tp->t_rcvtime = 0;
        tp->t_bw_rtttime = 0;
        /*
         * IPv4 TTL initialization is necessary for an IPv6 socket as well,
         * because the socket may be bound to an IPv6 wildcard address,
         * which may match an IPv4-mapped IPv6 address.
         */
        inp->inp_ip_ttl = ip_defttl;
        inp->inp_ppcb = (caddr_t)tp;
        return (tp);            /* XXX */
}

/*
 * Drop a TCP connection, reporting
 * the specified error.  If connection is synchronized,
 * then send a RST to peer.
 */
struct tcpcb *
tcp_drop(tp, errno)
        register struct tcpcb *tp;
        int errno;
{
        struct socket *so = tp->t_inpcb->inp_socket;
     
        if (TCPS_HAVERCVDSYN(tp->t_state)) {
                tp->t_state = TCPS_CLOSED;
                (void) tcp_output(tp);
                tcpstat.tcps_drops++;
        } else
                tcpstat.tcps_conndrops++;
        if (errno == ETIMEDOUT && tp->t_softerror)
                errno = tp->t_softerror;
        so->so_error = errno;
        return (tcp_close(tp));
}

/*
 * Close a TCP control block:
 *      discard all space held by the tcp
 *      discard internet protocol block
 *      wake up any sleepers
 */
struct tcpcb *
tcp_close(tp)
        register struct tcpcb *tp;
{
        struct inpcb *inp = tp->t_inpcb;
        struct socket *so = inp->inp_socket;
#if INET6
        int isipv6 = (inp->inp_vflag & INP_IPV6) != 0;
#endif /* INET6 */
        register struct rtentry *rt;
        int dosavessthresh;

        if ( inp->inp_ppcb == NULL) /* tcp_close was called previously, bail */
                return NULL;

        /* Clear the timers before we delete the PCB. */
        {
                int i;
                for (i = 0; i < TCPT_NTIMERS; i++) {
                        tp->t_timer[i] = 0;
                }
        }

        KERNEL_DEBUG(DBG_FNC_TCP_CLOSE | DBG_FUNC_START, tp,0,0,0,0);
        switch (tp->t_state) 
        {
        case TCPS_ESTABLISHED:
        case TCPS_FIN_WAIT_1:
        case TCPS_CLOSING:
        case TCPS_CLOSE_WAIT:
        case TCPS_LAST_ACK:
             break;
        }

        /*
         * If another thread for this tcp is currently in ip (indicated by
         * the TF_SENDINPROG flag), defer the cleanup until after it returns
         * back to tcp.  This is done to serialize the close until after all
         * pending output is finished, in order to avoid having the PCB be
         * detached and the cached route cleaned, only for ip to cache the
         * route back into the PCB again.  Note that we've cleared all the
         * timers at this point.  Set TF_CLOSING to indicate to tcp_output()
         * that is should call us again once it returns from ip; at that
         * point both flags should be cleared and we can proceed further
         * with the cleanup.
         */
        if (tp->t_flags & (TF_CLOSING|TF_SENDINPROG)) {
                tp->t_flags |= TF_CLOSING;
                return (NULL);
        }

        lck_mtx_lock(rt_mtx);
        /*
         * If we got enough samples through the srtt filter,
         * save the rtt and rttvar in the routing entry.
         * 'Enough' is arbitrarily defined as the 16 samples.
         * 16 samples is enough for the srtt filter to converge
         * to within 5% of the correct value; fewer samples and
         * we could save a very bogus rtt.
         *
         * Don't update the default route's characteristics and don't
         * update anything that the user "locked".
         */
        if (tp->t_rttupdated >= 16) {
                register u_long i = 0;

#if INET6
                if (isipv6) {
                        struct sockaddr_in6 *sin6;

                        if ((rt = inp->in6p_route.ro_rt) == NULL)
                                goto no_valid_rt;
                        sin6 = (struct sockaddr_in6 *)rt_key(rt);
                        if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
                                goto no_valid_rt;
                }
                else
#endif /* INET6 */
                rt = inp->inp_route.ro_rt;      
                if (rt == NULL ||
                    ((struct sockaddr_in *)rt_key(rt))->sin_addr.s_addr
                    == INADDR_ANY || rt->generation_id != route_generation) {
                        if (tp->t_state >= TCPS_CLOSE_WAIT)
                                tp->t_state = TCPS_CLOSING;

