xref: /dports/lang/erlang-runtime21/otp-OTP-21.3.8.24/erts/emulator/drivers/common/inet_drv.c

/*
 * %CopyrightBegin%
 *
 * Copyright Ericsson AB 1997-2019. All Rights Reserved.
 *
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at
 *
 *     http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 *
 * %CopyrightEnd%
 */

#ifdef HAVE_CONFIG_H
#include "config.h"
#endif

/* If we HAVE_SCTP_H and Solaris, we need to define the following in
   order to get SCTP working:
*/
#if (defined(HAVE_SCTP_H) && defined(__sun) && defined(__SVR4))
#define  SOLARIS10    1
/* WARNING: This is not quite correct, it may also be Solaris 11! */
#define  _XPG4_2
#define  __EXTENSIONS__
#endif

#include <stdio.h>
#include <stdlib.h>
#include <stddef.h>
#include <ctype.h>
#include <sys/types.h>
#include <errno.h>
#include <stdint.h>

#define IDENTITY(c) c
#define STRINGIFY_1(b) IDENTITY(#b)
#define STRINGIFY(a) STRINGIFY_1(a)

#ifdef HAVE_UNISTD_H
#include <unistd.h>
#endif
#ifdef HAVE_SYS_UIO_H
#include <sys/uio.h>
#endif

#ifdef HAVE_NET_IF_DL_H
#include <net/if_dl.h>
#endif
#ifdef HAVE_IFADDRS_H
#include <ifaddrs.h>
#endif
#ifdef HAVE_NETPACKET_PACKET_H
#include <netpacket/packet.h>
#endif
#ifdef HAVE_SYS_UN_H
#include <sys/un.h>
#endif

#ifdef HAVE_SENDFILE
#if defined(__linux__) || (defined(__sun) && defined(__SVR4))
    #include <sys/sendfile.h>
#elif defined(__FreeBSD__) || defined(__DragonFly__)
    /* Need to define __BSD_VISIBLE in order to expose prototype of sendfile */
    #define __BSD_VISIBLE 1
    #include <sys/socket.h>
#endif
#endif

#if defined(__APPLE__) && defined(__MACH__) && !defined(__DARWIN__)
    #define __DARWIN__ 1
#endif

/* All platforms fail on malloc errors. */
#define FATAL_MALLOC


#include "erl_driver.h"

/* The IS_SOCKET_ERROR macro below is used for portability reasons. While
   POSIX specifies that errors from socket-related system calls should be
   indicated with a -1 return value, some users have experienced non-Windows
   OS kernels that return negative values other than -1. While one can argue
   that such kernels are technically broken, comparing against values less
   than 0 covers their out-of-spec return values without imposing incorrect
   semantics on systems that manage to correctly return -1 for errors, thus
   increasing Erlang's portability.
*/
#ifdef __WIN32__
#define IS_SOCKET_ERROR(val) ((val) == SOCKET_ERROR)
#else
#define IS_SOCKET_ERROR(val) ((val) < 0)
#endif

#ifdef __WIN32__
#define LLU "%I64u"
#else
#define LLU "%llu"
#endif
typedef unsigned long long llu_t;

#ifndef INT16_MIN
#define INT16_MIN (-32768)
#endif
#ifndef INT16_MAX
#define INT16_MAX (32767)
#endif

#ifdef __WIN32__
#define  STRNCASECMP strncasecmp

#define INCL_WINSOCK_API_TYPEDEFS 1

#ifndef WINDOWS_H_INCLUDES_WINSOCK2_H
#include <winsock2.h>
#endif
#include <windows.h>
#include <Ws2tcpip.h>   /* NEED VC 6.0 or higher */

/* Visual studio 2008+: NTDDI_VERSION needs to be set for iphlpapi.h
   to define the right structures. It needs to be set to WINXP (or LONGHORN)
   for IPV6 to work and it's set lower by default, so we need to change it. */
#ifdef HAVE_SDKDDKVER_H
#  include <sdkddkver.h>
#  ifdef NTDDI_VERSION
#    undef NTDDI_VERSION
#  endif
#  define NTDDI_VERSION NTDDI_WINXP
#endif

#include <iphlpapi.h>


#undef WANT_NONBLOCKING
#include "sys.h"

#undef EWOULDBLOCK
#undef ETIMEDOUT

#ifdef EINPROGRESS
#undef EINPROGRESS
#endif
#ifdef EALREADY
#undef EALREADY
#endif
#ifdef ENOTSOCK
#undef ENOTSOCK
#endif
#ifdef EDESTADDRREQ
#undef EDESTADDRREQ
#endif
#ifdef EMSGSIZE
#undef EMSGSIZE
#endif
#ifdef EPROTOTYPE
#undef EPROTOTYPE
#endif
#ifdef ENOPROTOOPT
#undef ENOPROTOOPT
#endif
#ifdef EPROTONOSUPPORT
#undef EPROTONOSUPPORT
#endif
#ifdef EOPNOTSUPP
#undef EOPNOTSUPP
#endif
#ifdef EAFNOSUPPORT
#undef EAFNOSUPPORT
#endif
#ifdef EADDRINUSE
#undef EADDRINUSE
#endif
#ifdef EADDRNOTAVAIL
#undef EADDRNOTAVAIL
#endif
#ifdef ENETDOWN
#undef ENETDOWN
#endif
#ifdef ENETUNREACH
#undef ENETUNREACH
#endif
#ifdef ENETRESET
#undef ENETRESET
#endif
#ifdef ECONNABORTED
#undef ECONNABORTED
#endif
#ifdef ECONNRESET
#undef ECONNRESET
#endif
#ifdef ENOBUFS
#undef ENOBUFS
#endif
#ifdef EISCONN
#undef EISCONN
#endif
#ifdef ENOTCONN
#undef ENOTCONN
#endif
#ifdef ECONNREFUSED
#undef ECONNREFUSED
#endif
#ifdef ELOOP
#undef ELOOP
#endif
#ifdef EHOSTUNREACH
#undef EHOSTUNREACH
#endif


#define HAVE_MULTICAST_SUPPORT
#define HAVE_UDP

#define ERRNO_BLOCK             WSAEWOULDBLOCK

#define EWOULDBLOCK             WSAEWOULDBLOCK
#define EINPROGRESS             WSAEINPROGRESS
#define EALREADY                WSAEALREADY
#define ENOTSOCK                WSAENOTSOCK
#define EDESTADDRREQ            WSAEDESTADDRREQ
#define EMSGSIZE                WSAEMSGSIZE
#define EPROTOTYPE              WSAEPROTOTYPE
#define ENOPROTOOPT             WSAENOPROTOOPT
#define EPROTONOSUPPORT         WSAEPROTONOSUPPORT
#define ESOCKTNOSUPPORT         WSAESOCKTNOSUPPORT
#define EOPNOTSUPP              WSAEOPNOTSUPP
#define EPFNOSUPPORT            WSAEPFNOSUPPORT
#define EAFNOSUPPORT            WSAEAFNOSUPPORT
#define EADDRINUSE              WSAEADDRINUSE
#define EADDRNOTAVAIL           WSAEADDRNOTAVAIL
#define ENETDOWN                WSAENETDOWN
#define ENETUNREACH             WSAENETUNREACH
#define ENETRESET               WSAENETRESET
#define ECONNABORTED            WSAECONNABORTED
#define ECONNRESET              WSAECONNRESET
#define ENOBUFS                 WSAENOBUFS
#define EISCONN                 WSAEISCONN
#define ENOTCONN                WSAENOTCONN
#define ESHUTDOWN               WSAESHUTDOWN
#define ETOOMANYREFS            WSAETOOMANYREFS
#define ETIMEDOUT               WSAETIMEDOUT
#define ECONNREFUSED            WSAECONNREFUSED
#define ELOOP                   WSAELOOP
#undef ENAMETOOLONG
#define ENAMETOOLONG            WSAENAMETOOLONG
#define EHOSTDOWN               WSAEHOSTDOWN
#define EHOSTUNREACH            WSAEHOSTUNREACH
#undef ENOTEMPTY
#define ENOTEMPTY               WSAENOTEMPTY
#define EPROCLIM                WSAEPROCLIM
#define EUSERS                  WSAEUSERS
#define EDQUOT                  WSAEDQUOT
#define ESTALE                  WSAESTALE
#define EREMOTE                 WSAEREMOTE

#define INVALID_EVENT           WSA_INVALID_EVENT

static BOOL (WINAPI *fpSetHandleInformation)(HANDLE,DWORD,DWORD);

#define sock_open(af, type, proto) \
    make_noninheritable_handle(socket((af), (type), (proto)))
#define sock_close(s)              closesocket((s))
#define sock_shutdown(s, how)      shutdown((s), (how))

#define sock_accept(s, addr, len) \
    make_noninheritable_handle(accept((s), (addr), (len)))
#define sock_connect(s, addr, len) connect((s), (addr), (len))
#define sock_listen(s, b)          listen((s), (b))
#define sock_bind(s, addr, len)    bind((s), (addr), (len))
#define sock_getopt(s,t,n,v,l)     getsockopt((s),(t),(n),(v),(l))
#define sock_setopt(s,t,n,v,l)     setsockopt((s),(t),(n),(v),(l))
#define sock_name(s, addr, len)    getsockname((s), (addr), (len))
#define sock_peer(s, addr, len)    getpeername((s), (addr), (len))
#define sock_ntohs(x)              ntohs((x))
#define sock_ntohl(x)              ntohl((x))
#define sock_htons(x)              htons((x))
#define sock_htonl(x)              htonl((x))
#define sock_send(s,buf,len,flag)  send((s),(buf),(len),(flag))
#define sock_sendv(s, vec, size, np, flag) \
            WSASend((s),(WSABUF*)(vec),(size),(np),(flag),NULL,NULL)
#define sock_recv(s,buf,len,flag)  recv((s),(buf),(len),(flag))

#define sock_recvfrom(s,buf,blen,flag,addr,alen) \
	    recvfrom((s),(buf),(blen),(flag),(addr),(alen))
#define sock_sendto(s,buf,blen,flag,addr,alen) \
	    sendto((s),(buf),(blen),(flag),(addr),(alen))
#define sock_hostname(buf, len)    gethostname((buf), (len))

#define sock_getservbyname(name,proto) getservbyname((name),(proto))
#define sock_getservbyport(port,proto) getservbyport((port),(proto))

#define sock_errno() WSAGetLastError()
#define sock_create_event(d)       WSACreateEvent()
#define sock_close_event(e)        WSACloseEvent(e)

#define sock_select(D, Flags, OnOff) winsock_event_select(D, Flags, OnOff)

#define SET_BLOCKING(s)           ioctlsocket(s, FIONBIO, &zero_value)
#define SET_NONBLOCKING(s)        ioctlsocket(s, FIONBIO, &one_value)


static unsigned long zero_value = 0;
static unsigned long one_value = 1;

#define TCP_SHUT_WR    SD_SEND
#define TCP_SHUT_RD    SD_RECEIVE
#define TCP_SHUT_RDWR  SD_BOTH

#else /* !__WIN32__ */

#include <sys/time.h>
#ifdef NETDB_H_NEEDS_IN_H
#include <netinet/in.h>
#endif
#include <netdb.h>

#include <sys/socket.h>
#include <netinet/in.h>

#ifdef DEF_INADDR_LOOPBACK_IN_RPC_TYPES_H
#include <rpc/types.h>
#endif

#include <netinet/tcp.h>
#include <arpa/inet.h>

#include <sys/param.h>
#ifdef HAVE_ARPA_NAMESER_H
#include <arpa/nameser.h>
#endif

#ifdef HAVE_SYS_SOCKIO_H
#include <sys/sockio.h>
#endif

#ifdef HAVE_SYS_IOCTL_H
#include <sys/ioctl.h>
#endif

#include <net/if.h>

#ifdef HAVE_SCHED_H
#include <sched.h>
#endif

#ifdef HAVE_SETNS_H
#include <setns.h>
#endif

#define HAVE_UDP

/* SCTP support -- currently for UNIX platforms only: */
#undef HAVE_SCTP
#if defined(HAVE_SCTP_H)

#include <netinet/sctp.h>

/* SCTP Socket API Draft from version 11 on specifies that netinet/sctp.h must
   explicitly define HAVE_SCTP in case when SCTP is supported,  but Solaris 10
   still apparently uses Draft 10, and does not define that symbol, so we have
   to define it explicitly:
*/
#ifndef     HAVE_SCTP
#    define HAVE_SCTP
#endif

/* These changed in draft 11, so SOLARIS10 uses the old MSG_* */
#if ! HAVE_DECL_SCTP_UNORDERED
#     define    SCTP_UNORDERED  MSG_UNORDERED
#endif
#if ! HAVE_DECL_SCTP_ADDR_OVER
#     define    SCTP_ADDR_OVER  MSG_ADDR_OVER
#endif
#if ! HAVE_DECL_SCTP_ABORT
#     define    SCTP_ABORT      MSG_ABORT
#endif
#if ! HAVE_DECL_SCTP_EOF
#     define    SCTP_EOF        MSG_EOF
#endif

/* More Solaris 10 fixes: */
#if ! HAVE_DECL_SCTP_CLOSED && HAVE_DECL_SCTPS_IDLE
#    define SCTP_CLOSED SCTPS_IDLE
#    undef HAVE_DECL_SCTP_CLOSED
#    define HAVE_DECL_SCTP_CLOSED 1
#endif
#if ! HAVE_DECL_SCTP_BOUND && HAVE_DECL_SCTPS_BOUND
#    define SCTP_BOUND SCTPS_BOUND
#    undef HAVE_DECL_SCTP_BOUND
#    define HAVE_DECL_SCTP_BOUND 1
#endif
#if ! HAVE_DECL_SCTP_LISTEN && HAVE_DECL_SCTPS_LISTEN
#    define SCTP_LISTEN SCTPS_LISTEN
#    undef HAVE_DECL_SCTP_LISTEN
#    define HAVE_DECL_SCTP_LISTEN 1
#endif
#if ! HAVE_DECL_SCTP_COOKIE_WAIT && HAVE_DECL_SCTPS_COOKIE_WAIT
#    define SCTP_COOKIE_WAIT SCTPS_COOKIE_WAIT
#    undef HAVE_DECL_SCTP_COOKIE_WAIT
#    define HAVE_DECL_SCTP_COOKIE_WAIT 1
#endif
#if ! HAVE_DECL_SCTP_COOKIE_ECHOED && HAVE_DECL_SCTPS_COOKIE_ECHOED
#    define SCTP_COOKIE_ECHOED SCTPS_COOKIE_ECHOED
#    undef HAVE_DECL_SCTP_COOKIE_ECHOED
#    define HAVE_DECL_SCTP_COOKIE_ECHOED 1
#endif
#if ! HAVE_DECL_SCTP_ESTABLISHED && HAVE_DECL_SCTPS_ESTABLISHED
#    define SCTP_ESTABLISHED SCTPS_ESTABLISHED
#    undef HAVE_DECL_SCTP_ESTABLISHED
#    define HAVE_DECL_SCTP_ESTABLISHED 1
#endif
#if ! HAVE_DECL_SCTP_SHUTDOWN_PENDING && HAVE_DECL_SCTPS_SHUTDOWN_PENDING
#    define SCTP_SHUTDOWN_PENDING SCTPS_SHUTDOWN_PENDING
#    undef HAVE_DECL_SCTP_SHUTDOWN_PENDING
#    define HAVE_DECL_SCTP_SHUTDOWN_PENDING 1
#endif
#if ! HAVE_DECL_SCTP_SHUTDOWN_SENT && HAVE_DECL_SCTPS_SHUTDOWN_SENT
#    define SCTP_SHUTDOWN_SENT SCTPS_SHUTDOWN_SENT
#    undef HAVE_DECL_SCTP_SHUTDOWN_SENT
#    define HAVE_DECL_SCTP_SHUTDOWN_SENT 1
#endif
#if ! HAVE_DECL_SCTP_SHUTDOWN_RECEIVED && HAVE_DECL_SCTPS_SHUTDOWN_RECEIVED
#    define SCTP_SHUTDOWN_RECEIVED SCTPS_SHUTDOWN_RECEIVED
#    undef HAVE_DECL_SCTP_SHUTDOWN_RECEIVED
#    define HAVE_DECL_SCTP_SHUTDOWN_RECEIVED 1
#endif
#if ! HAVE_DECL_SCTP_SHUTDOWN_ACK_SENT && HAVE_DECL_SCTPS_SHUTDOWN_ACK_SENT
#    define SCTP_SHUTDOWN_ACK_SENT SCTPS_SHUTDOWN_ACK_SENT
#    undef HAVE_DECL_SCTP_SHUTDOWN_ACK_SENT
#    define HAVE_DECL_SCTP_SHUTDOWN_ACK_SENT 1
#endif
/* New spelling in lksctp 2.6.22 or maybe even earlier:
 *  adaption -> adaptation
 */
#if !defined(SCTP_ADAPTATION_LAYER) && defined (SCTP_ADAPTION_LAYER)
#     define SCTP_ADAPTATION_LAYER       SCTP_ADAPTION_LAYER
#     define SCTP_ADAPTATION_INDICATION  SCTP_ADAPTION_INDICATION
#     define sctp_adaptation_event       sctp_adaption_event
#     define sctp_setadaptation          sctp_setadaption
#     define sn_adaptation_event         sn_adaption_event
#     define sai_adaptation_ind          sai_adaption_ind
#     define ssb_adaptation_ind          ssb_adaption_ind
#     define sctp_adaptation_layer_event sctp_adaption_layer_event
#endif

#if defined(__GNUC__) && defined(HAVE_SCTP_BINDX)
static typeof(sctp_bindx) *p_sctp_bindx = NULL;
#else
static int (*p_sctp_bindx)
	(int sd, struct sockaddr *addrs, int addrcnt, int flags) = NULL;
#endif

#if defined(__GNUC__) && defined(HAVE_SCTP_PEELOFF)
static typeof(sctp_peeloff) *p_sctp_peeloff = NULL;
#else
static int (*p_sctp_peeloff)
        (int sd, sctp_assoc_t assoc_id) = NULL;
#endif

