/*
 * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
 * Use is subject to license terms.
 */

/*
 * kdc/network.c
 *
 * Copyright 1990,2000 by the Massachusetts Institute of Technology.
 *
 * Export of this software from the United States of America may
 *   require a specific license from the United States Government.
 *   It is the responsibility of any person or organization contemplating
 *   export to obtain such a license before exporting.
 *
 * WITHIN THAT CONSTRAINT, permission to use, copy, modify, and
 * distribute this software and its documentation for any purpose and
 * without fee is hereby granted, provided that the above copyright
 * notice appear in all copies and that both that copyright notice and
 * this permission notice appear in supporting documentation, and that
 * the name of M.I.T. not be used in advertising or publicity pertaining
 * to distribution of the software without specific, written prior
 * permission.  Furthermore if you modify this software you must label
 * your software as modified software and not distribute it in such a
 * fashion that it might be confused with the original M.I.T. software.
 * M.I.T. makes no representations about the suitability of
 * this software for any purpose.  It is provided "as is" without express
 * or implied warranty.
 *
 *
 * Network code for Kerberos v5 KDC.
 */

#include "k5-int.h"
#include "com_err.h"
#include "kdc_util.h"
#include "extern.h"
#include "kdc5_err.h"
#include "adm_proto.h"
#include <sys/ioctl.h>
#include <syslog.h>

#include <stddef.h>
#include <ctype.h>
#include "port-sockets.h"
/* #include "socket-utils.h" */

#ifdef HAVE_NETINET_IN_H
#include <sys/types.h>
#include <netinet/in.h>
#include <sys/socket.h>
#ifdef HAVE_SYS_SOCKIO_H
/* for SIOCGIFCONF, etc. */
#include <sys/sockio.h>
#endif
#include <sys/time.h>
#include <libintl.h>

#if HAVE_SYS_SELECT_H
#include <sys/select.h>
#endif
#include <arpa/inet.h>
#include <inet/ip.h>
#include <inet/ip6.h>

#ifndef ARPHRD_ETHER /* OpenBSD breaks on multiple inclusions */
#include <net/if.h>
#endif

#ifdef HAVE_SYS_FILIO_H
#include <sys/filio.h>          /* FIONBIO */
#endif

#include "fake-addrinfo.h"

/* Misc utility routines.  */
static void
set_sa_port(struct sockaddr *addr, int port)
{
    switch (addr->sa_family) {
    case AF_INET:
        sa2sin(addr)->sin_port = port;
        break;
#ifdef KRB5_USE_INET6
    case AF_INET6:
        sa2sin6(addr)->sin6_port = port;
        break;
#endif
    default:
        break;
    }
}

static int ipv6_enabled()
{
#ifdef KRB5_USE_INET6
    static int result = -1;
    if (result == -1) {
        int s;
        s = socket(AF_INET6, SOCK_STREAM, 0);
        if (s >= 0) {
            result = 1;
            close(s);
        } else
            result = 0;
    }
    return result;
#else
    return 0;
#endif
}

static int
setreuseaddr(int sock, int value)
{
    return setsockopt(sock, SOL_SOCKET, SO_REUSEADDR, &value, sizeof(value));
}

#if defined(KRB5_USE_INET6) && defined(IPV6_V6ONLY)
static int
setv6only(int sock, int value)
{
    return setsockopt(sock, IPPROTO_IPV6, IPV6_V6ONLY, &value, sizeof(value));
}
#endif


static const char *paddr (struct sockaddr *sa)
{
    static char buf[100];
    char portbuf[10];
    if (getnameinfo(sa, socklen(sa),
                    buf, sizeof(buf), portbuf, sizeof(portbuf),
                    NI_NUMERICHOST|NI_NUMERICSERV))
        strcpy(buf, "<unprintable>");
    else {
        unsigned int len = sizeof(buf) - strlen(buf);
        char *p = buf + strlen(buf);
        if (len > 2+strlen(portbuf)) {
            *p++ = '.';
            len--;
            strncpy(p, portbuf, len);
        }
    }
    return buf;
}

/* KDC data.  */

enum kdc_conn_type { CONN_UDP, CONN_TCP_LISTENER, CONN_TCP };