                        goto no_valid_rt;
                }

                if ((rt->rt_rmx.rmx_locks & RTV_RTT) == 0) {
                        i = tp->t_srtt *
                            (RTM_RTTUNIT / (TCP_RETRANSHZ * TCP_RTT_SCALE));
                        if (rt->rt_rmx.rmx_rtt && i)
                                /*
                                 * filter this update to half the old & half
                                 * the new values, converting scale.
                                 * See route.h and tcp_var.h for a
                                 * description of the scaling constants.
                                 */
                                rt->rt_rmx.rmx_rtt =
                                    (rt->rt_rmx.rmx_rtt + i) / 2;
                        else
                                rt->rt_rmx.rmx_rtt = i;
                        tcpstat.tcps_cachedrtt++;
                }
                if ((rt->rt_rmx.rmx_locks & RTV_RTTVAR) == 0) {
                        i = tp->t_rttvar *
                            (RTM_RTTUNIT / (TCP_RETRANSHZ * TCP_RTTVAR_SCALE));
                        if (rt->rt_rmx.rmx_rttvar && i)
                                rt->rt_rmx.rmx_rttvar =
                                    (rt->rt_rmx.rmx_rttvar + i) / 2;
                        else
                                rt->rt_rmx.rmx_rttvar = i;
                        tcpstat.tcps_cachedrttvar++;
                }
                /*
                 * The old comment here said:
                 * update the pipelimit (ssthresh) if it has been updated
                 * already or if a pipesize was specified & the threshhold
                 * got below half the pipesize.  I.e., wait for bad news
                 * before we start updating, then update on both good
                 * and bad news.
                 *
                 * But we want to save the ssthresh even if no pipesize is
                 * specified explicitly in the route, because such
                 * connections still have an implicit pipesize specified
                 * by the global tcp_sendspace.  In the absence of a reliable
                 * way to calculate the pipesize, it will have to do.
                 */
                i = tp->snd_ssthresh;
                if (rt->rt_rmx.rmx_sendpipe != 0)
                        dosavessthresh = (i < rt->rt_rmx.rmx_sendpipe / 2);
                else
                        dosavessthresh = (i < so->so_snd.sb_hiwat / 2);
                if (((rt->rt_rmx.rmx_locks & RTV_SSTHRESH) == 0 &&
                     i != 0 && rt->rt_rmx.rmx_ssthresh != 0)
                    || dosavessthresh) {
                        /*
                         * convert the limit from user data bytes to
                         * packets then to packet data bytes.
                         */
                        i = (i + tp->t_maxseg / 2) / tp->t_maxseg;
                        if (i < 2)
                                i = 2;
                        i *= (u_long)(tp->t_maxseg +
#if INET6
                                      (isipv6 ? sizeof (struct ip6_hdr) +
                                               sizeof (struct tcphdr) :
#endif
                                       sizeof (struct tcpiphdr)
#if INET6
                                       )
#endif
                                      );
                        if (rt->rt_rmx.rmx_ssthresh)
                                rt->rt_rmx.rmx_ssthresh =
                                    (rt->rt_rmx.rmx_ssthresh + i) / 2;
                        else
                                rt->rt_rmx.rmx_ssthresh = i;
                        tcpstat.tcps_cachedssthresh++;
                }
        }
        rt = inp->inp_route.ro_rt;
        if (rt) {
                /* 
                 * mark route for deletion if no information is
                 * cached.
                 */
                if ((so->so_flags & SOF_OVERFLOW) && tcp_lq_overflow && 
                    ((rt->rt_rmx.rmx_locks & RTV_RTT) == 0)){
                        if (rt->rt_rmx.rmx_rtt == 0)
                                rt->rt_flags |= RTF_DELCLONE;
                }
        }
    no_valid_rt:
        /* free the reassembly queue, if any */
        lck_mtx_unlock(rt_mtx);

        (void) tcp_freeq(tp);

        tcp_free_sackholes(tp);

        /* Free the packet list */
        if (tp->t_pktlist_head != NULL)
                m_freem_list(tp->t_pktlist_head);
        TCP_PKTLIST_CLEAR(tp);

#ifdef __APPLE__
        if (so->cached_in_sock_layer)
            inp->inp_saved_ppcb = (caddr_t) tp;
#endif

        soisdisconnected(so);
#if INET6
        if (INP_CHECK_SOCKAF(so, AF_INET6))
                in6_pcbdetach(inp);
        else
#endif /* INET6 */
        in_pcbdetach(inp);
        tcpstat.tcps_closed++;
        KERNEL_DEBUG(DBG_FNC_TCP_CLOSE | DBG_FUNC_END, tcpstat.tcps_closed,0,0,0,0);
        return ((struct tcpcb *)0);
}

int
tcp_freeq(tp)
        struct tcpcb *tp;
{

        register struct tseg_qent *q;
        int rv = 0;

        while((q = LIST_FIRST(&tp->t_segq)) != NULL) {
                LIST_REMOVE(q, tqe_q);
                m_freem(q->tqe_m);
                FREE(q, M_TSEGQ);
                tcp_reass_qsize--;
                rv = 1;
        }
        return (rv);
}

void
tcp_drain()
{
        if (do_tcpdrain)
        {
                struct inpcb *inpb;
                struct tcpcb *tcpb;
                struct tseg_qent *te;

        /*
         * Walk the tcpbs, if existing, and flush the reassembly queue,
         * if there is one...
         * XXX: The "Net/3" implementation doesn't imply that the TCP
         *      reassembly queue should be flushed, but in a situation
         *      where we're really low on mbufs, this is potentially
         *      usefull.        
         */
                if (!lck_rw_try_lock_exclusive(tcbinfo.mtx)) /* do it next time if the lock is in use */
                        return;

                for (inpb = LIST_FIRST(tcbinfo.listhead); inpb;
                        inpb = LIST_NEXT(inpb, inp_list)) {
                                if ((tcpb = intotcpcb(inpb))) {
                                        while ((te = LIST_FIRST(&tcpb->t_segq))
                                               != NULL) {
                                        LIST_REMOVE(te, tqe_q);
                                        m_freem(te->tqe_m);
                                        FREE(te, M_TSEGQ);
                                        tcp_reass_qsize--;
                                }
                        }
                }
                lck_rw_done(tcbinfo.mtx);

        }
}

/*
 * Notify a tcp user of an asynchronous error;
 * store error as soft error, but wake up user
 * (for now, won't do anything until can select for soft error).
 *
 * Do not wake up user since there currently is no mechanism for
 * reporting soft errors (yet - a kqueue filter may be added).
 */
static void
tcp_notify(inp, error)
        struct inpcb *inp;
        int error;
{
        struct tcpcb *tp;

        if (inp == NULL || (inp->inp_state == INPCB_STATE_DEAD)) 
                return; /* pcb is gone already */

        tp = (struct tcpcb *)inp->inp_ppcb;