#if defined(__GNUC__) && defined(HAVE_SCTP_GETLADDRS)
static typeof(sctp_getladdrs) *p_sctp_getladdrs = NULL;
#else
static int (*p_sctp_getladdrs)
        (int sd, sctp_assoc_t assoc_id, struct sockaddr **ss) = NULL;
#endif

#if defined(__GNUC__) && defined(HAVE_SCTP_FREELADDRS)
static typeof(sctp_freeladdrs) *p_sctp_freeladdrs = NULL;
#else
static void (*p_sctp_freeladdrs)(struct sockaddr *addrs) = NULL;
#endif

#if defined(__GNUC__) && defined(HAVE_SCTP_GETPADDRS)
static typeof(sctp_getpaddrs) *p_sctp_getpaddrs = NULL;
#else
static int (*p_sctp_getpaddrs)
        (int sd, sctp_assoc_t assoc_id, struct sockaddr **ss) = NULL;
#endif

#if defined(__GNUC__) && defined(HAVE_SCTP_FREEPADDRS)
static typeof(sctp_freepaddrs) *p_sctp_freepaddrs = NULL;
#else
static void (*p_sctp_freepaddrs)(struct sockaddr *addrs) = NULL;
#endif

#endif /* #if defined(HAVE_SCTP_H) */

#ifndef WANT_NONBLOCKING
#define WANT_NONBLOCKING
#endif
#include "sys.h"

/* #define INET_DRV_DEBUG 1 */
#ifdef INET_DRV_DEBUG
#define DEBUG 1
#undef DEBUGF
#define DEBUGF(X) printf X
#endif

#if !defined(HAVE_STRNCASECMP)
#define STRNCASECMP my_strncasecmp

static int my_strncasecmp(const char *s1, const char *s2, size_t n)
{
    int i;

    for (i=0;i<n-1 && s1[i] && s2[i] && toupper(s1[i]) == toupper(s2[i]);++i)
	;
    return (toupper(s1[i]) - toupper(s2[i]));
}


#else
#define  STRNCASECMP strncasecmp
#endif

#define INVALID_SOCKET -1
#define INVALID_EVENT  -1
#define SOCKET_ERROR   -1

#define SOCKET int
#define HANDLE long int
#define FD_READ    ERL_DRV_READ
#define FD_WRITE   ERL_DRV_WRITE
#define FD_CLOSE   0
#define FD_CONNECT ERL_DRV_WRITE
#define FD_ACCEPT  ERL_DRV_READ

#define sock_connect(s, addr, len)  connect((s), (addr), (len))
#define sock_listen(s, b)           listen((s), (b))
#define sock_bind(s, addr, len)     bind((s), (addr), (len))
#define sock_getopt(s,t,n,v,l)      getsockopt((s),(t),(n),(v),(l))
#define sock_setopt(s,t,n,v,l)      setsockopt((s),(t),(n),(v),(l))
#define sock_name(s, addr, len)     getsockname((s), (addr), (len))
#define sock_peer(s, addr, len)     getpeername((s), (addr), (len))
#define sock_ntohs(x)               ntohs((x))
#define sock_ntohl(x)               ntohl((x))
#define sock_htons(x)               htons((x))
#define sock_htonl(x)               htonl((x))

#define sock_accept(s, addr, len)   accept((s), (addr), (len))
#define sock_send(s,buf,len,flag)   send((s),(buf),(len),(flag))
#define sock_sendto(s,buf,blen,flag,addr,alen) \
                sendto((s),(buf),(blen),(flag),(addr),(alen))
#define sock_sendv(s, vec, size, np, flag) \
		(*(np) = writev((s), (struct iovec*)(vec), (size)))
#define sock_sendmsg(s,msghdr,flag) sendmsg((s),(msghdr),(flag))

#define sock_open(af, type, proto)  socket((af), (type), (proto))
#define sock_close(s)               close((s))
#define sock_shutdown(s, how)       shutdown((s), (how))

#define sock_hostname(buf, len)     gethostname((buf), (len))
#define sock_getservbyname(name,proto) getservbyname((name), (proto))
#define sock_getservbyport(port,proto) getservbyport((port), (proto))

#define sock_recv(s,buf,len,flag)   recv((s),(buf),(len),(flag))
#define sock_recvfrom(s,buf,blen,flag,addr,alen) \
                recvfrom((s),(buf),(blen),(flag),(addr),(alen))
#define sock_recvmsg(s,msghdr,flag) recvmsg((s),(msghdr),(flag))

#define sock_errno()                errno
#define sock_create_event(d)        ((d)->s) /* return file descriptor */
#define sock_close_event(e)                  /* do nothing */

#define inet_driver_select(port, e, mode, on) \
                                    driver_select(port, e, mode | (on?ERL_DRV_USE:0), on)

#define sock_select(d, flags, onoff) do { \
        ASSERT(!INET_IGNORED(d));         \
        (d)->event_mask = (onoff) ? \
                 ((d)->event_mask | (flags)) : \
                 ((d)->event_mask & ~(flags)); \
        DEBUGF(("(%s / %d) sock_select(%ld): flags=%02X, onoff=%d, event_mask=%02lX\r\n", \
		__FILE__, __LINE__, (long) (d)->port, (flags), (onoff), (unsigned long) (d)->event_mask)); \
        inet_driver_select((d)->port, (ErlDrvEvent)(long)(d)->event, (flags), (onoff)); \
   } while(0)

#define TCP_SHUT_WR    SHUT_WR
#define TCP_SHUT_RD    SHUT_RD
#define TCP_SHUT_RDWR  SHUT_RDWR

#endif /* !__WIN32__ */

#ifdef HAVE_SOCKLEN_T
#  define SOCKLEN_T socklen_t
#else
#  define SOCKLEN_T size_t
#endif

#include "packet_parser.h"

#define get_int24(s) ((((unsigned char*) (s))[0] << 16) | \
                      (((unsigned char*) (s))[1] << 8)  | \
                      (((unsigned char*) (s))[2]))

#define get_little_int32(s) ((((unsigned char*) (s))[3] << 24) | \
			     (((unsigned char*) (s))[2] << 16)  | \
			     (((unsigned char*) (s))[1] << 8) | \
			     (((unsigned char*) (s))[0]))

#if defined(HAVE_SYS_UN_H) || defined(SO_BINDTODEVICE)

/* strnlen doesn't exist everywhere */
static size_t my_strnlen(const char *s, size_t maxlen)
{
    size_t i = 0;
    while (i < maxlen && s[i] != '\0')
        i++;
    return i;
}

#endif

#ifdef VALGRIND
#  include <valgrind/memcheck.h>
#else
#  define VALGRIND_MAKE_MEM_DEFINED(ptr,size)
#endif

#ifndef __WIN32__
/* Calculate CMSG_NXTHDR without having a struct msghdr*.
 * CMSG_LEN only caters for alignment for start of data.
 * To get how much to advance we need to use CMSG_SPACE
 * on the payload length.  To get the payload length we
 * take the calculated cmsg->cmsg_len and subtract the
 * header length.  To get the header length we use
 * the pointer difference from the cmsg start pointer
 * to the CMSG_DATA(cmsg) pointer.
 *
 * Some platforms (seen on ppc Linux 2.6.29-3.ydl61.3)
 * may return 0 as the cmsg_len if the cmsg is to be ignored.
 */
#define LEN_CMSG_DATA(cmsg)                                             \
    ((cmsg)->cmsg_len < sizeof (struct cmsghdr) ? 0 :                   \
     (cmsg)->cmsg_len - ((char*)CMSG_DATA(cmsg) - (char*)(cmsg)))
#define NXT_CMSG_HDR(cmsg)                                              \
    ((struct cmsghdr*)(((char*)(cmsg)) + CMSG_SPACE(LEN_CMSG_DATA(cmsg))))
#endif

#if !defined(IPV6_PKTOPTIONS) && defined(IPV6_2292PKTOPTIONS)
#define IPV6_PKTOPTIONS IPV6_2292PKTOPTIONS
#endif

/*
  Magic errno value used locally for return of {error, system_limit}
  - the emulator definition of SYSTEM_LIMIT is not available here.
*/
#define INET_ERRNO_SYSTEM_LIMIT  (15 << 8)

/*----------------------------------------------------------------------------
** Interface constants.
**
** This section must be "identical" to the corresponding inet_int.hrl
*/

/* general address encode/decode tag */
#define INET_AF_UNSPEC      0
#define INET_AF_INET        1
#define INET_AF_INET6       2
#define INET_AF_ANY         3 /* INADDR_ANY or IN6ADDR_ANY_INIT */
#define INET_AF_LOOPBACK    4 /* INADDR_LOOPBACK or IN6ADDR_LOOPBACK_INIT */
#define INET_AF_LOCAL       5
#define INET_AF_UNDEFINED   6 /* Unknown */

/* open and INET_REQ_GETTYPE enumeration */
#define INET_TYPE_STREAM    1
#define INET_TYPE_DGRAM     2
#define INET_TYPE_SEQPACKET 3

/* INET_LOPT_MODE options */
#define INET_MODE_LIST      0
#define INET_MODE_BINARY    1

/* INET_LOPT_DELIVER options */
#define INET_DELIVER_PORT   0
#define INET_DELIVER_TERM   1

/* INET_LOPT_ACTIVE options */
#define INET_PASSIVE        0  /* false */
#define INET_ACTIVE         1  /* true */
#define INET_ONCE           2  /* true; active once then passive */
#define INET_MULTI          3  /* true; active N then passive */

/* INET_REQ_GETSTATUS enumeration */
#define INET_F_OPEN         0x0001
/* INET_F_BOUND removed - renumber when there comes a bigger rewrite */
#define INET_F_ACTIVE       0x0004
#define INET_F_LISTEN       0x0008
#define INET_F_CON          0x0010
#define INET_F_ACC          0x0020
#define INET_F_LST          0x0040
#define INET_F_BUSY         0x0080
#define INET_F_MULTI_CLIENT 0x0100 /* Multiple clients for one descriptor, i.e. multi-accept */

/* One numberspace for *_REQ_* so if an e.g UDP request is issued
** for a TCP socket, the driver can protest.
*/
#define INET_REQ_OPEN          1
#define INET_REQ_CLOSE         2
#define INET_REQ_CONNECT       3
#define INET_REQ_PEER          4
#define INET_REQ_NAME          5
#define INET_REQ_BIND          6
#define INET_REQ_SETOPTS       7
#define INET_REQ_GETOPTS       8
/* #define INET_REQ_GETIX         9  NOT USED ANY MORE */
/* #define INET_REQ_GETIF         10 REPLACE BY NEW STUFF */
#define INET_REQ_GETSTAT       11
#define INET_REQ_GETHOSTNAME   12
#define INET_REQ_FDOPEN        13
#define INET_REQ_GETFD         14
#define INET_REQ_GETTYPE       15
#define INET_REQ_GETSTATUS     16
#define INET_REQ_GETSERVBYNAME 17
#define INET_REQ_GETSERVBYPORT 18
#define INET_REQ_SETNAME       19
#define INET_REQ_SETPEER       20
#define INET_REQ_GETIFLIST     21
#define INET_REQ_IFGET         22
#define INET_REQ_IFSET         23
#define INET_REQ_SUBSCRIBE     24
#define INET_REQ_GETIFADDRS    25
#define INET_REQ_ACCEPT        26
#define INET_REQ_LISTEN        27
#define INET_REQ_IGNOREFD      28
#define INET_REQ_GETLADDRS     29
#define INET_REQ_GETPADDRS     30

/* TCP requests */
/* #define TCP_REQ_ACCEPT         40 MOVED */
/* #define TCP_REQ_LISTEN         41 MERGED */
#define TCP_REQ_RECV           42
#define TCP_REQ_UNRECV         43
#define TCP_REQ_SHUTDOWN       44
#define TCP_REQ_SENDFILE       45
/* UDP and SCTP requests */
#define PACKET_REQ_RECV        60 /* Common for UDP and SCTP         */
/* #define SCTP_REQ_LISTEN       61 MERGED Different from TCP; not for UDP */
#define SCTP_REQ_BINDX	       62 /* Multi-home SCTP bind            */
#define SCTP_REQ_PEELOFF       63

/* INET_REQ_SUBSCRIBE sub-requests */
#define INET_SUBS_EMPTY_OUT_Q  1

/* TCP additional flags */
#define TCP_ADDF_DELAY_SEND    1
#define TCP_ADDF_CLOSE_SENT    2 /* Close sent (active mode only) */
#define TCP_ADDF_DELAYED_CLOSE_RECV 4 /* If receive fails, report {error,closed} (passive mode) */
#define TCP_ADDF_DELAYED_CLOSE_SEND 8 /* If send fails, report {error,closed} (passive mode) */
#define TCP_ADDF_PENDING_SHUT_WR   16 /* Call shutdown(sock, SHUT_WR) when queue empties */
#define TCP_ADDF_PENDING_SHUT_RDWR 32 /* Call shutdown(sock, SHUT_RDWR) when queue empties */
#define TCP_ADDF_PENDING_SHUTDOWN \
		(TCP_ADDF_PENDING_SHUT_WR | TCP_ADDF_PENDING_SHUT_RDWR)
#define TCP_ADDF_SHOW_ECONNRESET   64 /* Tell user about incoming RST */
#define TCP_ADDF_DELAYED_ECONNRESET 128 /* An ECONNRESET error occurred on send or shutdown */
#define TCP_ADDF_SHUTDOWN_WR_DONE 256 /* A shutdown(sock, SHUT_WR) or SHUT_RDWR was made */
#define TCP_ADDF_LINGER_ZERO 	  512 /* Discard driver queue on port close */
#define TCP_ADDF_SENDFILE         1024 /* Send from an fd instead of the driver queue */

/* *_REQ_* replies */
#define INET_REP_ERROR       0
#define INET_REP_OK          1
#define INET_REP             2

/* INET_REQ_SETOPTS and INET_REQ_GETOPTS options */
#define INET_OPT_REUSEADDR  0   /* enable/disable local address reuse */
#define INET_OPT_KEEPALIVE  1   /* enable/disable keep connections alive */
#define INET_OPT_DONTROUTE  2   /* enable/disable routing for messages */
#define INET_OPT_LINGER     3   /* linger on close if data is present */
#define INET_OPT_BROADCAST  4   /* enable/disable transmission of broadcast */
#define INET_OPT_OOBINLINE  5   /* enable/disable out-of-band data in band */
#define INET_OPT_SNDBUF     6   /* set send buffer size */
#define INET_OPT_RCVBUF     7   /* set receive buffer size */
#define INET_OPT_PRIORITY   8   /* set priority */
#define INET_OPT_TOS        9   /* Set type of service */
#define TCP_OPT_NODELAY     10  /* don't delay send to coalesce packets */
#define UDP_OPT_MULTICAST_IF 11  /* set/get IP multicast interface */
#define UDP_OPT_MULTICAST_TTL 12 /* set/get IP multicast timetolive */
#define UDP_OPT_MULTICAST_LOOP 13 /* set/get IP multicast loopback */
#define UDP_OPT_ADD_MEMBERSHIP 14 /* add an IP group membership */
#define UDP_OPT_DROP_MEMBERSHIP 15 /* drop an IP group membership */
#define INET_OPT_IPV6_V6ONLY 16 /* IPv6 only socket, no mapped v4 addrs */
/* LOPT is local options */
#define INET_LOPT_BUFFER      20  /* min buffer size hint */
#define INET_LOPT_HEADER      21  /* list header size */
#define INET_LOPT_ACTIVE      22  /* enable/disable active receive */
#define INET_LOPT_PACKET      23  /* packet header type (TCP) */
#define INET_LOPT_MODE        24  /* list or binary mode */
#define INET_LOPT_DELIVER     25  /* port or term delivery */
#define INET_LOPT_EXITONCLOSE 26  /* exit port on active close or not ! */
#define INET_LOPT_TCP_HIWTRMRK     27  /* set local high watermark */
#define INET_LOPT_TCP_LOWTRMRK     28  /* set local low watermark */
                                /* 29  unused */
#define INET_LOPT_TCP_SEND_TIMEOUT 30  /* set send timeout */
#define INET_LOPT_TCP_DELAY_SEND   31  /* Delay sends until next poll */
#define INET_LOPT_PACKET_SIZE      32  /* Max packet size */
#define INET_LOPT_UDP_READ_PACKETS 33  /* Number of packets to read */
#define INET_OPT_RAW               34  /* Raw socket options */
#define INET_LOPT_TCP_SEND_TIMEOUT_CLOSE 35  /* auto-close on send timeout or not */
#define INET_LOPT_MSGQ_HIWTRMRK     36  /* set local msgq high watermark */
#define INET_LOPT_MSGQ_LOWTRMRK     37  /* set local msgq low watermark */
#define INET_LOPT_NETNS             38  /* Network namespace pathname */
#define INET_LOPT_TCP_SHOW_ECONNRESET 39  /* tell user about incoming RST */
#define INET_LOPT_LINE_DELIM        40  /* Line delimiting char */
#define INET_OPT_TCLASS             41  /* IPv6 transport class */
#define INET_OPT_BIND_TO_DEVICE     42  /* get/set network device the socket is bound to */
#define INET_OPT_RECVTOS            43  /* IP_RECVTOS ancillary data */
#define INET_OPT_RECVTCLASS         44  /* IPV6_RECVTCLASS ancillary data */
#define INET_OPT_PKTOPTIONS         45  /* IP(V6)_PKTOPTIONS get ancillary data */
#define INET_OPT_TTL                46  /* IP_TTL */
#define INET_OPT_RECVTTL            47  /* IP_RECVTTL ancillary data */
#define TCP_OPT_NOPUSH              48  /* super-Nagle, aka TCP_CORK */
/* SCTP options: a separate range, from 100: */
#define SCTP_OPT_RTOINFO		100
#define SCTP_OPT_ASSOCINFO		101
#define SCTP_OPT_INITMSG		102
#define SCTP_OPT_AUTOCLOSE		103
#define SCTP_OPT_NODELAY		104
#define SCTP_OPT_DISABLE_FRAGMENTS	105
#define SCTP_OPT_I_WANT_MAPPED_V4_ADDR	106
#define SCTP_OPT_MAXSEG			107
#define SCTP_OPT_SET_PEER_PRIMARY_ADDR  108
#define SCTP_OPT_PRIMARY_ADDR		109
#define SCTP_OPT_ADAPTATION_LAYER 	110
#define SCTP_OPT_PEER_ADDR_PARAMS	111
#define SCTP_OPT_DEFAULT_SEND_PARAM	112
#define SCTP_OPT_EVENTS			113
#define SCTP_OPT_DELAYED_ACK_TIME	114
#define SCTP_OPT_STATUS			115
#define SCTP_OPT_GET_PEER_ADDR_INFO	116