/* Per-connection info.  */
struct connection {
    int fd;
    enum kdc_conn_type type;
    void (*service)(struct connection *, const char *, int);
    /* Solaris Kerberos: for auditing */
    in_port_t port; /* local port */
    union {
        /* Type-specific information.  */
        struct {
            int x;
        } udp;
        struct {
            int x;
        } tcp_listener;
        struct {
            /* connection */
            struct sockaddr_storage addr_s;
            socklen_t addrlen;
            char addrbuf[56];
            krb5_fulladdr faddr;
            krb5_address kaddr;
            /* incoming */
            size_t bufsiz;
            size_t offset;
            char *buffer;
            size_t msglen;
            /* outgoing */
            krb5_data *response;
            unsigned char lenbuf[4];
            sg_buf sgbuf[2];
            sg_buf *sgp;
            int sgnum;
            /* crude denial-of-service avoidance support */
            time_t start_time;
        } tcp;
    } u;
};


#define SET(TYPE) struct { TYPE *data; int n, max; }

/* Start at the top and work down -- this should allow for deletions
   without disrupting the iteration, since we delete by overwriting
   the element to be removed with the last element.  */
#define FOREACH_ELT(set,idx,vvar) \
  for (idx = set.n-1; idx >= 0 && (vvar = set.data[idx], 1); idx--)

#define GROW_SET(set, incr, tmpptr) \
  (((int)(set.max + incr) < set.max                                     \
    || (((size_t)((int)(set.max + incr) * sizeof(set.data[0]))          \
         / sizeof(set.data[0]))                                         \
        != (set.max + incr)))                                           \
   ? 0                          /* overflow */                          \
   : ((tmpptr = realloc(set.data,                                       \
                        (int)(set.max + incr) * sizeof(set.data[0])))   \
      ? (set.data = tmpptr, set.max += incr, 1)                         \
      : 0))

/* 1 = success, 0 = failure */
#define ADD(set, val, tmpptr) \
  ((set.n < set.max || GROW_SET(set, 10, tmpptr))                       \
   ? (set.data[set.n++] = val, 1)                                       \
   : 0)

#define DEL(set, idx) \
  (set.data[idx] = set.data[--set.n])

#define FREE_SET_DATA(set) if(set.data) free(set.data);                 \
   (set.data = 0, set.max = 0)


/* Set<struct connection *> connections; */
static SET(struct connection *) connections;
#define n_sockets       connections.n
#define conns           connections.data

/* Set<u_short> udp_port_data, tcp_port_data; */
static SET(u_short) udp_port_data, tcp_port_data;

#include "cm.h"

static struct select_state sstate;

static krb5_error_code add_udp_port(int port)
{
    int i;
    void *tmp;
    u_short val;
    u_short s_port = port;

    if (s_port != port)
        return EINVAL;

    FOREACH_ELT (udp_port_data, i, val)
        if (s_port == val)
            return 0;
    if (!ADD(udp_port_data, s_port, tmp))
        return ENOMEM;
    return 0;
}

static krb5_error_code add_tcp_port(int port)
{
    int i;
    void *tmp;
    u_short val;
    u_short s_port = port;

    if (s_port != port)
        return EINVAL;

    FOREACH_ELT (tcp_port_data, i, val)
        if (s_port == val)
            return 0;
    if (!ADD(tcp_port_data, s_port, tmp))
        return ENOMEM;
    return 0;
}


#define USE_AF AF_INET
#define USE_TYPE SOCK_DGRAM
#define USE_PROTO 0
#define SOCKET_ERRNO errno
#include "foreachaddr.h"

struct socksetup {
    const char *prog;
    krb5_error_code retval;
};

static struct connection *
add_fd (struct socksetup *data, int sock, enum kdc_conn_type conntype,
        void (*service)(struct connection *, const char *, int))
{
    struct connection *newconn;
    void *tmp;

    newconn = malloc(sizeof(*newconn));
    if (newconn == 0) {
        data->retval = errno;
        com_err(data->prog, errno,
                gettext("cannot allocate storage for connection info"));
        return 0;
    }
    if (!ADD(connections, newconn, tmp)) {
        data->retval = errno;
        com_err(data->prog, data->retval, gettext("cannot save socket info"));
        free(newconn);
        return 0;
    }

    memset(newconn, 0, sizeof(*newconn));
    newconn->type = conntype;
    newconn->fd = sock;
    newconn->service = service;
    return newconn;
}

static void process_packet(struct connection *, const char *, int);
static void accept_tcp_connection(struct connection *, const char *, int);
static void process_tcp_connection(struct connection *, const char *, int);

static struct connection *
add_udp_fd (struct socksetup *data, int sock)
{
    return add_fd(data, sock, CONN_UDP, process_packet);
}

static struct connection *
add_tcp_listener_fd (struct socksetup *data, int sock)
{
    return add_fd(data, sock, CONN_TCP_LISTENER, accept_tcp_connection);
}

static struct connection *
add_tcp_data_fd (struct socksetup *data, int sock)
{
    return add_fd(data, sock, CONN_TCP, process_tcp_connection);
}

static void
delete_fd (struct connection *xconn)
{
    struct connection *conn;
    int i;