        /*
         * Ignore some errors if we are hooked up.
         * If connection hasn't completed, has retransmitted several times,
         * and receives a second error, give up now.  This is better
         * than waiting a long time to establish a connection that
         * can never complete.
         */
        if (tp->t_state == TCPS_ESTABLISHED &&
             (error == EHOSTUNREACH || error == ENETUNREACH ||
              error == EHOSTDOWN)) {
                return;
        } else if (tp->t_state < TCPS_ESTABLISHED && tp->t_rxtshift > 3 &&
            tp->t_softerror)
                tcp_drop(tp, error);
        else
                tp->t_softerror = error;
#if 0
        wakeup((caddr_t) &so->so_timeo);
        sorwakeup(so);
        sowwakeup(so);
#endif
}

static int
tcp_pcblist SYSCTL_HANDLER_ARGS
{
#pragma unused(oidp, arg1, arg2)
        int error, i, n;
        struct inpcb *inp, **inp_list;
        inp_gen_t gencnt;
        struct xinpgen xig;
        int slot;

        /*
         * The process of preparing the TCB list is too time-consuming and
         * resource-intensive to repeat twice on every request.
         */
        lck_rw_lock_shared(tcbinfo.mtx);
        if (req->oldptr == USER_ADDR_NULL) {
                n = tcbinfo.ipi_count;
                req->oldidx = 2 * (sizeof xig)
                        + (n + n/8) * sizeof(struct xtcpcb);
                lck_rw_done(tcbinfo.mtx);
                return 0;
        }

        if (req->newptr != USER_ADDR_NULL) {
                lck_rw_done(tcbinfo.mtx);
                return EPERM;
        }

        /*
         * OK, now we're committed to doing something.
         */
        gencnt = tcbinfo.ipi_gencnt;
        n = tcbinfo.ipi_count;

        bzero(&xig, sizeof(xig));
        xig.xig_len = sizeof xig;
        xig.xig_count = n;
        xig.xig_gen = gencnt;
        xig.xig_sogen = so_gencnt;
        error = SYSCTL_OUT(req, &xig, sizeof xig);
        if (error) {
                lck_rw_done(tcbinfo.mtx);
                return error;
        }
        /*
         * We are done if there is no pcb
         */
        if (n == 0) {
                lck_rw_done(tcbinfo.mtx);
                return 0; 
        }

        inp_list = _MALLOC(n * sizeof *inp_list, M_TEMP, M_WAITOK);
        if (inp_list == 0) {
                lck_rw_done(tcbinfo.mtx);
                return ENOMEM;
        }
        
        for (inp = LIST_FIRST(tcbinfo.listhead), i = 0; inp && i < n;
             inp = LIST_NEXT(inp, inp_list)) {
#ifdef __APPLE__
                if (inp->inp_gencnt <= gencnt && inp->inp_state != INPCB_STATE_DEAD)
#else
                if (inp->inp_gencnt <= gencnt && !prison_xinpcb(req->p, inp))
#endif
                        inp_list[i++] = inp;
        }

        for (slot = 0; slot < N_TIME_WAIT_SLOTS; slot++) {
                struct inpcb *inpnxt;

                for (inp = time_wait_slots[slot].lh_first; inp && i < n; inp = inpnxt) {
                        inpnxt = inp->inp_list.le_next;
                        if (inp->inp_gencnt <= gencnt && inp->inp_state != INPCB_STATE_DEAD)
                                inp_list[i++] = inp;
                }
        }

        n = i;

        error = 0;
        for (i = 0; i < n; i++) {
                inp = inp_list[i];
                if (inp->inp_gencnt <= gencnt && inp->inp_state != INPCB_STATE_DEAD) {
                        struct xtcpcb xt;
                        caddr_t inp_ppcb;

                        bzero(&xt, sizeof(xt));
                        xt.xt_len = sizeof xt;
                        /* XXX should avoid extra copy */
                        inpcb_to_compat(inp, &xt.xt_inp);
                        inp_ppcb = inp->inp_ppcb;
                        if (inp_ppcb != NULL) {
                                bcopy(inp_ppcb, &xt.xt_tp, sizeof xt.xt_tp);
                        }
                        else
                                bzero((char *) &xt.xt_tp, sizeof xt.xt_tp);
                        if (inp->inp_socket)
                                sotoxsocket(inp->inp_socket, &xt.xt_socket);
                        error = SYSCTL_OUT(req, &xt, sizeof xt);
                }
        }
        if (!error) {
                /*
                 * Give the user an updated idea of our state.
                 * If the generation differs from what we told
                 * her before, she knows that something happened
                 * while we were processing this request, and it
                 * might be necessary to retry.
                 */
                bzero(&xig, sizeof(xig));
                xig.xig_len = sizeof xig;
                xig.xig_gen = tcbinfo.ipi_gencnt;
                xig.xig_sogen = so_gencnt;
                xig.xig_count = tcbinfo.ipi_count;
                error = SYSCTL_OUT(req, &xig, sizeof xig);
        }
        FREE(inp_list, M_TEMP);
        lck_rw_done(tcbinfo.mtx);
        return error;
}

SYSCTL_PROC(_net_inet_tcp, TCPCTL_PCBLIST, pcblist, CTLFLAG_RD, 0, 0,
            tcp_pcblist, "S,xtcpcb", "List of active TCP connections");