/* INET_REQ_IFGET and INET_REQ_IFSET options */
#define INET_IFOPT_ADDR       1
#define INET_IFOPT_BROADADDR  2
#define INET_IFOPT_DSTADDR    3
#define INET_IFOPT_MTU        4
#define INET_IFOPT_NETMASK    5
#define INET_IFOPT_FLAGS      6
#define INET_IFOPT_HWADDR     7

/* INET_REQ_GETSTAT enumeration */
#define INET_STAT_RECV_CNT   1
#define INET_STAT_RECV_MAX   2
#define INET_STAT_RECV_AVG   3
#define INET_STAT_RECV_DVI   4
#define INET_STAT_SEND_CNT   5
#define INET_STAT_SEND_MAX   6
#define INET_STAT_SEND_AVG   7
#define INET_STAT_SEND_PND   8
#define INET_STAT_RECV_OCT   9      /* received octets */
#define INET_STAT_SEND_OCT   10     /* sent octets */

/* INET_IFOPT_FLAGS enumeration */
#define INET_IFF_UP            0x0001
#define INET_IFF_BROADCAST     0x0002
#define INET_IFF_LOOPBACK      0x0004
#define INET_IFF_POINTTOPOINT  0x0008
#define INET_IFF_RUNNING       0x0010
#define INET_IFF_MULTICAST     0x0020
/* Complement flags for turning them off */
#define INET_IFF_DOWN            0x0100
#define INET_IFF_NBROADCAST      0x0200
/* #define INET_IFF_NLOOPBACK    0x0400 */
#define INET_IFF_NPOINTTOPOINT   0x0800
/* #define INET_IFF_NRUNNING     0x1000 */
/* #define INET_IFF_NMULTICAST   0x2000 */

/* Flags for "sctp_sndrcvinfo". Used in a bitmask -- must be powers of 2:
** INET_REQ_SETOPTS:SCTP_OPT_DEFAULT_SEND_PARAM
*/
#define SCTP_FLAG_UNORDERED (1 /* am_unordered */)
#define SCTP_FLAG_ADDR_OVER (2 /* am_addr_over */)
#define SCTP_FLAG_ABORT     (4 /* am_abort */)
#define SCTP_FLAG_EOF       (8 /* am_eof */)
#define SCTP_FLAG_SNDALL   (16 /* am_sndall, NOT YET IMPLEMENTED */)

/* Flags for "sctp_set_opts" (actually for SCTP_OPT_PEER_ADDR_PARAMS).
** These flags are also used in a bitmask, so they must be powers of 2:
*/
#define SCTP_FLAG_HB_ENABLE	    (1 /* am_hb_enable */)
#define SCTP_FLAG_HB_DISABLE	    (2 /* am_hb_disable */)
#define SCTP_FLAG_HB_DEMAND	    (4 /* am_hb_demand */)
#define	SCTP_FLAG_PMTUD_ENABLE	    (8 /* am_pmtud_enable */)
#define	SCTP_FLAG_PMTUD_DISABLE    (16 /* am_pmtud_disable */)
#define SCTP_FLAG_SACDELAY_ENABLE  (32 /* am_sackdelay_enable */)
#define SCTP_FLAG_SACDELAY_DISABLE (64 /* am_sackdelay_disable */)

/* Flags for recv_cmsgflags */
#define INET_CMSG_RECVTOS        (1 << 0) /* am_recvtos, am_tos */
#define INET_CMSG_RECVTCLASS     (1 << 1) /* am_recvtclass, am_tclass */
#define INET_CMSG_RECVTTL        (1 << 2) /* am_recvttl, am_ttl */

/* Inet flags */
#define INET_FLG_BUFFER_SET      (1 << 0) /* am_buffer has been set by user */
#define INET_FLG_IS_IGNORED      (1 << 1) /* If a fd is ignored by the inet_drv.
                                             This flag should be set to true when
                                             the fd is used outside of inet_drv. */
#define INET_FLG_IS_IGNORED_RD   (1 << 2)
#define INET_FLG_IS_IGNORED_WR   (1 << 3)
#define INET_FLG_IS_IGNORED_PASS (1 << 4)

/*
** End of interface constants.
**--------------------------------------------------------------------------*/

#define INET_STATE_CLOSED          (0)
#define INET_STATE_OPEN            (INET_F_OPEN)
#define INET_STATE_CONNECTED       (INET_STATE_OPEN | INET_F_ACTIVE)
#define INET_STATE_LISTENING       (INET_STATE_OPEN | INET_F_LISTEN)
#define INET_STATE_CONNECTING      (INET_STATE_OPEN | INET_F_CON)
#define INET_STATE_ACCEPTING       (INET_STATE_LISTENING | INET_F_ACC)
#define INET_STATE_MULTI_ACCEPTING (INET_STATE_ACCEPTING | INET_F_MULTI_CLIENT)

#define IS_OPEN(d) \
 (((d)->state & INET_F_OPEN) == INET_F_OPEN)

#define IS_CONNECTED(d) \
  (((d)->state & INET_STATE_CONNECTED) == INET_STATE_CONNECTED)

#define IS_CONNECTING(d) \
  (((d)->state & INET_F_CON) == INET_F_CON)

#define IS_BUSY(d) \
  (((d)->state & INET_F_BUSY) == INET_F_BUSY)

#define INET_MAX_OPT_BUFFER (64*1024)

#define INET_DEF_BUFFER     1460        /* default buffer size */
#define INET_MIN_BUFFER     1           /* internal min buffer */

#define INET_HIGH_WATERMARK (1024*8) /* 8k pending high => busy  */
#define INET_LOW_WATERMARK  (1024*4) /* 4k pending => allow more */
#define INET_HIGH_MSGQ_WATERMARK (1024*8) /* 8k pending high => busy  */
#define INET_LOW_MSGQ_WATERMARK  (1024*4) /* 4k pending => allow more */

#define INET_INFINITY  0xffffffff  /* infinity value */

#define INET_MAX_ASYNC 1           /* max number of async queue ops */

/* INET_LOPT_UDP_PACKETS */
#define INET_PACKET_POLL     5   /* maximum number of packets to poll */

/* Max interface name */
#define INET_IFNAMSIZ          16

/* INET Ignore states */
#define INET_IGNORE_CLEAR(desc) ((desc)->flags & ~(INET_IGNORE_READ|INET_IGNORE_WRITE|INET_IGNORE_PASSIVE))
#define INET_IGNORED(desc)   ((desc)->flags & INET_FLG_IS_IGNORED)
#define INET_IGNORE_READ    (INET_FLG_IS_IGNORED|INET_FLG_IS_IGNORED_RD)
#define INET_IGNORE_WRITE   (INET_FLG_IS_IGNORED|INET_FLG_IS_IGNORED_WR)
#define INET_IGNORE_PASSIVE (INET_FLG_IS_IGNORED|INET_FLG_IS_IGNORED_PASS)

/* Max length of Erlang Term Buffer (for outputting structured terms):  */
#ifdef  HAVE_SCTP
#define PACKET_ERL_DRV_TERM_DATA_LEN  512
#else
#ifndef __WIN32__
/* Assume we have recvmsg() and might need room for ancillary data */
#define PACKET_ERL_DRV_TERM_DATA_LEN  64
#else
#define PACKET_ERL_DRV_TERM_DATA_LEN  32
#endif
#endif

typedef struct _tcp_descriptor tcp_descriptor;

#if defined(TCP_CORK)
#define INET_TCP_NOPUSH TCP_CORK
#elif defined(TCP_NOPUSH) && !defined(__DARWIN__)
#define INET_TCP_NOPUSH TCP_NOPUSH
#endif

#define BIN_REALLOC_MARGIN(x)  ((x)/4)  /* 25% */

/* The general purpose sockaddr */
typedef union {
    struct sockaddr sa;
    struct sockaddr_in sai;
#ifdef HAVE_IN6
    struct sockaddr_in6 sai6;
#endif
#ifdef HAVE_SYS_UN_H
    struct sockaddr_un sal;
#endif
} inet_address;


#define inet_address_port(x)			\
  ((((x)->sai.sin_family == AF_INET) ||		\
    ((x)->sai.sin_family == AF_INET6)) ?	\
   ((x)->sai.sin_port) : -1)

#ifdef HAVE_SYS_UN_H
#define localaddrlen(data)				\
  ((((unsigned char*)(data))[0] == INET_AF_LOCAL) ?	\
   (1 + 1 + ((unsigned char*)(data))[1]) : 1)
#else
#define localaddrlen(data) (1)
#endif

#if defined(HAVE_IN6) && defined(AF_INET6)
#define addrlen(data)					\
    ((((unsigned char*)(data))[0] == INET_AF_INET) ?	\
     (1 + 2 + 4) :					\
     ((((unsigned char*)(data))[0] == INET_AF_INET6) ?	\
      (1 + 2 + 16) : localaddrlen(data)))
#else
#define addrlen(data)					\
    ((((unsigned char*)(data))[0] == INET_AF_INET) ?	\
     (1 + 2 + 4) : localaddrlen(data))
#endif

typedef struct _multi_timer_data {
    ErlDrvTime when;
    ErlDrvTermData caller;
    void (*timeout_function)(ErlDrvData drv_data, ErlDrvTermData caller);
    struct _multi_timer_data *next;
    struct _multi_timer_data *prev;
} MultiTimerData;

static MultiTimerData *add_multi_timer(tcp_descriptor *desc, ErlDrvPort port,
                                       ErlDrvTermData caller, unsigned timeout,
                                       void (*timeout_fun)(ErlDrvData drv_data,
                                                           ErlDrvTermData caller));
static void fire_multi_timers(tcp_descriptor *desc, ErlDrvPort port,
			      ErlDrvData data);
static void remove_multi_timer(tcp_descriptor *desc, ErlDrvPort port, MultiTimerData *p);
static void cancel_multi_timer(tcp_descriptor *desc, ErlDrvPort port,
                               void (*timeout_fun)(ErlDrvData drv_data,
                                                   ErlDrvTermData caller));

static void tcp_inet_multi_timeout(ErlDrvData e, ErlDrvTermData caller);
static void clean_multi_timers(tcp_descriptor *desc, ErlDrvPort port);

typedef struct {
    int            id;      /* id used to identify reply */
    ErlDrvTermData caller;  /* recipient of async reply */
    int            req;     /* Request id (CONNECT/ACCEPT/RECV) */
    union {
	unsigned       value; /* Request timeout (since op issued,not started) */
	MultiTimerData *mtd;
    } tmo;
    ErlDrvMonitor monitor;
} inet_async_op;

typedef struct inet_async_multi_op_ {
    inet_async_op op;
    struct inet_async_multi_op_ *next;
} inet_async_multi_op;


typedef struct subs_list_ {
  ErlDrvTermData subscriber;
  struct subs_list_ *next;
} subs_list;

#define NO_PROCESS 0
#define NO_SUBSCRIBERS(SLP) ((SLP)->subscriber == NO_PROCESS)
static void send_to_subscribers(ErlDrvTermData, subs_list *, int,
				ErlDrvTermData [], int);
static void free_subscribers(subs_list*);
static int save_subscriber(subs_list *, ErlDrvTermData);

typedef struct {
    SOCKET s;                   /* the socket or INVALID_SOCKET if not open */
    HANDLE event;               /* Event handle (same as s in unix) */
    long  event_mask;           /* current FD events */
#ifdef __WIN32__
    long forced_events;           /* Mask of events that are forcefully signalled
				   on windows see winsock_event_select
				   for details */
    int send_would_block;       /* Last send attempt failed with "WOULDBLOCK" */
#endif
    ErlDrvPort  port;           /* the port identifier */
    ErlDrvTermData dport;       /* the port identifier as DriverTermData */
    int   state;                /* status */
    int   prebound;             /* only set when opened with inet_fdopen */
    int   mode;                 /* BINARY | LIST
				   (affect how to interpret hsz) */
    int   exitf;                /* exit port on close or not */
    int   deliver;              /* Delivery mode, TERM or PORT */

    ErlDrvTermData caller;      /* recipient of sync reply */
    ErlDrvTermData busy_caller; /* recipient of sync reply when caller busy.
				 * Only valid while INET_F_BUSY. */

    inet_async_op* oph;          /* queue head or NULL */
    inet_async_op* opt;          /* queue tail or NULL */
    inet_async_op  op_queue[INET_MAX_ASYNC];  /* call queue */
    int op_ref;                 /* queue reference generator  */

    int   active;               /* 0 = passive, 1 = active, 2 = active once */
    Sint16 active_count;        /* counter for {active,N} */
    int   stype;                /* socket type:
				    SOCK_STREAM/SOCK_DGRAM/SOCK_SEQPACKET   */
    int   sprotocol;            /* socket protocol:
				   IPPROTO_TCP|IPPROTO_UDP|IPPROTO_SCTP     */
    int   sfamily;              /* address family */
    enum PacketParseType htype; /* header type (TCP only?) */
    unsigned int psize;         /* max packet size (TCP only?) */
    inet_address remote;        /* remote address for connected sockets */
    inet_address peer_addr;     /* fake peer address */
    inet_address name_addr;     /* fake local address */

    inet_address* peer_ptr;     /* fake peername or NULL */
    inet_address* name_ptr;     /* fake sockname or NULL */
    SOCKLEN_T peer_addr_len;    /* fake peername size */
    SOCKLEN_T name_addr_len;    /* fake sockname size */

    int   bufsz;                /* minimum buffer constraint */
    unsigned int hsz;           /* the list header size, -1 is large !!! */
    /* statistics */
#ifdef ARCH_64
    Uint64        recv_oct;     /* number of received octets, 64 bits */
#else
    Uint32        recv_oct[2];  /* number of received octets, 64 bits */
#endif
    unsigned long recv_cnt;     /* number of packets received */
    unsigned long recv_max;     /* maximum packet size received */
    double recv_avg;            /* average packet size received */
    double recv_dvi;            /* avarage deviation from avg_size */
#ifdef ARCH_64
    Uint64        send_oct;     /* number of octets sent, 64 bits */
#else
    Uint32        send_oct[2];  /* number of octets sent, 64 bits */
#endif
    char          delimiter;    /* Line delimiting character (def: '\n')  */
    unsigned long send_cnt;     /* number of packets sent */
    unsigned long send_max;     /* maximum packet send */
    double send_avg;            /* average packet size sent */

    subs_list empty_out_q_subs; /* Empty out queue subscribers */
    int flags;
#ifdef HAVE_SETNS
    char *netns;                /* Socket network namespace name
				   as full file path */
#endif
    int recv_cmsgflags;         /* Which ancillary data to expect */
} inet_descriptor;



#define TCP_MAX_PACKET_SIZE 0x4000000  /* 64 M */

#define MAX_VSIZE 16		/* Max number of entries allowed in an I/O
				 * vector sock_sendv().
				 */

static int tcp_inet_init(void);
static void tcp_inet_stop(ErlDrvData);
static void tcp_inet_command(ErlDrvData, char*, ErlDrvSizeT);
static void tcp_inet_commandv(ErlDrvData, ErlIOVec*);
static void tcp_inet_flush(ErlDrvData drv_data);
static void tcp_inet_drv_input(ErlDrvData, ErlDrvEvent);
static void tcp_inet_drv_output(ErlDrvData data, ErlDrvEvent event);
static ErlDrvData tcp_inet_start(ErlDrvPort, char* command);
static ErlDrvSSizeT tcp_inet_ctl(ErlDrvData, unsigned int,
				 char*, ErlDrvSizeT, char**, ErlDrvSizeT);
static void tcp_inet_timeout(ErlDrvData);
static void tcp_inet_process_exit(ErlDrvData, ErlDrvMonitor *);
static void inet_stop_select(ErlDrvEvent, void*);
static void inet_emergency_close(ErlDrvData);
#ifdef __WIN32__
static void tcp_inet_event(ErlDrvData, ErlDrvEvent);
static void find_dynamic_functions(void);
#endif

static struct erl_drv_entry tcp_inet_driver_entry =
{
    tcp_inet_init,  /* inet_init will add this driver !! */
    tcp_inet_start,
    tcp_inet_stop,
    tcp_inet_command,
#ifdef __WIN32__
    tcp_inet_event,
    NULL,
#else
    tcp_inet_drv_input,
    tcp_inet_drv_output,
#endif
    "tcp_inet",
    NULL,
    NULL,
    tcp_inet_ctl,
    tcp_inet_timeout,
    tcp_inet_commandv,
    NULL,
    tcp_inet_flush,
    NULL,
    NULL,
    ERL_DRV_EXTENDED_MARKER,
    ERL_DRV_EXTENDED_MAJOR_VERSION,
    ERL_DRV_EXTENDED_MINOR_VERSION,
    ERL_DRV_FLAG_USE_PORT_LOCKING|ERL_DRV_FLAG_SOFT_BUSY,
    NULL,
    tcp_inet_process_exit,
    inet_stop_select,
    inet_emergency_close
};



#ifdef HAVE_UDP
static int        packet_inet_init(void);
static void       packet_inet_stop(ErlDrvData);
static void       packet_inet_command(ErlDrvData, char*, ErlDrvSizeT);
static void       packet_inet_drv_input(ErlDrvData data, ErlDrvEvent event);
static ErlDrvData udp_inet_start(ErlDrvPort, char* command);
#ifdef HAVE_SCTP
static ErlDrvData sctp_inet_start(ErlDrvPort, char* command);
#endif
static ErlDrvSSizeT packet_inet_ctl(ErlDrvData, unsigned int, char*,
				    ErlDrvSizeT, char**, ErlDrvSizeT);
static void       packet_inet_timeout(ErlDrvData);
#ifdef __WIN32__
static void       packet_inet_event(ErlDrvData, ErlDrvEvent);
static SOCKET     make_noninheritable_handle(SOCKET s);
static int        winsock_event_select(inet_descriptor *, int, int);
#endif

static struct erl_drv_entry udp_inet_driver_entry =
{
    packet_inet_init,  /* inet_init will add this driver !! */
    udp_inet_start,
    packet_inet_stop,
    packet_inet_command,
#ifdef __WIN32__
    packet_inet_event,
    NULL,
#else
    packet_inet_drv_input,
    NULL,
#endif
    "udp_inet",
    NULL,
    NULL,
    packet_inet_ctl,
    packet_inet_timeout,
    NULL,
    NULL,
    NULL,
    NULL,
    NULL,
    ERL_DRV_EXTENDED_MARKER,
    ERL_DRV_EXTENDED_MAJOR_VERSION,
    ERL_DRV_EXTENDED_MINOR_VERSION,
    ERL_DRV_FLAG_USE_PORT_LOCKING,
    NULL,
    NULL,
    inet_stop_select,
    inet_emergency_close
};
#endif