    FOREACH_ELT(connections, i, conn)
        if (conn == xconn) {
            DEL(connections, i);
            break;
        }
    free(xconn);
}

static int
setnbio(int sock)
{
    static const int one = 1;
    return ioctlsocket(sock, FIONBIO, (const void *)&one);
}

static int
setnolinger(int s)
{
    static const struct linger ling = { 0, 0 };
    return setsockopt(s, SOL_SOCKET, SO_LINGER, &ling, sizeof(ling));
}

/* Returns -1 or socket fd.  */
static int
setup_a_tcp_listener(struct socksetup *data, struct sockaddr *addr)
{
    int sock;

    sock = socket(addr->sa_family, SOCK_STREAM, 0);
    if (sock == -1) {
        com_err(data->prog, errno,
                gettext("Cannot create TCP server socket on %s"),
                paddr(addr));
        return -1;
    }
    /*
     * Solaris Kerberos: noticed that there where bind problems for tcp sockets
     * if kdc restarted quickly.  Setting SO_REUSEADDR allowed binds to succeed.
     */
    if (setreuseaddr(sock, 1) < 0) {
        com_err(data->prog, errno,
                gettext("enabling SO_REUSEADDR on TCP socket"));
        close(sock);
        return -1;
    }
    if (bind(sock, addr, socklen(addr)) == -1) {
        com_err(data->prog, errno,
                gettext("Cannot bind TCP server socket on %s"), paddr(addr));
        close(sock);
        return -1;
    }
    if (listen(sock, 5) < 0) {
        com_err(data->prog, errno,
                gettext("Cannot listen on TCP server socket on %s"),
                paddr(addr));
        close(sock);
        return -1;
    }
    if (setnbio(sock)) {
        com_err(data->prog, errno,
                gettext("cannot set listening tcp socket on %s non-blocking"),
                paddr(addr));
        close(sock);
        return -1;
    }
    if (setnolinger(sock)) {
        com_err(data->prog, errno,
                gettext("disabling SO_LINGER on TCP socket on %s"),
                paddr(addr));
        close(sock);
        return -1;
    }
    return sock;
}

static int
setup_tcp_listener_ports(struct socksetup *data)
{
    struct sockaddr_in sin4;
#ifdef KRB5_USE_INET6
    struct sockaddr_in6 sin6;
#endif
    int i, port;

    memset(&sin4, 0, sizeof(sin4));
    sin4.sin_family = AF_INET;
#ifdef HAVE_SA_LEN
    sin4.sin_len = sizeof(sin4);
#endif
    sin4.sin_addr.s_addr = INADDR_ANY;

#ifdef KRB5_USE_INET6
    memset(&sin6, 0, sizeof(sin6));
    sin6.sin6_family = AF_INET6;
#ifdef SIN6_LEN
    sin6.sin6_len = sizeof(sin6);
#endif
    sin6.sin6_addr = in6addr_any;
#endif

    FOREACH_ELT (tcp_port_data, i, port) {
        int s4, s6;

        set_sa_port((struct sockaddr *)&sin4, htons(port));
        if (!ipv6_enabled()) {
            s4 = setup_a_tcp_listener(data, (struct sockaddr *)&sin4);
            if (s4 < 0)
                return -1;
            s6 = -1;
        } else {
#ifndef KRB5_USE_INET6
            abort();
#else
            s4 = s6 = -1;

            set_sa_port((struct sockaddr *)&sin6, htons(port));

            s6 = setup_a_tcp_listener(data, (struct sockaddr *)&sin6);
            if (s6 < 0)
                return -1;
#ifdef IPV6_V6ONLY
            if (setv6only(s6, 0))
                com_err(data->prog, errno,
                        gettext("setsockopt(IPV6_V6ONLY,0) failed"));
#endif

            s4 = setup_a_tcp_listener(data, (struct sockaddr *)&sin4);
#endif /* KRB5_USE_INET6 */
        }