#ifndef __APPLE__
static int
tcp_getcred(SYSCTL_HANDLER_ARGS)
{
        struct sockaddr_in addrs[2];
        struct inpcb *inp;
        int error, s;

        error = suser(req->p);
        if (error)
                return (error);
        error = SYSCTL_IN(req, addrs, sizeof(addrs));
        if (error)
                return (error);
        s = splnet();
        inp = in_pcblookup_hash(&tcbinfo, addrs[1].sin_addr, addrs[1].sin_port,
            addrs[0].sin_addr, addrs[0].sin_port, 0, NULL);
        if (inp == NULL || inp->inp_socket == NULL) {
                error = ENOENT;
                goto out;
        }
        error = SYSCTL_OUT(req, inp->inp_socket->so_cred, sizeof(*(kauth_cred_t)0);
out:
        splx(s);
        return (error);
}

SYSCTL_PROC(_net_inet_tcp, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW,
    0, 0, tcp_getcred, "S,ucred", "Get the ucred of a TCP connection");

#if INET6
static int
tcp6_getcred(SYSCTL_HANDLER_ARGS)
{
        struct sockaddr_in6 addrs[2];
        struct inpcb *inp;
        int error, s, mapped = 0;

        error = suser(req->p);
        if (error)
                return (error);
        error = SYSCTL_IN(req, addrs, sizeof(addrs));
        if (error)
                return (error);
        if (IN6_IS_ADDR_V4MAPPED(&addrs[0].sin6_addr)) {
                if (IN6_IS_ADDR_V4MAPPED(&addrs[1].sin6_addr))
                        mapped = 1;
                else
                        return (EINVAL);
        }
        s = splnet();
        if (mapped == 1)
                inp = in_pcblookup_hash(&tcbinfo,
                        *(struct in_addr *)&addrs[1].sin6_addr.s6_addr[12],
                        addrs[1].sin6_port,
                        *(struct in_addr *)&addrs[0].sin6_addr.s6_addr[12],
                        addrs[0].sin6_port,
                        0, NULL);
        else
                inp = in6_pcblookup_hash(&tcbinfo, &addrs[1].sin6_addr,
                                 addrs[1].sin6_port,
                                 &addrs[0].sin6_addr, addrs[0].sin6_port,
                                 0, NULL);
        if (inp == NULL || inp->inp_socket == NULL) {
                error = ENOENT;
                goto out;
        }
        error = SYSCTL_OUT(req, inp->inp_socket->so_cred, 
                           sizeof(*(kauth_cred_t)0);
out:
        splx(s);
        return (error);
}

SYSCTL_PROC(_net_inet6_tcp6, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW,
            0, 0,
            tcp6_getcred, "S,ucred", "Get the ucred of a TCP6 connection");
#endif
#endif /* __APPLE__*/

void
tcp_ctlinput(cmd, sa, vip)
        int cmd;
        struct sockaddr *sa;
        void *vip;
{
        struct ip *ip = vip;
        struct tcphdr *th;
        struct in_addr faddr;
        struct inpcb *inp;
        struct tcpcb *tp;
        void (*notify)(struct inpcb *, int) = tcp_notify;
        tcp_seq icmp_seq;

        faddr = ((struct sockaddr_in *)sa)->sin_addr;
        if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY)
                return;

        if (cmd == PRC_QUENCH)
                notify = tcp_quench;
        else if (icmp_may_rst && (cmd == PRC_UNREACH_ADMIN_PROHIB ||
                cmd == PRC_UNREACH_PORT) && ip)
                notify = tcp_drop_syn_sent;
        else if (cmd == PRC_MSGSIZE)
                notify = tcp_mtudisc;
        else if (PRC_IS_REDIRECT(cmd)) {
                ip = 0;
                notify = in_rtchange;
        } else if (cmd == PRC_HOSTDEAD)
                ip = 0;
        else if ((unsigned)cmd > PRC_NCMDS || inetctlerrmap[cmd] == 0)
                return;
        if (ip) {
                th = (struct tcphdr *)((caddr_t)ip 
                                       + (IP_VHL_HL(ip->ip_vhl) << 2));
                inp = in_pcblookup_hash(&tcbinfo, faddr, th->th_dport,
                    ip->ip_src, th->th_sport, 0, NULL);
                if (inp != NULL && inp->inp_socket != NULL) {
                        tcp_lock(inp->inp_socket, 1, 0);
                        if (in_pcb_checkstate(inp, WNT_RELEASE, 1) == WNT_STOPUSING) {
                                tcp_unlock(inp->inp_socket, 1, 0);
                                return;
                        }
                        icmp_seq = htonl(th->th_seq);
                        tp = intotcpcb(inp);
                        if (SEQ_GEQ(icmp_seq, tp->snd_una) &&
                            SEQ_LT(icmp_seq, tp->snd_max))
                                (*notify)(inp, inetctlerrmap[cmd]);
                        tcp_unlock(inp->inp_socket, 1, 0);
                }
        } else
                in_pcbnotifyall(&tcbinfo, faddr, inetctlerrmap[cmd], notify);
}