#ifdef HAVE_SCTP
static struct erl_drv_entry sctp_inet_driver_entry =
{
    packet_inet_init,  /* inet_init will add this driver !! */
    sctp_inet_start,
    packet_inet_stop,
    packet_inet_command,
#ifdef __WIN32__
    packet_inet_event,
    NULL,
#else
    packet_inet_drv_input,
    NULL,
#endif
    "sctp_inet",
    NULL,
    NULL,
    packet_inet_ctl,
    packet_inet_timeout,
    NULL,
    NULL,
    NULL,
    NULL,
    NULL,
    ERL_DRV_EXTENDED_MARKER,
    ERL_DRV_EXTENDED_MAJOR_VERSION,
    ERL_DRV_EXTENDED_MINOR_VERSION,
    ERL_DRV_FLAG_USE_PORT_LOCKING,
    NULL,
    NULL, /* process_exit */
    inet_stop_select,
    inet_emergency_close
};
#endif

struct _tcp_descriptor {
    inet_descriptor inet;       /* common data structure (DON'T MOVE) */
    int   high;                 /* high watermark */
    int   low;                  /* low watermark */
    int   send_timeout;         /* timeout to use in send */
    int   send_timeout_close;   /* auto-close socket on send_timeout */
    int   busy_on_send;         /* busy on send with timeout! */
    int   i_bufsz;              /* current input buffer size (<= bufsz) */
    ErlDrvBinary* i_buf;        /* current binary buffer */
    char*         i_ptr;        /* current pos in buf */
    char*         i_ptr_start;  /* packet start pos in buf */
    int           i_remain;     /* remaining chars to read */
    int           tcp_add_flags;/* Additional TCP descriptor flags */
    int           http_state;   /* 0 = response|request  1=headers fields */
    inet_async_multi_op *multi_first;/* NULL == no multi-accept-queue, op is in ordinary queue */
    inet_async_multi_op *multi_last;
    MultiTimerData *mtd;       /* Timer structures for multiple accept */
    MultiTimerData *mtd_cache; /* A cache for timer allocations */
#ifdef HAVE_SENDFILE
    struct {
        ErlDrvSizeT ioq_skip;   /* The number of bytes in the queue at the time
                                 * sendfile was issued, which must be sent
                                 * before issuing the sendfile call itself. */
        int dup_file_fd;        /* The file handle to send from; this is
                                 * duplicated when sendfile is issued to
                                 * reduce (but not eliminate) the impact of a
                                 * nasty race, so we have to remember to close
                                 * it. */
        Uint64 bytes_sent;
        Uint64 offset;
        Uint64 length;
    } sendfile;
#endif
};

/* send function */
static int tcp_send(tcp_descriptor* desc, char* ptr, ErlDrvSizeT len);
static int tcp_sendv(tcp_descriptor* desc, ErlIOVec* ev);
static int tcp_recv(tcp_descriptor* desc, int request_len);
static int tcp_deliver(tcp_descriptor* desc, int len);

static int tcp_shutdown_error(tcp_descriptor* desc, int err);

#ifdef HAVE_SENDFILE
static int tcp_inet_sendfile(tcp_descriptor* desc);
static int tcp_sendfile_aborted(tcp_descriptor* desc, int socket_error);
#endif

static int tcp_inet_output(tcp_descriptor* desc, HANDLE event);
static int tcp_inet_input(tcp_descriptor* desc, HANDLE event);

static void tcp_desc_close(tcp_descriptor*);

#ifdef HAVE_UDP
typedef struct {
    inet_descriptor inet;   /* common data structure (DON'T MOVE) */
    int read_packets;       /* Number of packets to read per invocation */
    int i_bufsz;            /* current input buffer size */
    ErlDrvBinary* i_buf;    /* current binary buffer */
    char* i_ptr;            /* current pos in buf */
} udp_descriptor;


static int packet_inet_input(udp_descriptor* udesc, HANDLE event);
#endif

/* convert descriptor pointer to inet_descriptor pointer */
#define INETP(d) (&(d)->inet)

#define NEW_ASYNC_ID(desc) ((desc)->op_ref++ & 0xffff)

/* check for transition from active to passive */
#define INET_CHECK_ACTIVE_TO_PASSIVE(inet)                              \
    do {                                                                \
        if ((inet)->active == INET_ONCE)                                \
            (inet)->active = INET_PASSIVE;                              \
        else if ((inet)->active == INET_MULTI && --((inet)->active_count) == 0) { \
            (inet)->active = INET_PASSIVE;                              \
            packet_passive_message(inet);                               \
        }                                                               \
    } while (0)

static ErlDrvTermData am_ok;
static ErlDrvTermData am_undefined;
static ErlDrvTermData am_unspec;
static ErlDrvTermData am_tcp;
static ErlDrvTermData am_error;
static ErlDrvTermData am_einval;
static ErlDrvTermData am_inet_async;
static ErlDrvTermData am_inet_reply;
static ErlDrvTermData am_timeout;
static ErlDrvTermData am_closed;
static ErlDrvTermData am_tcp_passive;
static ErlDrvTermData am_tcp_closed;
static ErlDrvTermData am_tcp_error;
static ErlDrvTermData am_empty_out_q;
static ErlDrvTermData am_ssl_tls;
#ifdef HAVE_UDP
static ErlDrvTermData am_udp;
static ErlDrvTermData am_udp_passive;
static ErlDrvTermData am_udp_error;
#endif
#ifdef HAVE_SYS_UN_H
static ErlDrvTermData am_local;
#endif
#ifndef __WIN32__
static ErlDrvTermData am_tos;
static ErlDrvTermData am_tclass;
static ErlDrvTermData am_ttl;
#endif
#ifdef HAVE_SCTP
static ErlDrvTermData am_sctp;
static ErlDrvTermData am_sctp_passive;
static ErlDrvTermData am_sctp_error;
static ErlDrvTermData am_true;
static ErlDrvTermData am_false;
static ErlDrvTermData am_buffer;
static ErlDrvTermData am_mode;
static ErlDrvTermData am_list;
static ErlDrvTermData am_binary;
static ErlDrvTermData am_active;
static ErlDrvTermData am_once;
static ErlDrvTermData am_multi;
static ErlDrvTermData am_buffer;
static ErlDrvTermData am_linger;
static ErlDrvTermData am_recbuf;
static ErlDrvTermData am_sndbuf;
static ErlDrvTermData am_reuseaddr;
static ErlDrvTermData am_dontroute;
static ErlDrvTermData am_priority;
static ErlDrvTermData am_recvtos;
static ErlDrvTermData am_recvtclass;
static ErlDrvTermData am_recvttl;
static ErlDrvTermData am_ipv6_v6only;
static ErlDrvTermData am_netns;
static ErlDrvTermData am_bind_to_device;
#endif
#ifdef HAVE_SENDFILE
static ErlDrvTermData am_sendfile;
#endif

static char str_eafnosupport[] = "eafnosupport";
static char str_einval[] = "einval";

/* special errors for bad ports and sequences */
#define EXBADPORT "exbadport"
#define EXBADSEQ  "exbadseq"


static int inet_init(void);
static ErlDrvSSizeT ctl_reply(int, char*, ErlDrvSizeT, char**, ErlDrvSizeT);

struct erl_drv_entry inet_driver_entry =
{
    inet_init,  /* inet_init will add TCP, UDP and SCTP drivers */
    NULL, /* start */
    NULL, /* stop */
    NULL, /* output */
    NULL, /* ready_input */
    NULL, /* ready_output */
    "inet",
    NULL,
    NULL, /* handle */
    NULL, /* control */
    NULL, /* timeout */
    NULL, /* outputv */
    NULL, /* ready_async */
    NULL, /* flush */
    NULL, /* call */
    NULL, /* event */
    ERL_DRV_EXTENDED_MARKER,
    ERL_DRV_EXTENDED_MAJOR_VERSION,
    ERL_DRV_EXTENDED_MINOR_VERSION,
    0,
    NULL,
    NULL,
    NULL,
};

#if HAVE_IN6
#  if ! defined(HAVE_IN6ADDR_ANY) || ! HAVE_IN6ADDR_ANY
#    if HAVE_DECL_IN6ADDR_ANY_INIT
static const struct in6_addr in6addr_any = { { IN6ADDR_ANY_INIT } };
#    else
static const struct in6_addr in6addr_any =
    { { { 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0 } } };
#    endif /* HAVE_IN6ADDR_ANY_INIT */
#  endif /* ! HAVE_DECL_IN6ADDR_ANY */

#  if ! defined(HAVE_IN6ADDR_LOOPBACK) || ! HAVE_IN6ADDR_LOOPBACK
#    if HAVE_DECL_IN6ADDR_LOOPBACK_INIT
static const struct in6_addr in6addr_loopback =
    { { IN6ADDR_LOOPBACK_INIT } };
#    else
static const struct in6_addr in6addr_loopback =
    { { { 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1 } } };
#    endif /* HAVE_IN6ADDR_LOOPBACk_INIT */
#  endif /* ! HAVE_DECL_IN6ADDR_LOOPBACK */
#endif /* HAVE_IN6 */

/* XXX: is this a driver interface function ??? */
void erts_exit(int n, char*, ...);

/*
 * Malloc wrapper,
 * we would like to change the behaviour for different
 * systems here.
 */

#ifdef FATAL_MALLOC

static void *alloc_wrapper(ErlDrvSizeT size){
    void *ret = driver_alloc(size);
    if(ret == NULL)
	erts_exit(ERTS_ERROR_EXIT,"Out of virtual memory in malloc (%s)", __FILE__);
    return ret;
}
#define ALLOC(X) alloc_wrapper(X)

static void *realloc_wrapper(void *current, ErlDrvSizeT size){
    void *ret = driver_realloc(current,size);
    if(ret == NULL)
	erts_exit(ERTS_ERROR_EXIT,"Out of virtual memory in realloc (%s)", __FILE__);
    return ret;
}
#define REALLOC(X,Y) realloc_wrapper(X,Y)
#define FREE(P) driver_free((P))
#else /* FATAL_MALLOC */

#define ALLOC(X) driver_alloc((X))
#define REALLOC(X,Y) driver_realloc((X), (Y))
#define FREE(P) driver_free((P))

#endif /* FATAL_MALLOC */

#define INIT_ATOM(NAME) am_ ## NAME = driver_mk_atom(#NAME)

#define LOAD_ATOM_CNT 2
#define LOAD_ATOM(vec, i, atom) \
  (((vec)[(i)] = ERL_DRV_ATOM), \
  ((vec)[(i)+1] = (atom)), \
  ((i)+LOAD_ATOM_CNT))

#define LOAD_INT_CNT 2
#define LOAD_INT(vec, i, val) \
  (((vec)[(i)] = ERL_DRV_INT), \
  ((vec)[(i)+1] = (ErlDrvTermData)(val)), \
  ((i)+LOAD_INT_CNT))

#define LOAD_UINT_CNT 2
#define LOAD_UINT(vec, i, val) \
  (((vec)[(i)] = ERL_DRV_UINT), \
  ((vec)[(i)+1] = (ErlDrvTermData)(val)), \
  ((i)+LOAD_UINT_CNT))

#define LOAD_PORT_CNT 2
#define LOAD_PORT(vec, i, port) \
  (((vec)[(i)] = ERL_DRV_PORT), \
  ((vec)[(i)+1] = (port)), \
  ((i)+LOAD_PORT_CNT))

#define LOAD_PID_CNT 2
#define LOAD_PID(vec, i, pid) \
  (((vec)[(i)] = ERL_DRV_PID), \
  ((vec)[(i)+1] = (pid)), \
  ((i)+LOAD_PID_CNT))

#define LOAD_BINARY_CNT 4
#define LOAD_BINARY(vec, i, bin, offs, len) \
  (((vec)[(i)] = ERL_DRV_BINARY), \
  ((vec)[(i)+1] = (ErlDrvTermData)(bin)), \
  ((vec)[(i)+2] = (len)), \
  ((vec)[(i)+3] = (offs)), \
  ((i)+LOAD_BINARY_CNT))

#define LOAD_BUF2BINARY_CNT 3
#define LOAD_BUF2BINARY(vec, i, buf, len) \
  (((vec)[(i)] = ERL_DRV_BUF2BINARY), \
  ((vec)[(i)+1] = (ErlDrvTermData)(buf)), \
  ((vec)[(i)+2] = (len)), \
  ((i)+LOAD_BUF2BINARY_CNT))

#define LOAD_STRING_CNT 3
#define LOAD_STRING(vec, i, str, len) \
  (((vec)[(i)] = ERL_DRV_STRING), \
  ((vec)[(i)+1] = (ErlDrvTermData)(str)), \
  ((vec)[(i)+2] = (len)), \
  ((i)+LOAD_STRING_CNT))

#define LOAD_STRING_CONS_CNT 3
#define LOAD_STRING_CONS(vec, i, str, len) \
  (((vec)[(i)] = ERL_DRV_STRING_CONS), \
  ((vec)[(i)+1] = (ErlDrvTermData)(str)), \
  ((vec)[(i)+2] = (len)), \
  ((i)+LOAD_STRING_CONS_CNT))

#define LOAD_TUPLE_CNT 2
#define LOAD_TUPLE(vec, i, size) \
  (((vec)[(i)] = ERL_DRV_TUPLE), \
  ((vec)[(i)+1] = (size)), \
  ((i)+LOAD_TUPLE_CNT))

#define LOAD_NIL_CNT 1
#define LOAD_NIL(vec, i) \
  (((vec)[(i)] = ERL_DRV_NIL), \
  ((i)+LOAD_NIL_CNT))

#define LOAD_LIST_CNT 2
#define LOAD_LIST(vec, i, size) \
  (((vec)[(i)] = ERL_DRV_LIST), \
  ((vec)[(i)+1] = (size)), \
  ((i)+LOAD_LIST_CNT))


#ifdef HAVE_SCTP
    /* "IS_SCTP": tells the difference between a UDP and an SCTP socket: */
#   define IS_SCTP(desc)((desc)->sprotocol==IPPROTO_SCTP)

    /* For AssocID, 4 bytes should be enough -- checked by "init": */
#   define GET_ASSOC_ID		get_int32
#   define ASSOC_ID_LEN		4
#   define LOAD_ASSOC_ID        LOAD_UINT
#   define LOAD_ASSOC_ID_CNT    LOAD_UINT_CNT
#else
#   define IS_SCTP(desc) 0
#endif
#   define ANC_BUFF_SIZE   INET_DEF_BUFFER/2 /* XXX: not very good... */

#ifdef HAVE_UDP
static int load_address(ErlDrvTermData* spec, int i, char* buf)
{
    int n;
    switch (*buf++) { /* Family */
    case INET_AF_INET: {
        for (n = 2;  n < 2+4;  n++) {
	    spec[i++] = ERL_DRV_INT;
	    spec[i++] = (ErlDrvTermData) ((unsigned char)buf[n]);
	}
	spec[i++] = ERL_DRV_TUPLE;
	spec[i++] = 4;
	spec[i++] = ERL_DRV_INT;
	spec[i++] = (ErlDrvTermData) get_int16(buf);
	break;
    }
#if defined(HAVE_IN6) && defined(AF_INET6)
    case INET_AF_INET6: {
	for (n = 2;  n < 2+16;  n += 2) {
	    spec[i++] = ERL_DRV_INT;
	    spec[i++] = (ErlDrvTermData) get_int16(buf+n);
	}
	spec[i++] = ERL_DRV_TUPLE;
	spec[i++] = 8;
	spec[i++] = ERL_DRV_INT;
	spec[i++] = (ErlDrvTermData) get_int16(buf);
	break;
    }
#endif
#ifdef HAVE_SYS_UN_H
    case INET_AF_LOCAL: {
	int len = *(unsigned char*)buf++;
	i = LOAD_ATOM(spec, i, am_local);
	i = LOAD_BUF2BINARY(spec, i, buf, len);
	spec[i++] = ERL_DRV_TUPLE;
	spec[i++] = 2;
	spec[i++] = ERL_DRV_INT;
	spec[i++] = 0;
	break;
    }
#endif
    case INET_AF_UNSPEC: {
        i = LOAD_ATOM(spec, i, am_unspec);
	i = LOAD_BUF2BINARY(spec, i, buf, 0);
	spec[i++] = ERL_DRV_TUPLE;
	spec[i++] = 2;
	spec[i++] = ERL_DRV_INT;
	spec[i++] = 0;
	break;
    }
    default: { /* INET_AF_UNDEFINED */
        i = LOAD_ATOM(spec, i, am_undefined);
	i = LOAD_BUF2BINARY(spec, i, buf, 0);
	spec[i++] = ERL_DRV_TUPLE;
	spec[i++] = 2;
	spec[i++] = ERL_DRV_INT;
	spec[i++] = 0;
	break;
    }
    }
    return i;
 }
#endif


#ifdef HAVE_SCTP
/* For SCTP, we often need to return {IP, Port} tuples: */
static int inet_get_address(char* dst, inet_address* src, unsigned int* len);