        /* Sockets are created, prepare to listen on them.  */
        if (s4 >= 0) {
            FD_SET(s4, &sstate.rfds);
            if (s4 >= sstate.max)
                sstate.max = s4 + 1;
            if (add_tcp_listener_fd(data, s4) == 0)
                close(s4);
            else
                krb5_klog_syslog(LOG_INFO, "listening on fd %d: tcp %s",
                                 s4, paddr((struct sockaddr *)&sin4));
        }
#ifdef KRB5_USE_INET6
        if (s6 >= 0) {
            FD_SET(s6, &sstate.rfds);
            if (s6 >= sstate.max)
                sstate.max = s6 + 1;
            if (add_tcp_listener_fd(data, s6) == 0) {
                close(s6);
                s6 = -1;
            } else
                krb5_klog_syslog(LOG_INFO, "listening on fd %d: tcp %s",
                                 s6, paddr((struct sockaddr *)&sin6));
            if (s4 < 0)
                krb5_klog_syslog(LOG_INFO,
                                 "assuming IPv6 socket accepts IPv4");
        }
#endif
    }
    return 0;
}

static int
setup_udp_port(void *P_data, struct sockaddr *addr)
{
    struct socksetup *data = P_data;
    int sock = -1, i;
    char haddrbuf[NI_MAXHOST];
    int err;
    u_short port;

    err = getnameinfo(addr, socklen(addr), haddrbuf, sizeof(haddrbuf),
                      0, 0, NI_NUMERICHOST);
    if (err)
        strcpy(haddrbuf, "<unprintable>");

    switch (addr->sa_family) {
    case AF_INET:
        break;
#ifdef AF_INET6
    case AF_INET6:
#ifdef KRB5_USE_INET6
        break;
#else
        {
            static int first = 1;
            if (first) {
                krb5_klog_syslog (LOG_INFO, "skipping local ipv6 addresses");
                first = 0;
            }
            return 0;
        }
#endif
#endif
#ifdef AF_LINK /* some BSD systems, AIX */
    case AF_LINK:
        return 0;
#endif
#ifdef AF_DLI /* Direct Link Interface - DEC Ultrix/OSF1 link layer? */
    case AF_DLI:
        return 0;
#endif
    default:
        krb5_klog_syslog (LOG_INFO,
                          "skipping unrecognized local address family %d",
                          addr->sa_family);
        return 0;
    }

    FOREACH_ELT (udp_port_data, i, port) {
        sock = socket (addr->sa_family, SOCK_DGRAM, 0);
        if (sock == -1) {
            data->retval = errno;
            com_err(data->prog, data->retval,
                    gettext("Cannot create server socket for port %d address %s"),
                    port, haddrbuf);
            return 1;
        }
        set_sa_port(addr, htons(port));
        if (bind (sock, (struct sockaddr *)addr, socklen (addr)) == -1) {
            data->retval = errno;
            com_err(data->prog, data->retval,
                    gettext("Cannot bind server socket to port %d address %s"),
                    port, haddrbuf);
            return 1;
        }
        FD_SET (sock, &sstate.rfds);
        if (sock >= sstate.max)
            sstate.max = sock + 1;
        krb5_klog_syslog (LOG_INFO, "listening on fd %d: udp %s", sock,
                          paddr((struct sockaddr *)addr));
        if (add_udp_fd (data, sock) == 0)
            return 1;
    }
    return 0;
}

#if 1
static void klog_handler(const void *data, size_t len)
{
    static char buf[BUFSIZ];
    static int bufoffset;
    void *p;

#define flush_buf() \
  (bufoffset                                            \
   ? (((buf[0] == 0 || buf[0] == '\n')                  \
       ? (fork()==0?abort():(void)0)                    \
       : (void)0),                                      \
      krb5_klog_syslog(LOG_INFO, "%s", buf),            \
      memset(buf, 0, sizeof(buf)),                      \
      bufoffset = 0)                                    \
   : 0)

    p = memchr(data, 0, len);
    if (p)
        len = (const char *)p - (const char *)data;
scan_for_newlines:
    if (len == 0)
        return;
    p = memchr(data, '\n', len);
    if (p) {
        if (p != data)
            klog_handler(data, (size_t)((const char *)p - (const char *)data));
        flush_buf();
        len -= ((const char *)p - (const char *)data) + 1;
        data = 1 + (const char *)p;
        goto scan_for_newlines;
    } else if (len > sizeof(buf) - 1 || len + bufoffset > sizeof(buf) - 1) {
        size_t x = sizeof(buf) - len - 1;
        klog_handler(data, x);
        flush_buf();
        len -= x;
        data = (const char *)data + x;
        goto scan_for_newlines;
    } else {
        memcpy(buf + bufoffset, data, len);
        bufoffset += len;
    }
}
#endif

/* XXX */
extern int krb5int_debug_sendto_kdc;
extern void (*krb5int_sendtokdc_debug_handler)(const void*, size_t);

krb5_error_code
setup_network(const char *prog)
{
    struct socksetup setup_data;
    krb5_error_code retval;
    char *cp;
    int i, port;