#if INET6
void
tcp6_ctlinput(cmd, sa, d)
        int cmd;
        struct sockaddr *sa;
        void *d;
{
        struct tcphdr th;
        void (*notify)(struct inpcb *, int) = tcp_notify;
        struct ip6_hdr *ip6;
        struct mbuf *m;
        struct ip6ctlparam *ip6cp = NULL;
        const struct sockaddr_in6 *sa6_src = NULL;
        int off;
        struct tcp_portonly {
                u_int16_t th_sport;
                u_int16_t th_dport;
        } *thp;

        if (sa->sa_family != AF_INET6 ||
            sa->sa_len != sizeof(struct sockaddr_in6))
                return;

        if (cmd == PRC_QUENCH)
                notify = tcp_quench;
        else if (cmd == PRC_MSGSIZE)
                notify = tcp_mtudisc;
        else if (!PRC_IS_REDIRECT(cmd) &&
                 ((unsigned)cmd > PRC_NCMDS || inet6ctlerrmap[cmd] == 0))
                return;

        /* if the parameter is from icmp6, decode it. */
        if (d != NULL) {
                ip6cp = (struct ip6ctlparam *)d;
                m = ip6cp->ip6c_m;
                ip6 = ip6cp->ip6c_ip6;
                off = ip6cp->ip6c_off;
                sa6_src = ip6cp->ip6c_src;
        } else {
                m = NULL;
                ip6 = NULL;
                off = 0;        /* fool gcc */
                sa6_src = &sa6_any;
        }

        if (ip6) {
                /*
                 * XXX: We assume that when IPV6 is non NULL,
                 * M and OFF are valid.
                 */

                /* check if we can safely examine src and dst ports */
                if (m->m_pkthdr.len < off + sizeof(*thp))
                        return;

                bzero(&th, sizeof(th));
                m_copydata(m, off, sizeof(*thp), (caddr_t)&th);

                in6_pcbnotify(&tcbinfo, sa, th.th_dport,
                    (struct sockaddr *)ip6cp->ip6c_src,
                    th.th_sport, cmd, notify);
        } else
                in6_pcbnotify(&tcbinfo, sa, 0, (struct sockaddr *)sa6_src,
                              0, cmd, notify);
}
#endif /* INET6 */


/*
 * Following is where TCP initial sequence number generation occurs.
 *
 * There are two places where we must use initial sequence numbers:
 * 1.  In SYN-ACK packets.
 * 2.  In SYN packets.
 *
 * The ISNs in SYN-ACK packets have no monotonicity requirement, 
 * and should be as unpredictable as possible to avoid the possibility
 * of spoofing and/or connection hijacking.  To satisfy this
 * requirement, SYN-ACK ISNs are generated via the arc4random()
 * function.  If exact RFC 1948 compliance is requested via sysctl,
 * these ISNs will be generated just like those in SYN packets.
 *
 * The ISNs in SYN packets must be monotonic; TIME_WAIT recycling
 * depends on this property.  In addition, these ISNs should be
 * unguessable so as to prevent connection hijacking.  To satisfy
 * the requirements of this situation, the algorithm outlined in
 * RFC 1948 is used to generate sequence numbers.
 *
 * For more information on the theory of operation, please see
 * RFC 1948.
 *
 * Implementation details:
 *
 * Time is based off the system timer, and is corrected so that it
 * increases by one megabyte per second.  This allows for proper
 * recycling on high speed LANs while still leaving over an hour
 * before rollover.
 *
 * Two sysctls control the generation of ISNs:
 *
 * net.inet.tcp.isn_reseed_interval controls the number of seconds
 * between seeding of isn_secret.  This is normally set to zero,
 * as reseeding should not be necessary.
 *
 * net.inet.tcp.strict_rfc1948 controls whether RFC 1948 is followed
 * strictly.  When strict compliance is requested, reseeding is
 * disabled and SYN-ACKs will be generated in the same manner as
 * SYNs.  Strict mode is disabled by default.
 *
 */

#define ISN_BYTES_PER_SECOND 1048576

tcp_seq
tcp_new_isn(tp)
        struct tcpcb *tp;
{
        u_int32_t md5_buffer[4];
        tcp_seq new_isn;
        struct timeval timenow;
        u_char isn_secret[32];
        int isn_last_reseed = 0;
        MD5_CTX isn_ctx;

        /* Use arc4random for SYN-ACKs when not in exact RFC1948 mode. */
        if (((tp->t_state == TCPS_LISTEN) || (tp->t_state == TCPS_TIME_WAIT))
           && tcp_strict_rfc1948 == 0)
#ifdef __APPLE__
                return random();
#else
                return arc4random();
#endif
        getmicrotime(&timenow);

        /* Seed if this is the first use, reseed if requested. */
        if ((isn_last_reseed == 0) ||
            ((tcp_strict_rfc1948 == 0) && (tcp_isn_reseed_interval > 0) &&
             (((u_int)isn_last_reseed + (u_int)tcp_isn_reseed_interval*hz)
                < (u_int)timenow.tv_sec))) {
#ifdef __APPLE__
                read_random(&isn_secret, sizeof(isn_secret));
#else
                read_random_unlimited(&isn_secret, sizeof(isn_secret));
#endif
                isn_last_reseed = timenow.tv_sec;
        }
                
        /* Compute the md5 hash and return the ISN. */
        MD5Init(&isn_ctx);
        MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_fport, sizeof(u_short));
        MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_lport, sizeof(u_short));
#if INET6
        if ((tp->t_inpcb->inp_vflag & INP_IPV6) != 0) {
                MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->in6p_faddr,
                          sizeof(struct in6_addr));
                MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->in6p_laddr,
                          sizeof(struct in6_addr));
        } else
#endif
        {
                MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_faddr,
                          sizeof(struct in_addr));
                MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_laddr,
                          sizeof(struct in_addr));
        }
        MD5Update(&isn_ctx, (u_char *) &isn_secret, sizeof(isn_secret));
        MD5Final((u_char *) &md5_buffer, &isn_ctx);
        new_isn = (tcp_seq) md5_buffer[0];
        new_isn += timenow.tv_sec * (ISN_BYTES_PER_SECOND / hz);
        return new_isn;
}