/* Max of {{int()*8},int()} | {{int()*4},int()} |
 *        {{'local',binary()},int()}
 */
#define LOAD_INET_GET_ADDRESS_CNT					\
        (8*LOAD_INT_CNT + LOAD_TUPLE_CNT + LOAD_INT_CNT + LOAD_TUPLE_CNT)

static int load_inet_get_address
           (ErlDrvTermData* spec,    int i, inet_descriptor* desc,
	    struct sockaddr_storage* addr)
{
    /* The size of the buffer  used to stringify the addr  is the same as
       that of "sockaddr_storage" itself: only their layout is different:
    */
    unsigned int len  = sizeof(struct sockaddr_storage);
    unsigned int alen = len;
    char         abuf  [len];
    int res = inet_get_address(abuf, (inet_address*) addr, &alen);
    ASSERT(res==0); (void)res;

    /* Now "abuf" contains: Family(1b), Port(2b), IP(4|16b) */

    /* NB: the following functions are safe to use, as they create tuples
       of copied Ints on the "spec", and do not install any String pts --
       a ptr to "abuf" would be dangling upon exiting this function:   */
    i = load_address(spec, i, abuf);  /* IP,Port | Family,Addr */
    i = LOAD_TUPLE     (spec, i, 2);
    return i;
}

/* Loading Boolean flags as Atoms: */
#define LOAD_BOOL_CNT LOAD_ATOM_CNT
#define LOAD_BOOL(spec,   i,   flag)                          \
	LOAD_ATOM((spec), (i), (flag) ? am_true : am_false);
#endif /* HAVE_SCTP */

/* Assume a cache line size of 64 bytes */
#define INET_DRV_CACHE_LINE_SIZE ((ErlDrvUInt) 64)
#define INET_DRV_CACHE_LINE_MASK (INET_DRV_CACHE_LINE_SIZE - 1)

/*
** Binary Buffer Managment
** We keep a stack of usable buffers
*/
#define BUFFER_STACK_SIZE 14
#define BUFFER_STACK_MAX_MEM_SIZE (1024*1024)

ErlDrvTSDKey buffer_stack_key;

typedef struct {
    int mem_size;
    int pos;
    ErlDrvBinary* stk[BUFFER_STACK_SIZE];
} InetDrvBufStkBase;

typedef struct {
    InetDrvBufStkBase buf;
    char align[(((sizeof(InetDrvBufStkBase) - 1) / INET_DRV_CACHE_LINE_SIZE) + 1)
	       * INET_DRV_CACHE_LINE_SIZE];
} InetDrvBufStk;

static InetDrvBufStk *get_bufstk(void)
{
    InetDrvBufStk *bs = erl_drv_tsd_get(buffer_stack_key);
    if (bs)
	return bs;
    bs = driver_alloc(sizeof(InetDrvBufStk)
		      + INET_DRV_CACHE_LINE_SIZE - 1);
    if (!bs)
	return NULL;
    if ((((ErlDrvUInt) bs) & INET_DRV_CACHE_LINE_MASK) != 0)
	bs = ((InetDrvBufStk *)
	      ((((ErlDrvUInt) bs) & ~INET_DRV_CACHE_LINE_MASK)
	       + INET_DRV_CACHE_LINE_SIZE));
    erl_drv_tsd_set(buffer_stack_key, bs);
    bs->buf.pos = 0;
    bs->buf.mem_size = 0;

    ASSERT(bs == erl_drv_tsd_get(buffer_stack_key));

    return bs;
}

static ErlDrvBinary* alloc_buffer(ErlDrvSizeT minsz)
{
    InetDrvBufStk *bs = get_bufstk();

    DEBUGF(("alloc_buffer: "LLU"\r\n", (llu_t)minsz));

    if (bs && bs->buf.pos > 0) {
	long size;
	ErlDrvBinary* buf = bs->buf.stk[--bs->buf.pos];
	size = buf->orig_size;
	bs->buf.mem_size -= size;
	ASSERT(0 <= bs->buf.mem_size
	       && bs->buf.mem_size <= BUFFER_STACK_MAX_MEM_SIZE);
	if (size >= minsz)
	    return buf;

	driver_free_binary(buf);
    }

    ASSERT(!bs || bs->buf.pos != 0 || bs->buf.mem_size == 0);

    return driver_alloc_binary(minsz);
}

/*#define CHECK_DOUBLE_RELEASE 1*/
#ifdef CHECK_DOUBLE_RELEASE
static void
check_double_release(InetDrvBufStk *bs, ErlDrvBinary* buf)
{
#ifdef __GNUC__
#warning CHECK_DOUBLE_RELEASE is enabled, this is a custom build emulator
#endif
    int i;
    for (i = 0; i < bs->buf.pos; ++i) {
	if (bs->buf.stk[i] == buf) {
	    erts_exit(ERTS_ABORT_EXIT,
		     "Multiple buffer release in inet_drv, this "
		     "is a bug, save the core and send it to "
		     "support@erlang.ericsson.se!");
	}
    }
}
#endif

static void release_buffer(ErlDrvBinary* buf)
{
    InetDrvBufStk *bs;
    long size;

    DEBUGF(("release_buffer: %ld\r\n", (buf==NULL) ? 0 : buf->orig_size));

    if (!buf)
	return;

    size = buf->orig_size;

    if (size > BUFFER_STACK_MAX_MEM_SIZE)
	goto free_binary;

    bs = get_bufstk();
    if (!bs
	|| (bs->buf.mem_size + size > BUFFER_STACK_MAX_MEM_SIZE)
	|| (bs->buf.pos >= BUFFER_STACK_SIZE)) {
    free_binary:
	driver_free_binary(buf);
    }
    else {
#ifdef CHECK_DOUBLE_RELEASE
	check_double_release(bs, buf);
#endif
	ASSERT(bs->buf.pos != 0 || bs->buf.mem_size == 0);

	bs->buf.mem_size += size;
	bs->buf.stk[bs->buf.pos++] = buf;

	ASSERT(0 <= bs->buf.mem_size
	       && bs->buf.mem_size <= BUFFER_STACK_MAX_MEM_SIZE);
    }
}

#ifdef HAVE_UDP
static ErlDrvBinary* realloc_buffer(ErlDrvBinary* buf, ErlDrvSizeT newsz)
{
    DEBUGF(("realloc_buffer: %ld -> %ld\r\n", (buf==NULL) ? 0 : buf->orig_size, newsz));
    return driver_realloc_binary(buf, newsz);
}
#endif

/* use a TRICK, access the refc field to see if any one else has
 * a ref to this buffer then call driver_free_binary else
 * release_buffer instead
 */
static void free_buffer(ErlDrvBinary* buf)
{
    DEBUGF(("free_buffer: %ld\r\n", (buf==NULL) ? 0 : buf->orig_size));

    if (buf != NULL) {
	if (driver_binary_get_refc(buf) == 1)
	    release_buffer(buf);
	else
	    driver_free_binary(buf);
    }
}


#ifdef __WIN32__

static ErlDrvData dummy_start(ErlDrvPort port, char* command)
{
    return (ErlDrvData)port;
}

static ErlDrvSSizeT dummy_ctl(ErlDrvData data, unsigned int cmd,
			      char* buf, ErlDrvSizeT len, char** rbuf,
			      ErlDrvSizeT rsize)
{
    static char error[] = "no_winsock2";

    driver_failure_atom((ErlDrvPort)data, error);
    return ctl_reply(INET_REP_ERROR, error, sizeof(error), rbuf, rsize);
}

static void dummy_command(ErlDrvData data, char* buf, ErlDrvSizeT len)
{
}

static struct erl_drv_entry dummy_tcp_driver_entry =
{
    NULL,			/* init */
    dummy_start,		/* start */
    NULL,			/* stop */
    dummy_command,		/* command */
    NULL,			/* input */
    NULL,			/* output */
    "tcp_inet",			/* name */
    NULL,
    NULL,
    dummy_ctl,
    NULL,
    NULL
};

static struct erl_drv_entry dummy_udp_driver_entry =
{
    NULL,			/* init */
    dummy_start,		/* start */
    NULL,			/* stop */
    dummy_command,		/* command */
    NULL,			/* input */
    NULL,			/* output */
    "udp_inet",			/* name */
    NULL,
    NULL,
    dummy_ctl,
    NULL,
    NULL
};

#ifdef HAVE_SCTP
static struct erl_drv_entry dummy_sctp_driver_entry =
{				/* Though there is no SCTP for Win32 yet... */
    NULL,			/* init */
    dummy_start,		/* start */
    NULL,			/* stop */
    dummy_command,		/* command */
    NULL,			/* input */
    NULL,			/* output */
    "sctp_inet",		/* name */
    NULL,
    NULL,
    dummy_ctl,
    NULL,
    NULL
};
#endif

#endif

/* return lowercase string form of errno value */
static char *errno_str(int err)
{
    switch (err) {
    case INET_ERRNO_SYSTEM_LIMIT:
	return "system_limit";
    default:
	return erl_errno_id(err);
    }
}

/* general control reply function */
static ErlDrvSSizeT ctl_reply(int rep, char* buf, ErlDrvSizeT len,
			      char** rbuf, ErlDrvSizeT rsize)
{
    char* ptr;

    if ((len+1) > rsize) {
	ptr = ALLOC(len+1);
	*rbuf = ptr;
    }
    else
	ptr = *rbuf;
    *ptr++ = rep;
    memcpy(ptr, buf, len);
    return len+1;
}

/* general control error reply function */
static ErlDrvSSizeT ctl_error(int err, char** rbuf, ErlDrvSizeT rsize)
{
    char* s = errno_str(err);

    return ctl_reply(INET_REP_ERROR, s, strlen(s), rbuf, rsize);
}

static ErlDrvSSizeT ctl_xerror(char* xerr, char** rbuf, ErlDrvSizeT rsize)
{
    int n = strlen(xerr);
    return ctl_reply(INET_REP_ERROR, xerr, n, rbuf, rsize);
}


static ErlDrvTermData error_atom(int err)
{
    return driver_mk_atom(errno_str(err));
}


static void enq_old_multi_op(tcp_descriptor *desc, int id, int req,
			     ErlDrvTermData caller, MultiTimerData *timeout,
			     ErlDrvMonitor *monitorp)
{
    inet_async_multi_op *opp;

    opp = ALLOC(sizeof(inet_async_multi_op));

    opp->op.id = id;
    opp->op.caller = caller;
    opp->op.req = req;
    opp->op.tmo.mtd = timeout;
    memcpy(&(opp->op.monitor), monitorp, sizeof(ErlDrvMonitor));
    opp->next = NULL;

    if (desc->multi_first == NULL) {
	desc->multi_first = opp;
    } else {
	desc->multi_last->next = opp;
    }
    desc->multi_last = opp;
}

static void enq_multi_op(tcp_descriptor *desc, char *buf, int req,
			 ErlDrvTermData caller, MultiTimerData *timeout,
			 ErlDrvMonitor *monitorp)
{
    int id = NEW_ASYNC_ID(INETP(desc));
    enq_old_multi_op(desc,id,req,caller,timeout,monitorp);
    if (buf != NULL)
	put_int16(id, buf);
}

static int deq_multi_op(tcp_descriptor *desc, int *id_p, int *req_p,
			ErlDrvTermData *caller_p, MultiTimerData **timeout_p,
			ErlDrvMonitor *monitorp)
{
    inet_async_multi_op *opp;
    opp = desc->multi_first;
    if (!opp) {
	return -1;
    }
    desc->multi_first = opp->next;
    if (desc->multi_first == NULL) {
	desc->multi_last = NULL;
    }
    *id_p = opp->op.id;
    *req_p = opp->op.req;
    *caller_p = opp->op.caller;
    if (timeout_p != NULL) {
	*timeout_p = opp->op.tmo.mtd;
    }
    if (monitorp != NULL) {
	memcpy(monitorp,&(opp->op.monitor),sizeof(ErlDrvMonitor));
    }
    FREE(opp);
    return 0;
}

static int remove_multi_op(tcp_descriptor *desc, int *id_p, int *req_p,
			   ErlDrvTermData caller, MultiTimerData **timeout_p,
			   ErlDrvMonitor *monitorp)
{
    inet_async_multi_op *opp, *slap;
    for (opp = desc->multi_first, slap = NULL;
	 opp != NULL && opp->op.caller != caller;
	 slap = opp, opp = opp->next)
	;
    if (!opp) {
	return -1;
    }
    if (slap == NULL) {
	desc->multi_first = opp->next;
    } else {
	slap->next = opp->next;
    }
    if (desc->multi_last == opp) {
	desc->multi_last = slap;
    }
    *id_p = opp->op.id;
    *req_p = opp->op.req;
    if (timeout_p != NULL) {
	*timeout_p = opp->op.tmo.mtd;
    }
    if (monitorp != NULL) {
	memcpy(monitorp,&(opp->op.monitor),sizeof(ErlDrvMonitor));
    }
    FREE(opp);
    return 0;
}

/* setup a new async id + caller (format async_id into buf) */

static int enq_async_w_tmo(inet_descriptor* desc, char* buf, int req, unsigned timeout,
			   ErlDrvMonitor *monitorp)
{
    int id = NEW_ASYNC_ID(desc);
    inet_async_op* opp;

    if ((opp = desc->oph) == NULL)            /* queue empty */
	opp = desc->oph = desc->opt = desc->op_queue;
    else if (desc->oph == desc->opt) { /* queue full */
	DEBUGF(("enq(%ld): queue full\r\n", (long)desc->port));
	return -1;
    }

    opp->id = id;
    opp->caller = driver_caller(desc->port);
    opp->req = req;
    opp->tmo.value = timeout;
    if (monitorp != NULL) {
	memcpy(&(opp->monitor),monitorp,sizeof(ErlDrvMonitor));
    }

    DEBUGF(("enq(%ld): %d %ld %d\r\n",
	    (long) desc->port, opp->id, opp->caller, opp->req));

    opp++;
    if (opp >= desc->op_queue + INET_MAX_ASYNC)
	desc->oph = desc->op_queue;
    else
	desc->oph = opp;

    if (buf != NULL)
	put_int16(id, buf);
    return 0;
}

static int enq_async(inet_descriptor* desc, char* buf, int req)
{
    return enq_async_w_tmo(desc,buf,req,INET_INFINITY, NULL);
}

static int deq_async_w_tmo(inet_descriptor* desc, int* ap, ErlDrvTermData* cp,
			   int* rp, unsigned *tp, ErlDrvMonitor *monitorp)
{
    inet_async_op* opp;

    if ((opp = desc->opt) == NULL) {  /* queue empty */
	DEBUGF(("deq(%ld): queue empty\r\n", (long)desc->port));
	return -1;
    }
    *ap = opp->id;
    *cp = opp->caller;
    *rp = opp->req;
    if (tp != NULL) {
	*tp = opp->tmo.value;
    }
    if (monitorp != NULL) {
	memcpy(monitorp,&(opp->monitor),sizeof(ErlDrvMonitor));
    }

    DEBUGF(("deq(%ld): %d %ld %d\r\n",
	    (long)desc->port, opp->id, opp->caller, opp->req));

    opp++;
    if (opp >= desc->op_queue + INET_MAX_ASYNC)
	desc->opt = desc->op_queue;
    else
	desc->opt = opp;

    if (desc->opt == desc->oph)
	desc->opt = desc->oph = NULL;
    return 0;
}

static int deq_async(inet_descriptor* desc, int* ap, ErlDrvTermData* cp, int* rp)
{
    return deq_async_w_tmo(desc,ap,cp,rp,NULL,NULL);
}
/* send message:
**     {inet_async, Port, Ref, ok}
*/
static int
send_async_ok(ErlDrvTermData Port, int Ref,ErlDrvTermData recipient)
{
    ErlDrvTermData spec[2*LOAD_ATOM_CNT + LOAD_PORT_CNT +
			LOAD_INT_CNT + LOAD_TUPLE_CNT];
    int i = 0;

    i = LOAD_ATOM(spec, i, am_inet_async);
    i = LOAD_PORT(spec, i, Port);
    i = LOAD_INT(spec, i, Ref);
    i = LOAD_ATOM(spec, i, am_ok);
    i = LOAD_TUPLE(spec, i, 4);

    ASSERT(i == sizeof(spec)/sizeof(*spec));

    return erl_drv_send_term(Port, recipient, spec, i);
}

/* send message:
**     {inet_async, Port, Ref, {ok,Port2}}
*/
static int
send_async_ok_port(ErlDrvTermData Port, int Ref,
		   ErlDrvTermData recipient, ErlDrvTermData Port2)
{
    ErlDrvTermData spec[2*LOAD_ATOM_CNT + 2*LOAD_PORT_CNT +
			LOAD_INT_CNT + 2*LOAD_TUPLE_CNT];
    int i = 0;

    i = LOAD_ATOM(spec, i, am_inet_async);
    i = LOAD_PORT(spec, i, Port);
    i = LOAD_INT(spec, i, Ref);
    {
	i = LOAD_ATOM(spec, i, am_ok);
	i = LOAD_PORT(spec, i, Port2);
	i = LOAD_TUPLE(spec, i, 2);
    }
    i = LOAD_TUPLE(spec, i, 4);

    ASSERT(i == sizeof(spec)/sizeof(*spec));

    return erl_drv_send_term(Port, recipient, spec, i);
}

/* send message:
**      {inet_async, Port, Ref, {error,Reason}}
*/
static int
send_async_error(ErlDrvTermData Port, int Ref,
		 ErlDrvTermData recipient, ErlDrvTermData Reason)
{
    ErlDrvTermData spec[3*LOAD_ATOM_CNT + LOAD_PORT_CNT +
			LOAD_INT_CNT + 2*LOAD_TUPLE_CNT];
    int i = 0;

    i = LOAD_ATOM(spec, i, am_inet_async);
    i = LOAD_PORT(spec, i, Port);
    i = LOAD_INT(spec, i, Ref);
    {
	i = LOAD_ATOM(spec, i, am_error);
	i = LOAD_ATOM(spec, i, Reason);
	i = LOAD_TUPLE(spec, i, 2);
    }
    i = LOAD_TUPLE(spec, i, 4);
    ASSERT(i == sizeof(spec)/sizeof(*spec));
    DEBUGF(("send_async_error %ld %ld\r\n", recipient, Reason));
    return erl_drv_send_term(Port, recipient, spec, i);
}


static int async_ok(inet_descriptor* desc)
{
    int req;
    int aid;
    ErlDrvTermData caller;

    if (deq_async(desc, &aid, &caller, &req) < 0)
	return -1;
    return send_async_ok(desc->dport, aid, caller);
}

static int async_ok_port(inet_descriptor* desc, ErlDrvTermData Port2)
{
    int req;
    int aid;
    ErlDrvTermData caller;

    if (deq_async(desc, &aid, &caller, &req) < 0)
	return -1;
    return send_async_ok_port(desc->dport, aid, caller, Port2);
}

static int async_error_am(inet_descriptor* desc, ErlDrvTermData reason)
{
    int req;
    int aid;
    ErlDrvTermData caller;

    if (deq_async(desc, &aid, &caller, &req) < 0)
	return -1;
    return send_async_error(desc->dport, aid, caller, reason);
}