    FD_ZERO(&sstate.rfds);
    FD_ZERO(&sstate.wfds);
    FD_ZERO(&sstate.xfds);
    sstate.max = 0;

/*    krb5int_debug_sendto_kdc = 1; */
    krb5int_sendtokdc_debug_handler = klog_handler;

    /* Handle each realm's ports */
    for (i=0; i<kdc_numrealms; i++) {
        cp = kdc_realmlist[i]->realm_ports;
        while (cp && *cp) {
            if (*cp == ',' || isspace((int) *cp)) {
                cp++;
                continue;
            }
            port = strtol(cp, &cp, 10);
            if (cp == 0)
                break;
            retval = add_udp_port(port);
            if (retval)
                return retval;
        }

        cp = kdc_realmlist[i]->realm_tcp_ports;
        while (cp && *cp) {
            if (*cp == ',' || isspace((int) *cp)) {
                cp++;
                continue;
            }
            port = strtol(cp, &cp, 10);
            if (cp == 0)
                break;
            retval = add_tcp_port(port);
            if (retval)
                return retval;
        }
    }

    setup_data.prog = prog;
    setup_data.retval = 0;
    krb5_klog_syslog (LOG_INFO, "setting up network...");
    /* To do: Use RFC 2292 interface (or follow-on) and IPV6_PKTINFO,
       so we might need only one UDP socket; fall back to binding
       sockets on each address only if IPV6_PKTINFO isn't
       supported.  */
    if (foreach_localaddr (&setup_data, setup_udp_port, 0, 0)) {
        return setup_data.retval;
    }
    setup_tcp_listener_ports(&setup_data);
    krb5_klog_syslog (LOG_INFO, "set up %d sockets", n_sockets);
    if (n_sockets == 0) {
        com_err(prog, 0, gettext("no sockets set up?"));
        exit (1);
    }

    return 0;
}

static void init_addr(krb5_fulladdr *faddr, struct sockaddr *sa)
{
    switch (sa->sa_family) {
    case AF_INET:
        faddr->address->addrtype = ADDRTYPE_INET;
        faddr->address->length = IPV4_ADDR_LEN;
        faddr->address->contents = (krb5_octet *) &sa2sin(sa)->sin_addr;
        faddr->port = ntohs(sa2sin(sa)->sin_port);
        break;
#ifdef KRB5_USE_INET6
    case AF_INET6:
        if (IN6_IS_ADDR_V4MAPPED(&sa2sin6(sa)->sin6_addr)) {
            faddr->address->addrtype = ADDRTYPE_INET;
            faddr->address->length = IPV4_ADDR_LEN;
            /* offset to RAM address of ipv4 part of ipv6 address */
            faddr->address->contents = (IPV6_ADDR_LEN - IPV4_ADDR_LEN) +
                (krb5_octet *) &sa2sin6(sa)->sin6_addr;
        } else {
            faddr->address->addrtype = ADDRTYPE_INET6;
            faddr->address->length = IPV6_ADDR_LEN;
            faddr->address->contents = (krb5_octet *) &sa2sin6(sa)->sin6_addr;
        }
        faddr->port = ntohs(sa2sin6(sa)->sin6_port);
        break;
#endif
    default:
        faddr->address->addrtype = -1;
        faddr->address->length = 0;
        faddr->address->contents = 0;
        faddr->port = 0;
        break;
    }
}

static void process_packet(struct connection *conn, const char *prog,
                           int selflags)
{
    int cc;
    socklen_t saddr_len;
    krb5_fulladdr faddr;
    krb5_error_code retval;
    struct sockaddr_storage saddr;
    krb5_address addr;
    krb5_data request;
    krb5_data *response;
    char pktbuf[MAX_DGRAM_SIZE];
    int port_fd = conn->fd;

    response = NULL;
    saddr_len = sizeof(saddr);
    cc = recvfrom(port_fd, pktbuf, sizeof(pktbuf), 0,
                  (struct sockaddr *)&saddr, &saddr_len);
    if (cc == -1) {
        if (errno != EINTR
            /* This is how Linux indicates that a previous
               transmission was refused, e.g., if the client timed out
               before getting the response packet.  */
            && errno != ECONNREFUSED
            )
            com_err(prog, errno, gettext("while receiving from network"));
        return;
    }
    if (!cc)
        return;         /* zero-length packet? */