/*
 * When a source quench is received, close congestion window
 * to one segment.  We will gradually open it again as we proceed.
 */
void
tcp_quench(
        struct inpcb *inp,
        __unused int errno
)
{
        struct tcpcb *tp = intotcpcb(inp);

        if (tp) {
                tp->snd_cwnd = tp->t_maxseg;
                tp->t_bytes_acked = 0;
        }
}

/*
 * When a specific ICMP unreachable message is received and the
 * connection state is SYN-SENT, drop the connection.  This behavior
 * is controlled by the icmp_may_rst sysctl.
 */
void
tcp_drop_syn_sent(inp, errno)
        struct inpcb *inp;
        int errno;
{
        struct tcpcb *tp = intotcpcb(inp);

        if (tp && tp->t_state == TCPS_SYN_SENT)
                tcp_drop(tp, errno);
}

/*
 * When `need fragmentation' ICMP is received, update our idea of the MSS
 * based on the new value in the route.  Also nudge TCP to send something,
 * since we know the packet we just sent was dropped.
 * This duplicates some code in the tcp_mss() function in tcp_input.c.
 */
void
tcp_mtudisc(
        struct inpcb *inp,
        __unused int errno
)
{
        struct tcpcb *tp = intotcpcb(inp);
        struct rtentry *rt;
        struct rmxp_tao *taop;
        struct socket *so = inp->inp_socket;
        int offered;
        int mss;
#if INET6
        int isipv6 = (tp->t_inpcb->inp_vflag & INP_IPV6) != 0;
#endif /* INET6 */

        if (tp) {
                lck_mtx_lock(rt_mtx);
#if INET6
                if (isipv6)
                        rt = tcp_rtlookup6(inp);
                else
#endif /* INET6 */
                rt = tcp_rtlookup(inp);
                if (!rt || !rt->rt_rmx.rmx_mtu) {
                        tp->t_maxopd = tp->t_maxseg =
#if INET6
                                isipv6 ? tcp_v6mssdflt :
#endif /* INET6 */
                                tcp_mssdflt;
                        lck_mtx_unlock(rt_mtx);
                        return;
                }
                taop = rmx_taop(rt->rt_rmx);
                offered = taop->tao_mssopt;
                mss = rt->rt_rmx.rmx_mtu -
#if INET6
                        (isipv6 ?
                         sizeof(struct ip6_hdr) + sizeof(struct tcphdr) :
#endif /* INET6 */
                         sizeof(struct tcpiphdr)
#if INET6
                         )
#endif /* INET6 */
                        ;

                lck_mtx_unlock(rt_mtx);
                if (offered)
                        mss = min(mss, offered);
                /*
                 * XXX - The above conditional probably violates the TCP
                 * spec.  The problem is that, since we don't know the
                 * other end's MSS, we are supposed to use a conservative
                 * default.  But, if we do that, then MTU discovery will
                 * never actually take place, because the conservative
                 * default is much less than the MTUs typically seen
                 * on the Internet today.  For the moment, we'll sweep
                 * this under the carpet.
                 *
                 * The conservative default might not actually be a problem
                 * if the only case this occurs is when sending an initial
                 * SYN with options and data to a host we've never talked
                 * to before.  Then, they will reply with an MSS value which
                 * will get recorded and the new parameters should get
                 * recomputed.  For Further Study.
                 */
                if (tp->t_maxopd <= mss)
                        return;
                tp->t_maxopd = mss;

                if ((tp->t_flags & (TF_REQ_TSTMP|TF_NOOPT)) == TF_REQ_TSTMP &&
                    (tp->t_flags & TF_RCVD_TSTMP) == TF_RCVD_TSTMP)
                        mss -= TCPOLEN_TSTAMP_APPA;

                if (so->so_snd.sb_hiwat < mss)
                        mss = so->so_snd.sb_hiwat;

                tp->t_maxseg = mss;

                tcpstat.tcps_mturesent++;
                tp->t_rtttime = 0;
                tp->snd_nxt = tp->snd_una;
                tcp_output(tp);
        }
}

/*
 * Look-up the routing entry to the peer of this inpcb.  If no route
 * is found and it cannot be allocated then return NULL.  This routine
 * is called by TCP routines that access the rmx structure and by tcp_mss
 * to get the interface MTU.
 */
struct rtentry *
tcp_rtlookup(inp)
        struct inpcb *inp;
{
        struct route *ro;
        struct rtentry *rt;
        struct tcpcb *tp;

        ro = &inp->inp_route;
        if (ro == NULL)
                return (NULL);
        rt = ro->ro_rt;

        lck_mtx_assert(rt_mtx, LCK_MTX_ASSERT_OWNED);

        if (rt == NULL || !(rt->rt_flags & RTF_UP) || rt->generation_id != route_generation) {
                /* No route yet, so try to acquire one */
                if (inp->inp_faddr.s_addr != INADDR_ANY) {
                        ro->ro_dst.sa_family = AF_INET;
                        ro->ro_dst.sa_len = sizeof(struct sockaddr_in);
                        ((struct sockaddr_in *) &ro->ro_dst)->sin_addr =
                                inp->inp_faddr;
                        rtalloc_ign_locked(ro, 0UL);
                        rt = ro->ro_rt;
                }
        }
        if (rt != NULL && rt->rt_ifp != NULL)
                somultipages(inp->inp_socket,
                    (rt->rt_ifp->if_hwassist & IFNET_MULTIPAGES));