/* dequeue all operations */
static int async_error_am_all(inet_descriptor* desc, ErlDrvTermData reason)
{
    int req;
    int aid;
    ErlDrvTermData caller;

    while (deq_async(desc, &aid, &caller, &req) == 0) {
	send_async_error(desc->dport, aid, caller, reason);
    }
    return 0;
}


static int async_error(inet_descriptor* desc, int err)
{
    return async_error_am(desc, error_atom(err));
}

/* send:
**   {inet_reply, S, ok}
*/

static int inet_reply_ok(inet_descriptor* desc)
{
    ErlDrvTermData spec[2*LOAD_ATOM_CNT + LOAD_PORT_CNT + LOAD_TUPLE_CNT];
    ErlDrvTermData caller = desc->caller;
    int i = 0;

    desc->caller = 0;
    if (is_not_internal_pid(caller))
        return 0;

    i = LOAD_ATOM(spec, i, am_inet_reply);
    i = LOAD_PORT(spec, i, desc->dport);
    i = LOAD_ATOM(spec, i, am_ok);
    i = LOAD_TUPLE(spec, i, 3);
    ASSERT(i == sizeof(spec)/sizeof(*spec));

    return erl_drv_send_term(desc->dport, caller, spec, i);
}

#ifdef HAVE_SCTP
static int inet_reply_ok_port(inet_descriptor* desc, ErlDrvTermData dport)
{
    ErlDrvTermData spec[2*LOAD_ATOM_CNT + 2*LOAD_PORT_CNT + 2*LOAD_TUPLE_CNT];
    ErlDrvTermData caller = desc->caller;
    int i = 0;

    i = LOAD_ATOM(spec, i, am_inet_reply);
    i = LOAD_PORT(spec, i, desc->dport);
    i = LOAD_ATOM(spec, i, am_ok);
    i = LOAD_PORT(spec, i, dport);
    i = LOAD_TUPLE(spec, i, 2);
    i = LOAD_TUPLE(spec, i, 3);
    ASSERT(i == sizeof(spec)/sizeof(*spec));

    desc->caller = 0;
    return erl_drv_send_term(desc->dport, caller, spec, i);
}
#endif

/* send:
**   {inet_reply, S, {error, Reason}}
*/
static int inet_reply_error_am(inet_descriptor* desc, ErlDrvTermData reason)
{
    ErlDrvTermData spec[3*LOAD_ATOM_CNT + LOAD_PORT_CNT + 2*LOAD_TUPLE_CNT];
    ErlDrvTermData caller = desc->caller;
    int i = 0;

    i = LOAD_ATOM(spec, i, am_inet_reply);
    i = LOAD_PORT(spec, i, desc->dport);
    i = LOAD_ATOM(spec, i, am_error);
    i = LOAD_ATOM(spec, i, reason);
    i = LOAD_TUPLE(spec, i, 2);
    i = LOAD_TUPLE(spec, i, 3);
    ASSERT(i == sizeof(spec)/sizeof(*spec));
    desc->caller = 0;

    DEBUGF(("inet_reply_error_am %ld %ld\r\n", caller, reason));
    return erl_drv_send_term(desc->dport, caller, spec, i);
}

/* send:
**   {inet_reply, S, {error, Reason}}
*/
static int inet_reply_error(inet_descriptor* desc, int err)
{
    return inet_reply_error_am(desc, error_atom(err));
}

/*
** Deliver port data from buffer
*/
static int inet_port_data(inet_descriptor* desc, const char* buf, int len)
{
    unsigned int hsz = desc->hsz;

    DEBUGF(("inet_port_data(%ld): len = %d\r\n", (long)desc->port, len));

    if ((desc->mode == INET_MODE_LIST) || (hsz > len))
	return driver_output2(desc->port, (char*)buf, len, NULL, 0);
    else if (hsz > 0)
	return driver_output2(desc->port, (char*)buf, hsz, (char*)buf+hsz, len-hsz);
    else
	return driver_output(desc->port, (char*)buf, len);
}

/*
** Deliver port data from binary (for an active mode socket)
*/
static int
inet_port_binary_data(inet_descriptor* desc, ErlDrvBinary* bin, int offs, int len)
{
    unsigned int hsz = desc->hsz;

    DEBUGF(("inet_port_binary_data(%ld): offs=%d, len = %d\r\n",
	    (long)desc->port, offs, len));

    if ((desc->mode == INET_MODE_LIST) || (hsz > len))
	return driver_output2(desc->port, bin->orig_bytes+offs, len, NULL, 0);
    else
	return driver_output_binary(desc->port, bin->orig_bytes+offs, hsz,
				    bin, offs+hsz, len-hsz);
}

static ErlDrvTermData am_http_eoh;
static ErlDrvTermData am_http_header;
static ErlDrvTermData am_http_request;
static ErlDrvTermData am_http_response;
static ErlDrvTermData am_http_error;
static ErlDrvTermData am_abs_path;
static ErlDrvTermData am_absoluteURI;
static ErlDrvTermData am_star;
static ErlDrvTermData am_http;
static ErlDrvTermData am_https;
static ErlDrvTermData am_scheme;

static int http_load_string(tcp_descriptor* desc, ErlDrvTermData* spec, int i,
			    const char* str, int len)
{
    if (desc->inet.htype >= TCP_PB_HTTP_BIN) {
	ASSERT(desc->inet.htype == TCP_PB_HTTP_BIN ||
	       desc->inet.htype == TCP_PB_HTTPH_BIN);
	i = LOAD_BUF2BINARY(spec, i, str, len);
    } else {
	i = LOAD_STRING(spec, i, str, len);
    }
    return i;
}

static int http_response_inetdrv(void *arg, int major, int minor,
				 int status, const char* phrase, int phrase_len)
{
    tcp_descriptor* desc = (tcp_descriptor*) arg;
    int i = 0;
    ErlDrvTermData spec[27];
    ErlDrvTermData caller = ERL_DRV_NIL;

    if (desc->inet.active == INET_PASSIVE) {
        /* {inet_async,S,Ref,{ok,{http_response,Version,Status,Phrase}}} */
        int req;
        int aid;

        if (deq_async(INETP(desc), &aid, &caller, &req) < 0)
            return -1;
        i = LOAD_ATOM(spec, i,  am_inet_async);
        i = LOAD_PORT(spec, i,  desc->inet.dport);
        i = LOAD_INT(spec, i,   aid);
        i = LOAD_ATOM(spec, i,  am_ok);
    }
    else {
        /* {http, S, {http_response,Version,Status,Phrase}} */
        i = LOAD_ATOM(spec, i, am_http);
        i = LOAD_PORT(spec, i, desc->inet.dport);
    }
    i = LOAD_ATOM(spec, i,  am_http_response);
    i = LOAD_INT(spec, i, major);
    i = LOAD_INT(spec, i, minor);
    i = LOAD_TUPLE(spec, i, 2);
    i = LOAD_INT(spec, i, status);
    i = http_load_string(desc, spec, i, phrase, phrase_len);
    i = LOAD_TUPLE(spec, i, 4);

    if (desc->inet.active == INET_PASSIVE) {
        i = LOAD_TUPLE(spec, i, 2);
        i = LOAD_TUPLE(spec, i, 4);
        ASSERT(i<=27);
        return erl_drv_send_term(desc->inet.dport, caller, spec, i);
    }
    else {
        i = LOAD_TUPLE(spec, i, 3);
        ASSERT(i<=27);
        return erl_drv_output_term(desc->inet.dport, spec, i);
    }
}

static int http_load_uri(tcp_descriptor* desc, ErlDrvTermData* spec, int i,
			 const PacketHttpURI* uri)
{
    ErlDrvTermData scheme;

    switch (uri->type) {
    case URI_STAR:
        i = LOAD_ATOM(spec, i, am_star);
        break;
    case URI_ABS_PATH:
        i = LOAD_ATOM(spec, i, am_abs_path);
        i = http_load_string(desc, spec, i, uri->s1_ptr, uri->s1_len);
        i = LOAD_TUPLE(spec, i, 2);
        break;
    case URI_HTTP:
        scheme = am_http;
        goto http_common;
    case URI_HTTPS:
        scheme = am_https;
    http_common:
        i = LOAD_ATOM(spec, i, am_absoluteURI);
        i = LOAD_ATOM(spec, i, scheme);
        i = http_load_string(desc, spec, i, uri->s1_ptr, uri->s1_len);
        if (uri->port == 0) {
            i = LOAD_ATOM(spec, i, am_undefined);
        } else {
            i = LOAD_INT(spec, i, uri->port);
        }
        i = http_load_string(desc, spec, i, uri->s2_ptr, uri->s2_len);
        i = LOAD_TUPLE(spec, i, 5);
        break;

    case URI_STRING:
        i = http_load_string(desc, spec, i, uri->s1_ptr, uri->s1_len);
        break;
    case URI_SCHEME:
        i = LOAD_ATOM(spec, i, am_scheme);
        i = http_load_string(desc, spec, i, uri->s1_ptr, uri->s1_len);
        i = http_load_string(desc, spec, i, uri->s2_ptr, uri->s2_len);
        i = LOAD_TUPLE(spec, i, 3);
    }
    return i;
}


static int
http_request_inetdrv(void* arg, const http_atom_t* meth, const char* meth_ptr,
		     int meth_len, const PacketHttpURI* uri,
		     int major, int minor)
{
    tcp_descriptor* desc = (tcp_descriptor*) arg;
    int i = 0;
    ErlDrvTermData spec[43];
    ErlDrvTermData caller = ERL_DRV_NIL;

    if (desc->inet.active == INET_PASSIVE) {
        /* {inet_async, S, Ref, {ok,{http_request,Meth,Uri,Version}}} */
        int req;
        int aid;

        if (deq_async(INETP(desc), &aid, &caller, &req) < 0)
            return -1;
        i = LOAD_ATOM(spec, i,  am_inet_async);
        i = LOAD_PORT(spec, i,  desc->inet.dport);
        i = LOAD_INT(spec, i,   aid);
        i = LOAD_ATOM(spec, i,  am_ok);
    }
    else {
        /* {http, S, {http_request,Meth,Uri,Version}}} */
        i = LOAD_ATOM(spec, i, am_http);
        i = LOAD_PORT(spec, i, desc->inet.dport);
    }

    i = LOAD_ATOM(spec, i,  am_http_request);
    if (meth != NULL)
      i = LOAD_ATOM(spec, i, meth->atom);
    else
      i = http_load_string(desc, spec, i, meth_ptr, meth_len);
    i = http_load_uri(desc, spec, i, uri);
    i = LOAD_INT(spec, i, major);
    i = LOAD_INT(spec, i, minor);
    i = LOAD_TUPLE(spec, i, 2);
    i = LOAD_TUPLE(spec, i, 4);

    if (desc->inet.active == INET_PASSIVE) {
        i = LOAD_TUPLE(spec, i, 2);
        i = LOAD_TUPLE(spec, i, 4);
        ASSERT(i <= 43);
        return erl_drv_send_term(desc->inet.dport, caller, spec, i);
    }
    else {
        i = LOAD_TUPLE(spec, i, 3);
        ASSERT(i <= 43);
        return erl_drv_output_term(desc->inet.dport, spec, i);
    }
}

static int
http_header_inetdrv(void* arg, const http_atom_t* name, const char* name_ptr,
		    int name_len, const char* value_ptr, int value_len)
{
    tcp_descriptor* desc = (tcp_descriptor*) arg;
    int i = 0;
    ErlDrvTermData spec[26];
    ErlDrvTermData caller = ERL_DRV_NIL;

    if (desc->inet.active == INET_PASSIVE) {
        /* {inet_async,S,Ref,{ok,{http_header,Bit,Name,IValue,Value}} */
        int req;
        int aid;


        if (deq_async(INETP(desc), &aid, &caller, &req) < 0)
            return -1;
        i = LOAD_ATOM(spec, i,  am_inet_async);
        i = LOAD_PORT(spec, i,  desc->inet.dport);
        i = LOAD_INT(spec, i,   aid);
        i = LOAD_ATOM(spec, i,  am_ok);
    }
    else {
        /* {http, S, {http_header,Bit,Name,IValue,Value}} */
        i = LOAD_ATOM(spec, i, am_http);
        i = LOAD_PORT(spec, i, desc->inet.dport);
    }

    i = LOAD_ATOM(spec, i,  am_http_header);
    if (name != NULL) {
      i = LOAD_INT(spec, i,  name->index+1);
      i = LOAD_ATOM(spec, i, name->atom);
    }
    else {
      i = LOAD_INT(spec, i,  0);
      i = http_load_string(desc, spec, i, name_ptr, name_len);
    }
    i = LOAD_ATOM(spec, i, am_undefined);
    i = http_load_string(desc, spec, i, value_ptr, value_len);
    i = LOAD_TUPLE(spec, i, 5);

    if (desc->inet.active == INET_PASSIVE) {
        i = LOAD_TUPLE(spec, i, 2);
        i = LOAD_TUPLE(spec, i, 4);
        ASSERT(i <= 26);
        return erl_drv_send_term(desc->inet.dport, caller, spec, i);
    }
    else {
        i = LOAD_TUPLE(spec, i, 3);
        ASSERT(i <= 26);
        return erl_drv_output_term(desc->inet.dport, spec, i);
    }
}

static int http_eoh_inetdrv(void* arg)
{
  tcp_descriptor* desc = (tcp_descriptor*) arg;
  int i = 0;
  ErlDrvTermData spec[14];

  if (desc->inet.active == INET_PASSIVE) {
    /* {inet_async,S,Ref,{ok,http_eoh}} */
    int req;
    int aid;
    ErlDrvTermData caller;

    if (deq_async(INETP(desc), &aid, &caller, &req) < 0)
      return -1;
    i = LOAD_ATOM(spec, i,  am_inet_async);
    i = LOAD_PORT(spec, i,  desc->inet.dport);
    i = LOAD_INT(spec, i,   aid);
    i = LOAD_ATOM(spec, i,  am_ok);
    i = LOAD_ATOM(spec, i,  am_http_eoh);
    i = LOAD_TUPLE(spec, i, 2);
    i = LOAD_TUPLE(spec, i, 4);
    ASSERT(i <= 14);
    return erl_drv_send_term(desc->inet.dport, caller, spec, i);
  }
  else {
      /* {http, S, http_eoh} */
      i = LOAD_ATOM(spec, i,  am_http);
      i = LOAD_PORT(spec, i,  desc->inet.dport);
      i = LOAD_ATOM(spec, i,  am_http_eoh);
      i = LOAD_TUPLE(spec, i, 3);
      ASSERT(i <= 14);
      return erl_drv_output_term(desc->inet.dport, spec, i);
  }
}

static int http_error_inetdrv(void* arg, const char* buf, int len)
{
  tcp_descriptor* desc = (tcp_descriptor*) arg;
  int i = 0;
  ErlDrvTermData spec[19];

  if (desc->inet.active == INET_PASSIVE) {
    /* {inet_async,S,Ref,{ok,{http_error,Line}}} */
    int req;
    int aid;
    ErlDrvTermData caller;

    if (deq_async(INETP(desc), &aid, &caller, &req) < 0)
      return -1;
    i = LOAD_ATOM(spec, i,  am_inet_async);
    i = LOAD_PORT(spec, i,  desc->inet.dport);
    i = LOAD_INT(spec, i,   aid);
    i = LOAD_ATOM(spec, i,  am_ok);
    i = LOAD_ATOM(spec, i,  am_http_error);
    i = http_load_string(desc, spec, i, buf, len);
    i = LOAD_TUPLE(spec, i, 2);
    i = LOAD_TUPLE(spec, i, 2);
    i = LOAD_TUPLE(spec, i, 4);
    ASSERT(i <= 19);
    return erl_drv_send_term(desc->inet.dport, caller, spec, i);
  }
  else {
      /* {http, S, {http_error,Line} */
      i = LOAD_ATOM(spec, i,  am_http);
      i = LOAD_PORT(spec, i,  desc->inet.dport);
      i = LOAD_ATOM(spec, i,  am_http_error);
      i = http_load_string(desc, spec, i, buf, len);
      i = LOAD_TUPLE(spec, i, 2);
      i = LOAD_TUPLE(spec, i, 3);
      ASSERT(i <= 19);
      return erl_drv_output_term(desc->inet.dport, spec, i);
  }
}


static
int ssl_tls_inetdrv(void* arg, unsigned type, unsigned major, unsigned minor,
                    const char* buf, int len, const char* prefix, int plen)
{
    tcp_descriptor* desc = (tcp_descriptor*) arg;
    int i = 0;
    ErlDrvTermData spec[30];
    ErlDrvTermData caller = ERL_DRV_NIL;
    ErlDrvBinary* bin;
    int ret;

    if ((bin = driver_alloc_binary(plen+len)) == NULL)
        return async_error(&desc->inet, ENOMEM);
    memcpy(bin->orig_bytes+plen, buf, len);
    if (plen) {
        memcpy(bin->orig_bytes, prefix, plen);
        len += plen;
    }

    if (desc->inet.active == INET_PASSIVE) {
        /* {inet_async,S,Ref,{ok,{ssl_tls,...}}} */
        int req;
        int aid;

        if (deq_async(INETP(desc), &aid, &caller, &req) < 0) {
            ret = -1;
            goto done;
        }
        i = LOAD_ATOM(spec, i,  am_inet_async);
        i = LOAD_PORT(spec, i,  desc->inet.dport);
        i = LOAD_INT(spec, i,   aid);
        i = LOAD_ATOM(spec, i,  am_ok);
    }