    request.length = cc;
    request.data = pktbuf;
    faddr.address = &addr;
    init_addr(&faddr, ss2sa(&saddr));
    /* this address is in net order */
    if ((retval = dispatch(&request, &faddr, &response))) {
        com_err(prog, retval, gettext("while dispatching (udp)"));
        return;
    }
    cc = sendto(port_fd, response->data, (socklen_t) response->length, 0,
                (struct sockaddr *)&saddr, saddr_len);
    if (cc == -1) {
        char addrbuf[46];
        krb5_free_data(kdc_context, response);
        if (inet_ntop(((struct sockaddr *)&saddr)->sa_family,
                      addr.contents, addrbuf, sizeof(addrbuf)) == 0) {
            strcpy(addrbuf, "?");
        }
        com_err(prog, errno, gettext("while sending reply to %s/%d"),
                addrbuf, faddr.port);
        return;
    }
    if (cc != response->length) {
        krb5_free_data(kdc_context, response);
        com_err(prog, 0, gettext("short reply write %d vs %d\n"),
                response->length, cc);
        return;
    }
    krb5_free_data(kdc_context, response);
    return;
}

static int tcp_data_counter;
/* Solaris kerberos: getting this value from elsewhere */
extern int max_tcp_data_connections;

static void kill_tcp_connection(struct connection *);

static void accept_tcp_connection(struct connection *conn, const char *prog,
                                  int selflags)
{
    int s;
    struct sockaddr_storage addr_s;
    struct sockaddr *addr = (struct sockaddr *)&addr_s;
    socklen_t addrlen = sizeof(addr_s);
    struct socksetup sockdata;
    struct connection *newconn;
    char tmpbuf[10];

    s = accept(conn->fd, addr, &addrlen);
    if (s < 0)
        return;
    setnbio(s), setnolinger(s);

    sockdata.prog = prog;
    sockdata.retval = 0;

    newconn = add_tcp_data_fd(&sockdata, s);
    if (newconn == 0)
        return;

    if (getnameinfo((struct sockaddr *)&addr_s, addrlen,
                    newconn->u.tcp.addrbuf, sizeof(newconn->u.tcp.addrbuf),
                    tmpbuf, sizeof(tmpbuf),
                    NI_NUMERICHOST | NI_NUMERICSERV))
        strcpy(newconn->u.tcp.addrbuf, "???");
    else {
        char *p, *end;
        p = newconn->u.tcp.addrbuf;
        end = p + sizeof(newconn->u.tcp.addrbuf);
        p += strlen(p);
        if (end - p > 2 + strlen(tmpbuf)) {
            *p++ = '.';
            strcpy(p, tmpbuf);
        }
    }
#if 0
    krb5_klog_syslog(LOG_INFO, "accepted TCP connection on socket %d from %s",
                     s, newconn->u.tcp.addrbuf);
#endif

    newconn->u.tcp.addr_s = addr_s;
    newconn->u.tcp.addrlen = addrlen;
    newconn->u.tcp.bufsiz = 1024 * 1024;
    newconn->u.tcp.buffer = malloc(newconn->u.tcp.bufsiz);
    newconn->u.tcp.start_time = time(0);

    if (++tcp_data_counter > max_tcp_data_connections) {
        struct connection *oldest_tcp = NULL;
        struct connection *c;
        int i;

        krb5_klog_syslog(LOG_INFO, "too many connections");

        FOREACH_ELT (connections, i, c) {
            if (c->type != CONN_TCP)
                continue;
            if (c == newconn)
                continue;
#if 0
            krb5_klog_syslog(LOG_INFO, "fd %d started at %ld", c->fd,
                             c->u.tcp.start_time);
#endif
            if (oldest_tcp == NULL
                || oldest_tcp->u.tcp.start_time > c->u.tcp.start_time)
                oldest_tcp = c;
        }
        if (oldest_tcp != NULL) {
            krb5_klog_syslog(LOG_INFO, "dropping tcp fd %d from %s",
                             oldest_tcp->fd, oldest_tcp->u.tcp.addrbuf);
            kill_tcp_connection(oldest_tcp);
            oldest_tcp = NULL;
        }
    }
    if (newconn->u.tcp.buffer == 0) {
        com_err(prog, errno, gettext("allocating buffer for new TCP session from %s"),
                newconn->u.tcp.addrbuf);
        delete_fd(newconn);
        close(s);
        tcp_data_counter--;
        return;
    }
    newconn->u.tcp.offset = 0;
    newconn->u.tcp.faddr.address = &newconn->u.tcp.kaddr;
    init_addr(&newconn->u.tcp.faddr, ss2sa(&newconn->u.tcp.addr_s));
    SG_SET(&newconn->u.tcp.sgbuf[0], newconn->u.tcp.lenbuf, 4);
    SG_SET(&newconn->u.tcp.sgbuf[1], 0, 0);