        /*
         * Update MTU discovery determination. Don't do it if:
         *      1) it is disabled via the sysctl
         *      2) the route isn't up
         *      3) the MTU is locked (if it is, then discovery has been
         *         disabled)
         */

         tp = intotcpcb(inp);

        if (!path_mtu_discovery || ((rt != NULL) && 
            (!(rt->rt_flags & RTF_UP) || (rt->rt_rmx.rmx_locks & RTV_MTU)))) 
                tp->t_flags &= ~TF_PMTUD;
        else
                tp->t_flags |= TF_PMTUD;

#ifdef IFEF_NOWINDOWSCALE
        if (tp->t_state == TCPS_SYN_SENT && rt != NULL && rt->rt_ifp != NULL &&
                (rt->rt_ifp->if_eflags & IFEF_NOWINDOWSCALE) != 0)
        {
                // Timestamps are not enabled on this interface
                tp->t_flags &= ~(TF_REQ_SCALE);
        }
#endif

        return rt;
}

#if INET6
struct rtentry *
tcp_rtlookup6(inp)
        struct inpcb *inp;
{
        struct route_in6 *ro6;
        struct rtentry *rt;
        struct tcpcb *tp;

        lck_mtx_assert(rt_mtx, LCK_MTX_ASSERT_OWNED);

        ro6 = &inp->in6p_route;
        rt = ro6->ro_rt;
        if (rt == NULL || !(rt->rt_flags & RTF_UP)) {
                /* No route yet, so try to acquire one */
                if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
                        struct sockaddr_in6 *dst6;

                        dst6 = (struct sockaddr_in6 *)&ro6->ro_dst;
                        dst6->sin6_family = AF_INET6;
                        dst6->sin6_len = sizeof(*dst6);
                        dst6->sin6_addr = inp->in6p_faddr;
                        rtalloc_ign_locked((struct route *)ro6, 0UL);
                        rt = ro6->ro_rt;
                }
        }
        if (rt != NULL && rt->rt_ifp != NULL)
                somultipages(inp->inp_socket,
                    (rt->rt_ifp->if_hwassist & IFNET_MULTIPAGES));
        /*
         * Update path MTU Discovery determination
         * while looking up the route:
         *  1) we have a valid route to the destination
         *  2) the MTU is not locked (if it is, then discovery has been
         *    disabled)
         */


         tp = intotcpcb(inp);

        /*
         * Update MTU discovery determination. Don't do it if:
         *      1) it is disabled via the sysctl
         *      2) the route isn't up
         *      3) the MTU is locked (if it is, then discovery has been
         *         disabled)
         */

        if (!path_mtu_discovery || ((rt != NULL) && 
            (!(rt->rt_flags & RTF_UP) || (rt->rt_rmx.rmx_locks & RTV_MTU)))) 
                tp->t_flags &= ~TF_PMTUD;
        else
                tp->t_flags |= TF_PMTUD;

        return rt;
}
#endif /* INET6 */

#if IPSEC
/* compute ESP/AH header size for TCP, including outer IP header. */
size_t
ipsec_hdrsiz_tcp(tp)
        struct tcpcb *tp;
{
        struct inpcb *inp;
        struct mbuf *m;
        size_t hdrsiz;
        struct ip *ip;
#if INET6
        struct ip6_hdr *ip6 = NULL;
#endif /* INET6 */
        struct tcphdr *th;

        if ((tp == NULL) || ((inp = tp->t_inpcb) == NULL))
                return 0;
        MGETHDR(m, M_DONTWAIT, MT_DATA);        /* MAC-OK */
        if (!m)
                return 0;

#if INET6
        if ((inp->inp_vflag & INP_IPV6) != 0) {
                ip6 = mtod(m, struct ip6_hdr *);
                th = (struct tcphdr *)(ip6 + 1);
                m->m_pkthdr.len = m->m_len =
                        sizeof(struct ip6_hdr) + sizeof(struct tcphdr);
                tcp_fillheaders(tp, ip6, th);
                hdrsiz = ipsec6_hdrsiz(m, IPSEC_DIR_OUTBOUND, inp);
        } else
#endif /* INET6 */
      {
        ip = mtod(m, struct ip *);
        th = (struct tcphdr *)(ip + 1);
        m->m_pkthdr.len = m->m_len = sizeof(struct tcpiphdr);
        tcp_fillheaders(tp, ip, th);
        hdrsiz = ipsec4_hdrsiz(m, IPSEC_DIR_OUTBOUND, inp);
      }
        m_free(m);
        return hdrsiz;
}
#endif /*IPSEC*/

/*
 * Return a pointer to the cached information about the remote host.
 * The cached information is stored in the protocol specific part of
 * the route metrics.
 */
struct rmxp_tao *
tcp_gettaocache(inp)
        struct inpcb *inp;
{
        struct rtentry *rt;
        struct rmxp_tao *taop;

        lck_mtx_lock(rt_mtx);
#if INET6
        if ((inp->inp_vflag & INP_IPV6) != 0)
                rt = tcp_rtlookup6(inp);
        else
#endif /* INET6 */
        rt = tcp_rtlookup(inp);

        /* Make sure this is a host route and is up. */
        if (rt == NULL ||
            (rt->rt_flags & (RTF_UP|RTF_HOST)) != (RTF_UP|RTF_HOST)) {
                lck_mtx_unlock(rt_mtx);
                return NULL;
        }
        
        taop = rmx_taop(rt->rt_rmx);
        lck_mtx_unlock(rt_mtx);
        return (taop);
}