    /* {ssl_tls,S,ContentType,{Major,Minor},Bin} */
    i = LOAD_ATOM(spec, i,  am_ssl_tls);
    i = LOAD_PORT(spec, i,  desc->inet.dport);
    i = LOAD_INT(spec, i,   type);
    i = LOAD_INT(spec, i,   major);
    i = LOAD_INT(spec, i,   minor);
    i = LOAD_TUPLE(spec, i, 2);
    i = LOAD_BINARY(spec, i, bin, 0, len);
    i = LOAD_TUPLE(spec, i, 5);

    if (desc->inet.active == INET_PASSIVE) {
        i = LOAD_TUPLE(spec, i, 2);
        i = LOAD_TUPLE(spec, i, 4);
        ASSERT(i <= sizeof(spec)/sizeof(*spec));
        ret = erl_drv_send_term(desc->inet.dport, caller, spec, i);
    }
    else {
        ASSERT(i <= sizeof(spec)/sizeof(*spec));
        ret = erl_drv_output_term(desc->inet.dport, spec, i);
    }
done:
    driver_free_binary(bin);
    return ret;
}


static PacketCallbacks packet_callbacks =
{
    http_response_inetdrv,
    http_request_inetdrv,
    http_eoh_inetdrv,
    http_header_inetdrv,
    http_error_inetdrv,
    ssl_tls_inetdrv
};


/*
** passive mode reply:
**        {inet_async, S, Ref, {ok,[H1,...Hsz | Data]}}
** NB: this is for TCP only;
** UDP and SCTP use inet_async_binary_data .
*/
static int inet_async_data(inet_descriptor* desc, const char* buf, int len)
{
    unsigned int hsz = desc->hsz;
    ErlDrvTermData spec[20];
    ErlDrvTermData caller;
    int req;
    int aid;
    int i = 0;

    DEBUGF(("inet_async_data(%ld): len = %d\r\n", (long)desc->port, len));

    if (deq_async(desc, &aid, &caller, &req) < 0)
	return -1;

    i = LOAD_ATOM(spec, i, am_inet_async);
    i = LOAD_PORT(spec, i, desc->dport);
    i = LOAD_INT(spec, i, aid);

    i = LOAD_ATOM(spec, i, am_ok);
    if ((desc->mode == INET_MODE_LIST) || (hsz > len)) {
	i = LOAD_STRING(spec, i, buf, len); /* => [H1,H2,...Hn] */
	i = LOAD_TUPLE(spec, i, 2);
	i = LOAD_TUPLE(spec, i, 4);
	ASSERT(i == 15);
	desc->caller = 0;
	return erl_drv_send_term(desc->dport, caller, spec, i);
    }
    else {
	/* INET_MODE_BINARY => [H1,H2,...HSz | Binary] */
	int sz = len - hsz;
	int code;

	i = LOAD_BUF2BINARY(spec, i, buf+hsz, sz);
	if (hsz > 0)
	    i = LOAD_STRING_CONS(spec, i, buf, hsz);
	i = LOAD_TUPLE(spec, i, 2);
	i = LOAD_TUPLE(spec, i, 4);
	ASSERT(i <= 20);
	desc->caller = 0;
	code = erl_drv_send_term(desc->dport, caller, spec, i);
	return code;
    }
}

#ifndef __WIN32__
static int load_cmsg_int(ErlDrvTermData *spec, int i,
                    struct cmsghdr *cmsg) {
    union u {
        byte uint8;
        Uint16 uint16;
        Uint32 uint32;
        Uint64 uint64;
    } *p;
    p = (union u*) CMSG_DATA(cmsg);
    switch (LEN_CMSG_DATA(cmsg) * CHAR_BIT) {
    case 8:
        return LOAD_INT(spec, i, p->uint8);
    case 16:
        return LOAD_INT(spec, i, p->uint16);

    case 32:
        return LOAD_INT(spec, i, p->uint32);

    case 64:
        return LOAD_INT(spec, i, p->uint64);
    }
    return LOAD_INT(spec, i, 0);
}

static int parse_ancillary_data_item(ErlDrvTermData *spec, int i,
                                     struct cmsghdr *cmsg, int *n) {
#define LOAD_CMSG_INT(proto, type, am)              \
    if (cmsg->cmsg_level == (proto) &&          \
        cmsg->cmsg_type == (type)) {            \
        i = LOAD_ATOM(spec, i, (am));           \
        i = load_cmsg_int(spec, i, cmsg);            \
        i = LOAD_TUPLE(spec, i, 2);             \
        (*n)++;                                 \
        return i;                               \
    }
#if defined(IPPROTO_IP) && defined(IP_TOS)
    LOAD_CMSG_INT(IPPROTO_IP, IP_TOS, am_tos);
#endif
#if defined(IPPROTO_IPV6) && defined(IPV6_TCLASS)
    LOAD_CMSG_INT(IPPROTO_IPV6, IPV6_TCLASS, am_tclass);
#endif
#if defined(IPPROTO_IP) && defined(IP_TTL)
    LOAD_CMSG_INT(IPPROTO_IP, IP_TTL, am_ttl);
#endif
    /* BSD uses the RECV* names in CMSG fields */
#if defined(IPPROTO_IP) && defined(IP_RECVTOS)
    LOAD_CMSG_INT(IPPROTO_IP, IP_RECVTOS, am_tos);
#endif
#if defined(IPPROTO_IPV6) && defined(IPV6_RECVTCLASS)
    LOAD_CMSG_INT(IPPROTO_IPV6, IPV6_RECVTCLASS, am_tclass);
#endif
#if defined(IPPROTO_IP) && defined(IP_RECVTTL)
    LOAD_CMSG_INT(IPPROTO_IP, IP_RECVTTL, am_ttl);
#endif
#undef LOAD_CMSG_INT
    return i;
}
#endif /* #ifndef __WIN32__ */

#ifdef HAVE_SCTP
/*
** SCTP-related atoms:
*/
static ErlDrvTermData   am_sctp_rtoinfo, /* Option names */
    am_sctp_associnfo,                 am_sctp_initmsg,
    am_sctp_autoclose,                 am_sctp_nodelay,
    am_sctp_disable_fragments,         am_sctp_i_want_mapped_v4_addr,
    am_sctp_maxseg,                    am_sctp_set_peer_primary_addr,
    am_sctp_primary_addr,              am_sctp_adaptation_layer,
    am_sctp_peer_addr_params,          am_sctp_default_send_param,
    am_sctp_events,                    am_sctp_delayed_ack_time,
    am_sctp_status,                    am_sctp_get_peer_addr_info,

    /* Record names */
    am_sctp_sndrcvinfo,                am_sctp_assoc_change,
    am_sctp_paddr_change,              am_sctp_remote_error,
    am_sctp_send_failed,               am_sctp_shutdown_event,
    am_sctp_adaptation_event,          am_sctp_pdapi_event,
    am_sctp_assocparams,               am_sctp_prim,
    am_sctp_setpeerprim,               am_sctp_setadaptation,
    am_sctp_paddrparams,               am_sctp_event_subscribe,
    am_sctp_assoc_value,               am_sctp_paddrinfo,

    /* For #sctp_sndrcvinfo{}: */
    am_unordered,                      am_addr_over,
    am_abort,                          am_eof,

    /* For #sctp_assoc_change{}: */
    am_comm_up,                        am_comm_lost,
    am_restart,                        am_shutdown_comp,
    am_cant_assoc,

    /* For #sctp_paddr_change{}: */
    am_addr_available,                 am_addr_unreachable,
    am_addr_removed,                   am_addr_added,
    am_addr_made_prim,                 am_addr_confirmed,

    /* For #sctp_remote_error{}: */
    am_short_recv,                     am_wrong_anc_data,

    /* For #sctp_pdap_event{}: */
    am_partial_delivery_aborted,

    /* For #sctp_paddrparams{}: */
    am_hb_enable,                      am_hb_disable,
    am_hb_demand,                      am_pmtud_enable,
    am_pmtud_disable,                  am_sackdelay_enable,
    am_sackdelay_disable,

    /* For #sctp_paddrinfo{}: */
    am_active,                         am_inactive,
#    if HAVE_DECL_SCTP_UNCONFIRMED
    am_unconfirmed,
#    endif

    /* For #sctp_status{}: */
#    if HAVE_DECL_SCTP_EMPTY
    am_empty,
#    endif
#    if HAVE_DECL_SCTP_BOUND
    am_bound,
#    endif
#    if HAVE_DECL_SCTP_LISTEN
    am_listen,
#    endif
    am_cookie_wait,                    am_cookie_echoed,
    am_established,                    am_shutdown_pending,
    am_shutdown_sent,                  am_shutdown_received,
    am_shutdown_ack_sent;

/*
** Parsing of "sctp_sndrcvinfo": ancillary data coming with received msgs.
** This function is mainly used by "sctp_parse_ancillary_data",  but also
** by "sctp_parse_async_event" in case of SCTP_SEND_FAILED:
*/
#define SCTP_PARSE_SNDRCVINFO_CNT                            \
        (5*LOAD_ATOM_CNT + 5*LOAD_INT_CNT + 2*LOAD_UINT_CNT + \
	 LOAD_NIL_CNT + LOAD_LIST_CNT + LOAD_ASSOC_ID_CNT + LOAD_TUPLE_CNT)
static int sctp_parse_sndrcvinfo
	   (ErlDrvTermData * spec, int i, struct sctp_sndrcvinfo * sri)
{
    int n;

    i = LOAD_ATOM	(spec, i, am_sctp_sndrcvinfo);
    i = LOAD_INT	(spec, i, sri->sinfo_stream);
    i = LOAD_INT	(spec, i, sri->sinfo_ssn);
    /* Now Flags, as a list: */
    n = 0;
    if (sri->sinfo_flags & SCTP_UNORDERED)
	{ i = LOAD_ATOM (spec, i, am_unordered);     n++; }

    if (sri->sinfo_flags & SCTP_ADDR_OVER)
	{ i = LOAD_ATOM (spec, i, am_addr_over);     n++; }

    if (sri->sinfo_flags & SCTP_ABORT)
	{ i = LOAD_ATOM (spec, i, am_abort);	     n++; }

    if (sri->sinfo_flags & SCTP_EOF)
	{ i = LOAD_ATOM (spec, i, am_eof);	     n++; }

    /* SCTP_SENDALL is not yet supported by the Linux kernel     */
    i = LOAD_NIL	(spec, i);
    i = LOAD_LIST	(spec, i, n+1);

    /* Continue with other top-level fields: */
    i = LOAD_INT	(spec, i, sock_ntohl(sri->sinfo_ppid));
    i = LOAD_INT	(spec, i, sri->sinfo_context);
    i = LOAD_INT	(spec, i, sri->sinfo_timetolive);
    i = LOAD_UINT	(spec, i, sri->sinfo_tsn);
    i = LOAD_UINT	(spec, i, sri->sinfo_cumtsn);
    i = LOAD_ASSOC_ID	(spec, i, sri->sinfo_assoc_id);

    /* Close up the record: */
    i = LOAD_TUPLE	(spec, i, 10);
    return i;
}

/*
** This function skips non-SCTP ancillary data, returns SCTP-specific anc.data
** (currently "sctp_sndrcvinfo" only) as a list of records:
*/
static int sctp_parse_ancillary_data
	   (ErlDrvTermData * spec, int i, struct msghdr * mptr)
{
    /* First of all, check for ancillary data: */
    struct cmsghdr * cmsg, * frst_msg = CMSG_FIRSTHDR(mptr);
    int    s = 0;
    for (cmsg = frst_msg; cmsg != NULL; cmsg = CMSG_NXTHDR(mptr,cmsg))
    {
	struct sctp_sndrcvinfo * sri;
#ifndef __WIN32
        int old_s;

        /* Parse ancillary data common to UDP */
        old_s = s;
        i = parse_ancillary_data_item(spec, i, cmsg, &s);
        if (s > old_s) continue;
	/* Skip other possible ancillary data, e.g. from IPv6: */
	if (cmsg->cmsg_level != IPPROTO_SCTP ||
	    cmsg->cmsg_type  != SCTP_SNDRCV)
	continue;
#endif

	if (((char*)cmsg + cmsg->cmsg_len) - (char*)frst_msg >
	    mptr->msg_controllen)
	    /* MUST check this in Linux --  the returned "cmsg" may actually
	       go too far! */
	    break;

	/* The ONLY kind of ancillary SCTP data which can occur on receiving
	   is "sctp_sndrcvinfo" (on sending, "sctp_initmsg" can be specified
	   by the user). So parse this type:
	*/
	sri = (struct sctp_sndrcvinfo*) CMSG_DATA(cmsg);
	i = sctp_parse_sndrcvinfo (spec, i, sri);
	s ++;
    }
    /* Now make the list of tuples created above. Normally, it will be [] or
       a singleton list.   The list must first be closed with NIL, otherwise
       traversing it in Erlang would be problematic:
    */
    i = LOAD_NIL (spec, i);
    i = LOAD_LIST(spec, i, s+1);
    return i;
}

/*
** Parsing of ERROR and ABORT SCTP chunks. The function returns a list of error
** causes (as atoms).  The chunks also contain some extended cause info, but it
** is not very detailed anyway, and of no interest at the user level   (it only
** concerns the protocol implementation), so we omit it:
*/
static int sctp_parse_error_chunk
       (ErlDrvTermData * spec, int i, char * chunk, int chlen)
{
    /* The "chunk" itself contains its length, which must not be greater than
       the "chlen" derived from the over-all msg size:
    */
    char *causes, *cause;
    int coff,  /* Cause offset */
	ccode, /* Cause code */
	clen,  /* cause length */
	s;
    int len = sock_ntohs (*((uint16_t*)(chunk+2)));
    ASSERT(len >= 4 && len <= chlen);

    causes = chunk + 4;
    coff   = 0;
    len -= 4;  /* Total length of the "causes" fields */
    cause  = causes;
    s      = 0;

    while (coff < len)
    {
	ccode = sock_ntohs (*((uint16_t*)(cause)));
	clen  = sock_ntohs (*((uint16_t*)(cause + 2)));
	if (clen <= 0)
	    /* Strange, but must guard against that!  */
	    break;

	/* Install the corresp atom for this "ccode": */
	i = LOAD_INT (spec, i, ccode);
	cause += clen;
	coff  += clen;
	s ++;
    }
    i = LOAD_NIL (spec, i);
    i = LOAD_LIST(spec, i, s+1);
    return i;
}

/*
** Parsing of SCTP notification events. NB: they are NOT ancillary data: they
** are sent IN PLACE OF, not in conjunction with, the normal data:
*/
static int sctp_parse_async_event
      (ErlDrvTermData * spec, int i,    int ok_pos,
       ErlDrvTermData   error_atom,     inet_descriptor* desc,
       ErlDrvBinary   * bin,  int offs, int sz)
{
    char* body			   = bin->orig_bytes + offs;
    union sctp_notification * nptr = (union sctp_notification *) body;

    switch (nptr->sn_header.sn_type)
    {
	case SCTP_ASSOC_CHANGE:
	{   /* {sctp_assoc_change,
		State		: Atom(),
		Error		: Atom(),
		OutBoundStreams : Int(),
		InBoundStreams  : Int(),
		AssocID		: Int(),
		// AbortCauses	: [Atom()]   // NOT YET IMPLEMENTED
	       }
	    */
	    struct sctp_assoc_change* sptr = &(nptr->sn_assoc_change);
	    ASSERT(sptr->sac_length <= sz);  /* No buffer overrun */

	    i = LOAD_ATOM (spec, i, am_sctp_assoc_change);

	    switch (sptr->sac_state)
	    {
	    case SCTP_COMM_UP:
		i = LOAD_ATOM (spec, i, am_comm_up);
		break;
	    case SCTP_COMM_LOST:
		i = LOAD_ATOM (spec, i, am_comm_lost);
		break;
	    case SCTP_RESTART:
		i = LOAD_ATOM (spec, i, am_restart);
		break;
	    case SCTP_SHUTDOWN_COMP:
		i = LOAD_ATOM (spec, i, am_shutdown_comp);
		break;
	    case SCTP_CANT_STR_ASSOC:
		i = LOAD_ATOM (spec, i, am_cant_assoc);
		break;
	    default:
		ASSERT(0);
	    }
	    i = LOAD_INT (spec, i, sptr->sac_error);
	    i = LOAD_INT (spec, i, sptr->sac_outbound_streams);
	    i = LOAD_INT (spec, i, sptr->sac_inbound_streams);
	    i = LOAD_INT (spec, i, sptr->sac_assoc_id);

	    /* The ABORT chunk may or may not be present at the end, depending
	       on whether there was really an ABORT.  In the Linux Kernel SCTP
	       implementation, this chunk is not delivered anyway, so we leave
	       it out. Just close up the tuple:
	    */
	    i = LOAD_TUPLE (spec, i, 6);
	    break;
	}

	case SCTP_PEER_ADDR_CHANGE:
	{   /* {sctp_paddr_change,
		AffectedAddr	: String(),
		State		: Atom(),
		Error		: Atom(),
		AssocID		: Int()
	       }
	    */
	    struct sctp_paddr_change* sptr = &(nptr->sn_paddr_change);
	    ASSERT(sptr->spc_length <= sz);  /* No buffer overrun */

	    i = LOAD_ATOM	(spec, i, am_sctp_paddr_change);
	    i = load_inet_get_address(spec, i, desc, &sptr->spc_aaddr);

	    switch (sptr->spc_state)
	    {
	    case SCTP_ADDR_AVAILABLE:
		i = LOAD_ATOM (spec, i, am_addr_available);
		break;
	    case SCTP_ADDR_UNREACHABLE:
		i = LOAD_ATOM (spec, i, am_addr_unreachable);
		break;
	    case SCTP_ADDR_REMOVED:
		i = LOAD_ATOM (spec, i, am_addr_removed);
		break;
	    case SCTP_ADDR_ADDED:
		i = LOAD_ATOM (spec, i, am_addr_added);
		break;
	    case SCTP_ADDR_MADE_PRIM:
		i = LOAD_ATOM (spec, i, am_addr_made_prim);
		break;
#if HAVE_DECL_SCTP_ADDR_CONFIRMED
	    case SCTP_ADDR_CONFIRMED:
		i = LOAD_ATOM (spec, i, am_addr_confirmed);
		break;
#endif
	    default:
		ASSERT(0);
	    }
	    i = LOAD_INT   (spec, i, sptr->spc_error);
	    i = LOAD_INT   (spec, i, sptr->spc_assoc_id);
	    i = LOAD_TUPLE (spec, i, 5);
	    break;
	}

	case SCTP_REMOTE_ERROR:
	{   /* This is an error condition, so we return an error term
	       {sctp_remote_error,
		Error		: Int(),
		AssocID		: Int(),
		RemoteCauses	: [Atom()] // Remote Error flags
	       }
	    */
	    char *chunk;
	    int chlen;
	    struct sctp_remote_error * sptr = &(nptr->sn_remote_error);
	    ASSERT(sptr->sre_length <= sz);   /* No buffer overrun */