    FD_SET(s, &sstate.rfds);
    if (sstate.max <= s)
        sstate.max = s + 1;
}

static void
kill_tcp_connection(struct connection *conn)
{
    if (conn->u.tcp.response)
        krb5_free_data(kdc_context, conn->u.tcp.response);
    if (conn->u.tcp.buffer)
        free(conn->u.tcp.buffer);
    FD_CLR(conn->fd, &sstate.rfds);
    FD_CLR(conn->fd, &sstate.wfds);
    if (sstate.max == conn->fd + 1)
        while (sstate.max > 0
               && ! FD_ISSET(sstate.max-1, &sstate.rfds)
               && ! FD_ISSET(sstate.max-1, &sstate.wfds)
               /* && ! FD_ISSET(sstate.max-1, &sstate.xfds) */
            )
            sstate.max--;
    close(conn->fd);
    conn->fd = -1;
    delete_fd(conn);
    tcp_data_counter--;
}

static krb5_error_code
make_toolong_error (krb5_data **out)
{
    krb5_error errpkt;
    krb5_error_code retval;
    krb5_data *scratch;

    retval = krb5_us_timeofday(kdc_context, &errpkt.stime, &errpkt.susec);
    if (retval)
        return retval;
    errpkt.error = KRB_ERR_FIELD_TOOLONG;
    errpkt.server = tgs_server;
    errpkt.client = NULL;
    errpkt.cusec = 0;
    errpkt.ctime = 0;
    errpkt.text.length = 0;
    errpkt.text.data = 0;
    errpkt.e_data.length = 0;
    errpkt.e_data.data = 0;
    scratch = malloc(sizeof(*scratch));
    if (scratch == NULL)
        return ENOMEM;
    retval = krb5_mk_error(kdc_context, &errpkt, scratch);
    if (retval) {
        free(scratch);
        return retval;
    }

    *out = scratch;
    return 0;
}

static void
process_tcp_connection(struct connection *conn, const char *prog, int selflags)
{
    if (selflags & SSF_WRITE) {
        ssize_t nwrote;
        SOCKET_WRITEV_TEMP tmp;

        nwrote = SOCKET_WRITEV(conn->fd, conn->u.tcp.sgp, conn->u.tcp.sgnum,
                               tmp);
        if (nwrote < 0) {
            goto kill_tcp_connection;
        }
        if (nwrote == 0)
            /* eof */
            goto kill_tcp_connection;
        while (nwrote) {
            sg_buf *sgp = conn->u.tcp.sgp;
            if (nwrote < SG_LEN(sgp)) {
                SG_ADVANCE(sgp, nwrote);
                nwrote = 0;
            } else {
                nwrote -= SG_LEN(sgp);
                conn->u.tcp.sgp++;
                conn->u.tcp.sgnum--;
                if (conn->u.tcp.sgnum == 0 && nwrote != 0)
                    abort();
            }
        }
        if (conn->u.tcp.sgnum == 0) {
            /* finished sending */
            /* We should go back to reading, though if we sent a
               FIELD_TOOLONG error in reply to a length with the high
               bit set, RFC 4120 says we have to close the TCP
               stream.  */
            goto kill_tcp_connection;
        }
    } else if (selflags & SSF_READ) {
        /* Read message length and data into one big buffer, already
           allocated at connect time.  If we have a complete message,
           we stop reading, so we should only be here if there is no
           data in the buffer, or only an incomplete message.  */
        size_t len;
        ssize_t nread;
        if (conn->u.tcp.offset < 4) {
            /* msglen has not been computed */
            /* XXX Doing at least two reads here, letting the kernel
               worry about buffering.  It'll be faster when we add
               code to manage the buffer here.  */
            len = 4 - conn->u.tcp.offset;
            nread = SOCKET_READ(conn->fd,
                                conn->u.tcp.buffer + conn->u.tcp.offset, len);
            if (nread < 0)
                /* error */
                goto kill_tcp_connection;
            if (nread == 0)
                /* eof */
                goto kill_tcp_connection;
            conn->u.tcp.offset += nread;
            if (conn->u.tcp.offset == 4) {
                unsigned char *p = (unsigned char *)conn->u.tcp.buffer;
                conn->u.tcp.msglen = ((p[0] << 24)
                                      | (p[1] << 16)
                                      | (p[2] <<  8)
                                      | p[3]);
                if (conn->u.tcp.msglen > conn->u.tcp.bufsiz - 4) {
                    krb5_error_code err;
                    /* message too big */
                    krb5_klog_syslog(LOG_ERR, "TCP client %s wants %lu bytes, cap is %lu",
                                     conn->u.tcp.addrbuf, (unsigned long) conn->u.tcp.msglen,
                                     (unsigned long) conn->u.tcp.bufsiz - 4);
                    /* XXX Should return an error.  */
                    err = make_toolong_error (&conn->u.tcp.response);
                    if (err) {
                        krb5_klog_syslog(LOG_ERR,
                                         "error constructing KRB_ERR_FIELD_TOOLONG error! %s",
                                         error_message(err));
                        goto kill_tcp_connection;
                    }
                    goto have_response;
                }
            }
        } else {
            /* msglen known */
            krb5_data request;
            krb5_error_code err;