/*
 * Clear all the TAO cache entries, called from tcp_init.
 *
 * XXX
 * This routine is just an empty one, because we assume that the routing
 * routing tables are initialized at the same time when TCP, so there is
 * nothing in the cache left over.
 */
static void
tcp_cleartaocache()
{
}

int
tcp_lock(so, refcount, lr)
        struct socket *so;
        int refcount;
        int lr;
{
        int lr_saved;
        if (lr == 0) 
                lr_saved = (unsigned int) __builtin_return_address(0);
        else lr_saved = lr;

        if (so->so_pcb) {
                lck_mtx_lock(((struct inpcb *)so->so_pcb)->inpcb_mtx);
        }
        else  {
                panic("tcp_lock: so=%p NO PCB! lr=%x\n", so, lr_saved);
                lck_mtx_lock(so->so_proto->pr_domain->dom_mtx);
        }

        if (so->so_usecount < 0)
                panic("tcp_lock: so=%p so_pcb=%p lr=%x ref=%x\n",
                so, so->so_pcb, lr_saved, so->so_usecount);

        if (refcount)
                so->so_usecount++;
        so->lock_lr[so->next_lock_lr] = (u_int32_t)lr_saved;
        so->next_lock_lr = (so->next_lock_lr+1) % SO_LCKDBG_MAX;
        return (0);
}

int
tcp_unlock(so, refcount, lr)
        struct socket *so;
        int refcount;
        int lr;
{
        int lr_saved;
        if (lr == 0) 
                lr_saved = (unsigned int) __builtin_return_address(0);
        else lr_saved = lr;

#ifdef MORE_TCPLOCK_DEBUG
        printf("tcp_unlock: so=%p sopcb=%x lock=%x ref=%x lr=%x\n", 
                so, so->so_pcb, ((struct inpcb *)so->so_pcb)->inpcb_mtx, so->so_usecount, lr_saved);
#endif
        if (refcount)
                so->so_usecount--;

        if (so->so_usecount < 0)
                panic("tcp_unlock: so=%p usecount=%x\n", so, so->so_usecount);  
        if (so->so_pcb == NULL) 
                panic("tcp_unlock: so=%p NO PCB usecount=%x lr=%x\n", so, so->so_usecount, lr_saved);
        else {
                lck_mtx_assert(((struct inpcb *)so->so_pcb)->inpcb_mtx, LCK_MTX_ASSERT_OWNED);
                so->unlock_lr[so->next_unlock_lr] = (u_int32_t)lr_saved;
                so->next_unlock_lr = (so->next_unlock_lr+1) % SO_LCKDBG_MAX;
                lck_mtx_unlock(((struct inpcb *)so->so_pcb)->inpcb_mtx);
        }
        return (0);
}

lck_mtx_t *
tcp_getlock(
        struct socket *so,
        __unused int locktype)
{
        struct inpcb *inp = sotoinpcb(so);

        if (so->so_pcb)  {
                if (so->so_usecount < 0)
                        panic("tcp_getlock: so=%p usecount=%x\n", so, so->so_usecount); 
                return(inp->inpcb_mtx);
        }
        else {
                panic("tcp_getlock: so=%p NULL so_pcb\n", so);
                return (so->so_proto->pr_domain->dom_mtx);
        }
}
long
tcp_sbspace(struct tcpcb *tp)
{
        struct sockbuf *sb = &tp->t_inpcb->inp_socket->so_rcv;
        long space, newspace;

        space =  ((long) lmin((sb->sb_hiwat - sb->sb_cc), 
                (sb->sb_mbmax - sb->sb_mbcnt)));

#if TRAFFIC_MGT
        if (tp->t_inpcb->inp_socket->so_traffic_mgt_flags & TRAFFIC_MGT_SO_BACKGROUND) {
                if (tcp_background_io_enabled &&
                        tp->t_inpcb->inp_socket->so_traffic_mgt_flags & TRAFFIC_MGT_SO_BG_SUPPRESSED) {
                        tp->t_flags |= TF_RXWIN0SENT;
                        return 0; /* Triggers TCP window closing by responding there is no space */
                }
        }
#endif /* TRAFFIC_MGT */

        /* Avoid inscreasing window size if the current window
         * is already very low, we could be in "persist" mode and
         * we could break some apps (see rdar://5409343)
         */

        if (space < tp->t_maxseg) 
                return space;

        /* Clip window size for slower link */ 

        if (((tp->t_flags & TF_SLOWLINK) != 0) && slowlink_wsize > 0 )  
                return lmin(space, slowlink_wsize);

        /*
         * Check for ressources constraints before over-ajusting the amount of space we can
         * advertise in the TCP window size updates.
         */

        if (sbspace_factor && (tp->t_inpcb->inp_pcbinfo->ipi_count < tcp_sockthreshold) &&
                    (total_mb_cnt / 8) < (mbstat.m_clusters / sbspace_factor)) {
                if (space < (long)(sb->sb_maxused - sb->sb_cc)) {/* make sure we don't constrain the window if we have enough ressources */
                        space = (long) lmax((sb->sb_maxused - sb->sb_cc), tp->rcv_maxbyps);
                }
                newspace = (long) lmax(((long)sb->sb_maxused - sb->sb_cc), (long)tp->rcv_maxbyps);

                if (newspace > space)
                        space = newspace;
        }
        return space;
}
/* DSEP Review Done pl-20051213-v02 @3253,@3391,@3400 */