            len = conn->u.tcp.msglen - (conn->u.tcp.offset - 4);
            nread = SOCKET_READ(conn->fd,
                                conn->u.tcp.buffer + conn->u.tcp.offset, len);
            if (nread < 0)
                /* error */
                goto kill_tcp_connection;
            if (nread == 0)
                /* eof */
                goto kill_tcp_connection;
            conn->u.tcp.offset += nread;
            if (conn->u.tcp.offset < conn->u.tcp.msglen + 4)
                return;
            /* have a complete message, and exactly one message */
            request.length = conn->u.tcp.msglen;
            request.data = conn->u.tcp.buffer + 4;
            err = dispatch(&request, &conn->u.tcp.faddr,
                           &conn->u.tcp.response);
            if (err) {
                com_err(prog, err, gettext("while dispatching (tcp)"));
                goto kill_tcp_connection;
            }
        have_response:
            conn->u.tcp.lenbuf[0] = 0xff & (conn->u.tcp.response->length >> 24);
            conn->u.tcp.lenbuf[1] = 0xff & (conn->u.tcp.response->length >> 16);
            conn->u.tcp.lenbuf[2] = 0xff & (conn->u.tcp.response->length >> 8);
            conn->u.tcp.lenbuf[3] = 0xff & (conn->u.tcp.response->length >> 0);
            SG_SET(&conn->u.tcp.sgbuf[1], conn->u.tcp.response->data,
                   conn->u.tcp.response->length);
            conn->u.tcp.sgp = conn->u.tcp.sgbuf;
            conn->u.tcp.sgnum = 2;
            FD_CLR(conn->fd, &sstate.rfds);
            FD_SET(conn->fd, &sstate.wfds);
        }
    } else
        abort();

    return;

kill_tcp_connection:
    kill_tcp_connection(conn);
}

static void service_conn(struct connection *conn, const char *prog,
                         int selflags)
{
    conn->service(conn, prog, selflags);
}

krb5_error_code
listen_and_process(const char *prog)
{
    int                 nfound;
    /* This struct contains 3 fd_set objects; on some platforms, they
       can be rather large.  Making this static avoids putting all
       that junk on the stack.  */
    static struct select_state sout;
    int                 i, sret;
    krb5_error_code     err;

    if (conns == (struct connection **) NULL)
        return KDC5_NONET;

    while (!signal_requests_exit) {
        if (signal_requests_hup) {
            krb5_klog_reopen(kdc_context);
            signal_requests_hup = 0;
        }
        sstate.end_time.tv_sec = sstate.end_time.tv_usec = 0;
        err = krb5int_cm_call_select(&sstate, &sout, &sret);
        if (err) {
            com_err(prog, err, gettext("while selecting for network input(1)"));
            continue;
        }
        if (sret == -1) {
            if (errno != EINTR)
                com_err(prog, errno, gettext("while selecting for network input(2)"));
            continue;
        }
        nfound = sret;
        for (i=0; i<n_sockets && nfound > 0; i++) {
            int sflags = 0;
            if (conns[i]->fd < 0)
                abort();
            if (FD_ISSET(conns[i]->fd, &sout.rfds))
                sflags |= SSF_READ, nfound--;
            if (FD_ISSET(conns[i]->fd, &sout.wfds))
                sflags |= SSF_WRITE, nfound--;
            if (sflags)
                service_conn(conns[i], prog, sflags);
        }
    }
    return 0;
}

krb5_error_code
closedown_network(const char *prog)
{
    int i;
    struct connection *conn;

    if (conns == (struct connection **) NULL)
        return KDC5_NONET;

    FOREACH_ELT (connections, i, conn) {
        if (conn->fd >= 0)
            (void) close(conn->fd);
        DEL (connections, i);
        /* There may also be per-connection data in the tcp structure
           (tcp.buffer, tcp.response) that we're not freeing here.
           That should only happen if we quit with a connection in
           progress.  */
        free(conn);
    }
    FREE_SET_DATA(connections);
    FREE_SET_DATA(udp_port_data);
    FREE_SET_DATA(tcp_port_data);

    return 0;
}

#endif /* INET */