	    /* Over-write the prev part of the response with an error: */
	    (void)LOAD_ATOM(spec, ok_pos, error_atom);

	    /* Continue from the curr pos: */
	    i = LOAD_ATOM  (spec, i, am_sctp_remote_error);

	    i = LOAD_INT   (spec, i, sock_ntohs(sptr->sre_error));
	    i = LOAD_INT   (spec, i, sptr->sre_assoc_id);

#	    ifdef HAVE_STRUCT_SCTP_REMOTE_ERROR_SRE_DATA
	    chunk = (char*) (&(sptr->sre_data));
#	    else
	    chunk = ((char*) &(sptr->sre_assoc_id))
		+ sizeof(sptr->sre_assoc_id);
#	    endif
	    chlen = sptr->sre_length  - (chunk - (char *)sptr);
	    i = sctp_parse_error_chunk(spec, i, chunk, chlen);

	    i = LOAD_TUPLE (spec, i, 4);
	    /* The {error, {...}} will be closed by the caller */
	    break;
	}

	case SCTP_SEND_FAILED:
	{   /* {sctp_send_failed,
		DataSent	: Atom()	// true or false
		Error		: Atom(),
		OrigInfo	: Tuple(),
		AssocID		: Int(),
		OrigData	: Binary()
	       }
	       This is also an ERROR condition -- overwrite the 'ok':
	    */
	    char *chunk;
	    int chlen, choff;
	    struct sctp_send_failed * sptr = &(nptr->sn_send_failed);
	    ASSERT(sptr->ssf_length <= sz);	/* No buffer overrun */

	    /* Over-write 'ok' with 'error', continue from curr "i": */
	    (void)LOAD_ATOM(spec, ok_pos, error_atom);

	    i = LOAD_ATOM  (spec, i, am_sctp_send_failed);
	    switch (sptr->ssf_flags) {
	    case SCTP_DATA_SENT:
		i = LOAD_ATOM (spec, i, am_true);
		break;
	    case SCTP_DATA_UNSENT:
		i = LOAD_ATOM (spec, i, am_false);
		break;
	    default:
		ASSERT(0);
	    }
	    i = LOAD_INT      (spec, i, sptr->ssf_error);
	    /* Now parse the orig SCTP_SNDRCV info */
	    i = sctp_parse_sndrcvinfo (spec, i, &sptr->ssf_info);
	    i = LOAD_ASSOC_ID (spec, i, sptr->ssf_assoc_id);

	    /* Load the orig data chunk, as an unparsed binary. Note that
	       in LOAD_BINARY below, we must specify the offset wrt bin->
	       orig_bytes. In Solaris 10, we don't have ssf_data:
	    */
#	    ifdef HAVE_STRUCT_SCTP_SEND_FAILED_SSF_DATA
	    chunk = (char*) (&(sptr->ssf_data));
#	    else
	    chunk = ((char*) &(sptr->ssf_assoc_id))
		+ sizeof(sptr->ssf_assoc_id);
#	    endif
	    chlen = sptr->ssf_length - (chunk - (char*) sptr);
	    choff = chunk - bin->orig_bytes;

	    i = LOAD_BINARY(spec, i, bin, choff, chlen);
	    i = LOAD_TUPLE (spec, i, 6);
	    /* The {error, {...}} tuple is not yet closed */
	    break;
	}

	case SCTP_SHUTDOWN_EVENT:
	{   /* {sctp_shutdown_event,
		AssocID		: Int()
	       }
	    */
	    struct sctp_shutdown_event * sptr = &(nptr->sn_shutdown_event);

	    ASSERT (sptr->sse_length == sizeof(struct sctp_shutdown_event) &&
		    sptr->sse_length <= sz);	/* No buffer overrun */

	    i = LOAD_ATOM  (spec, i, am_sctp_shutdown_event);
	    i = LOAD_INT   (spec, i, sptr->sse_assoc_id);
	    i = LOAD_TUPLE (spec, i, 2);
	    break;
	}

	case SCTP_ADAPTATION_INDICATION:
	{   /* {sctp_adaptation_event,
		Indication	: Atom(),
		AssocID		: Int()
	       }
	    */
	    struct sctp_adaptation_event * sptr =
		&(nptr->sn_adaptation_event);
	    ASSERT (sptr->sai_length == sizeof(struct sctp_adaptation_event)
		    && sptr->sai_length <= sz);	/* No buffer overrun */

	    i = LOAD_ATOM  (spec, i, am_sctp_adaptation_event);
	    i = LOAD_INT   (spec, i, sock_ntohl(sptr->sai_adaptation_ind));
	    i = LOAD_INT   (spec, i, sptr->sai_assoc_id);
	    i = LOAD_TUPLE (spec, i, 3);
	    break;
	}

	case SCTP_PARTIAL_DELIVERY_EVENT:
	{   /* It is not clear  whether this event  is sent  to the sender
		(when the receiver gets only a part of a message),   or to
		the receiver itself.  In any case, we do not support partial
		delivery of msgs in this implementation, so this is an error
		condition:
		{sctp_pdapi_event, sctp_partial_delivery_aborted, AssocID}:
	    */
	    struct sctp_pdapi_event * sptr;
	    (void) LOAD_ATOM  (spec, ok_pos, error_atom);

	    sptr = &(nptr->sn_pdapi_event);
	    ASSERT (sptr->pdapi_length == sizeof(struct sctp_pdapi_event) &&
		    sptr->pdapi_length <= sz);  /* No buffer overrun */

	    i = LOAD_ATOM  (spec, i, am_sctp_pdapi_event);

	    /* Currently, there is only one indication possible: */
	    ASSERT (sptr->pdapi_indication == SCTP_PARTIAL_DELIVERY_ABORTED);

	    i = LOAD_ATOM  (spec, i, am_partial_delivery_aborted);
	    i = LOAD_INT   (spec, i, sptr->pdapi_assoc_id);
	    i = LOAD_TUPLE (spec, i, 3);
	    /* The {error, {...}} tuple is not yet closed */
	    break;
	}

	/* XXX: No more supported SCTP Event types. The standard also provides
	   SCTP_AUTHENTICATION_EVENT, but it is not implemented in the Linux
	   kernel, hence not supported here either. It is not possible to
	   request delivery of such events in this implementation, so they
	   cannot occur:
	*/
	default:   ASSERT(0);
    }
    return i;
}
#endif  /* HAVE_SCTP */

#ifndef __WIN32__
static int udp_parse_ancillary_data(ErlDrvTermData *spec, int i,
                                struct msghdr *mptr) {
    struct cmsghdr *cmsg;
    int n;

    n = 0;
    for (cmsg = CMSG_FIRSTHDR(mptr);
         cmsg != NULL;
         cmsg = CMSG_NXTHDR(mptr, cmsg)) {
        i = parse_ancillary_data_item(spec, i, cmsg, &n);
    }
    i = LOAD_NIL(spec, i);
    return LOAD_LIST(spec, i, n+1);
}

static int compile_ancillary_data(struct msghdr *mhdr,
                                  char *ptr, ErlDrvSizeT anc_len) {
    struct cmsghdr *cmsg;
    size_t controllen = 0;
    cmsg = CMSG_FIRSTHDR(mhdr);
    for (;;) {
        if (anc_len == 0) {
            /* End of options to compile */
            mhdr->msg_controllen = controllen;
            return 0;
        }
        if (cmsg == NULL) {
            /* End of destination before end of options */
            return 1;
        }

#define COMPILE_ANCILLARY_DATA_ITEM(Level, Opt, Type, Get, Size)        \
        do {                                                            \
            if (anc_len < (Size)) return 1;                             \
            sys_memset(cmsg, '\0', CMSG_SPACE(sizeof(Type)));           \
            cmsg->cmsg_level = Level;                                   \
            cmsg->cmsg_type = Opt;                                      \
            cmsg->cmsg_len = CMSG_LEN(sizeof(Type));                    \
            *((Type *) CMSG_DATA(cmsg)) = Get(ptr);                     \
            controllen += CMSG_SPACE(sizeof(Type));                     \
            cmsg = CMSG_NXTHDR(mhdr, cmsg);                             \
            ptr += 4;                                                   \
            anc_len -= 4;                                               \
        } while (0)
#define SIZEOF_ANCILLARY_DATA (2 * CMSG_SPACE(sizeof(int)))
        /* (IP_TOS | IPV6_TCLASS) + IP_TTL */

        switch (anc_len--, *ptr++) {
        case INET_OPT_TOS: {
#if defined(IPPROTO_IP) && defined(IP_TOS)
            COMPILE_ANCILLARY_DATA_ITEM(IPPROTO_IP, IP_TOS, int, get_int32, 4);
#else
            return 1; /* Socket option not implemented */
#endif
            break;
        }
        case INET_OPT_TTL: {
#if defined(IPPROTO_IP) && defined(IP_TTL)
            COMPILE_ANCILLARY_DATA_ITEM(IPPROTO_IP, IP_TTL, int, get_int32, 4);
#else
            return 1; /* Socket option not implemented */
#endif
            break;
        }
        case INET_OPT_TCLASS: {
#if defined(IPPROTO_IPV6) && defined(IPV6_TCLASS)
            COMPILE_ANCILLARY_DATA_ITEM(IPPROTO_IPV6, IPV6_TCLASS, int, get_int32, 4);
#else
            return 1; /* Socket option not implemented */
#endif
            break;
        }
        default:
            /* Unknow socket option */
            return 1;
        }
#undef COMPILE_ANCILLARY_DATA_ITEM
    }
}
#endif /* ifndef __WIN32__ */

/*
** passive mode reply:
** for UDP:
**        {inet_async, S, Ref, {ok, Data=[H1,...,Hsz | BinData]}}
** or (in the list mode)
**	  {inet_async, S, Ref, {ok, Data=[H1,...,Hsz]}}
**
** for SCTP:
**	  {inet_async, S, Ref, {ok, {[H1,...,HSz], [AncilData], Data_OR_Event}}}
** where  each AncilDatum:Tuple();
**	  Data:List() or Binary(), but if List(), then without the Addr part,
**				   which is moved in front;
**	  Event:Tuple();
** or
** 	  {inet_async, S, Ref, {error, {[H1,...,HSz], [AncilData], ErrorTerm}}}
**
** Cf: the output of send_async_error() is
**	  {inet_async, S, Ref, {error, Cause:Atom()}}
*/
static int
inet_async_binary_data
	(inet_descriptor* desc, unsigned  int phsz,
	 ErlDrvBinary   * bin,  int offs, int len, void *mp)
{
    unsigned int hsz = desc->hsz + phsz;
    ErlDrvTermData spec [PACKET_ERL_DRV_TERM_DATA_LEN];
    ErlDrvTermData caller = desc->caller;
    int aid;
    int req;
    int i = 0;
#ifdef HAVE_SCTP
    int ok_pos;
#endif

    DEBUGF(("inet_async_binary_data(%ld): offs=%d, len=%d\r\n",
	    (long)desc->port, offs, len));

    if (deq_async(desc, &aid, &caller, &req) < 0)
	return -1;

    i = LOAD_ATOM(spec, i, am_inet_async);	/* 'inet_async' */
    i = LOAD_PORT(spec, i, desc->dport);	/* S		*/
    i = LOAD_INT (spec, i, aid);		/* Ref		*/

#ifdef HAVE_SCTP
    /* Need to memoise the position of the 'ok' atom written, as it may
       later be overridden by an 'error': */
    ok_pos = i;
#endif
    i = LOAD_ATOM(spec, i, am_ok);

#ifdef HAVE_SCTP
    if (IS_SCTP(desc))
    {	/* For SCTP we always have desc->hsz==0 (i.e., no application-level
	   headers are used), so hsz==phsz (see above): */
	int sz;
        struct msghdr *mptr;

        mptr = mp;
	ASSERT (hsz == phsz && hsz != 0);
	sz = len - hsz;  /* Size of the msg data proper, w/o the addr */

	/* We always put the Addr as a list in front */
	i = LOAD_STRING(spec, i, bin->orig_bytes+offs, hsz);

	/* Put in the list (possibly empty) of Ancillary Data: */
	i = sctp_parse_ancillary_data (spec, i, mptr);

	/* Then: Data or Event (Notification)? */
	if (mptr->msg_flags & MSG_NOTIFICATION)
	    /* This is an Event, parse it. It may indicate a normal or an error
	       condition; in the latter case,   the 'ok' above is overridden by
	       an 'error', and the Event we receive contains the error term: */
	    i = sctp_parse_async_event
		(spec, i, ok_pos, am_error, desc, bin, offs+hsz, sz);
        else
    	    /* This is SCTP data, not a notification event.   The data can be
	       returned as a List or as a Binary, similar to the generic case:
	    */
	    if (desc->mode == INET_MODE_LIST)
		/* INET_MODE_LIST   => [H1,H2,...Hn], addr and data together,
		   butthe Addr has already been parsed, so start at offs+hsz:
		*/
		i = LOAD_STRING(spec, i, bin->orig_bytes+offs+hsz, sz);
	    else
	    	/* INET_MODE_BINARY => Binary */
		i = LOAD_BINARY(spec, i, bin, offs+hsz, sz);

	/* Close up the {[H1,...,HSz], [AncilData], Event_OR_Data} tuple. This
	   is valid even in the case when Event is a error notification:  */
	i = LOAD_TUPLE (spec, i, 3);
    }
    else
#endif  /* HAVE_SCTP */
    {
        /* Generic case. Both Addr and Data
         * (or a single list of them together) are returned: */

        if ((desc->mode == INET_MODE_LIST) || (hsz > len)) {
            /* INET_MODE_LIST => [H1,H2,...Hn] */
            i = LOAD_STRING(spec, i, bin->orig_bytes+offs, len);
        }
        else {
            /* INET_MODE_BINARY => [H1,H2,...HSz | Binary] or [Binary]: */
            int sz = len - hsz;
            i = LOAD_BINARY(spec, i, bin, offs+hsz, sz);
            if (hsz > 0)
                i = LOAD_STRING_CONS(spec, i, bin->orig_bytes+offs, hsz);
        }

#ifndef __WIN32__
        if (mp) {
            /* We got ancillary data from an UDP recvmsg.
             * Insert an additional tuple level {[F|AddrData],AncData} */
            i = udp_parse_ancillary_data(spec, i, (struct msghdr*)mp);
            i = LOAD_TUPLE(spec, i, 2);
        }
#endif
    }

    /* Close up the {ok, ...} or {error, ...} tuple: */
    i = LOAD_TUPLE(spec, i, 2);

    /* Close up the outer {inet_async, S, Ref, {ok|error, ...}} tuple: */
    i = LOAD_TUPLE(spec, i, 4);

    ASSERT(i <= PACKET_ERL_DRV_TERM_DATA_LEN);
    desc->caller = 0;
    return erl_drv_send_term(desc->dport, caller, spec, i);
}

/*
** active mode message:
**        {tcp, S, [H1,...Hsz | Data]}
*/
static int tcp_message(inet_descriptor* desc, const char* buf, int len)
{
    unsigned int hsz = desc->hsz;
    ErlDrvTermData spec[20];
    int i = 0;

    DEBUGF(("tcp_message(%ld): len = %d\r\n", (long)desc->port, len));
    /* XXX fprintf(stderr,"tcp_message send.\r\n"); */

    i = LOAD_ATOM(spec, i, am_tcp);
    i = LOAD_PORT(spec, i, desc->dport);

    if ((desc->mode == INET_MODE_LIST) || (hsz > len)) {
	i = LOAD_STRING(spec, i, buf, len); /* => [H1,H2,...Hn] */
	i = LOAD_TUPLE(spec, i, 3);
	ASSERT(i <= 20);
	return erl_drv_output_term(desc->dport, spec, i);
    }
    else {
	/* INET_MODE_BINARY => [H1,H2,...HSz | Binary] */
	int sz = len - hsz;
	int code;

	i = LOAD_BUF2BINARY(spec, i, buf+hsz, sz);
	if (hsz > 0)
	    i = LOAD_STRING_CONS(spec, i, buf, hsz);
	i = LOAD_TUPLE(spec, i, 3);
	ASSERT(i <= 20);
	code = erl_drv_output_term(desc->dport, spec, i);
	return code;
    }
}

/*
** active mode message:
**        {tcp, S, [H1,...Hsz | Data]}
*/
static int
tcp_binary_message(inet_descriptor* desc, ErlDrvBinary* bin, int offs, int len)
{
    unsigned int hsz = desc->hsz;
    ErlDrvTermData spec[20];
    int i = 0;

    DEBUGF(("tcp_binary_message(%ld): len = %d\r\n", (long)desc->port, len));

    i = LOAD_ATOM(spec, i, am_tcp);
    i = LOAD_PORT(spec, i, desc->dport);

    if ((desc->mode == INET_MODE_LIST) || (hsz > len)) {
	/* INET_MODE_LIST => [H1,H2,...Hn] */
	i = LOAD_STRING(spec, i, bin->orig_bytes+offs, len);
    }
    else {
	/* INET_MODE_BINARY => [H1,H2,...HSz | Binary] */
	int sz = len - hsz;

	i = LOAD_BINARY(spec, i, bin, offs+hsz, sz);
	if (hsz > 0)
	    i = LOAD_STRING_CONS(spec, i, bin->orig_bytes+offs, hsz);
    }
    i = LOAD_TUPLE(spec, i, 3);
    ASSERT(i <= 20);
    return erl_drv_output_term(desc->dport, spec, i);
}

/*
** send:  active mode  {tcp_closed, S}
*/
static int tcp_closed_message(tcp_descriptor* desc)
{
    ErlDrvTermData spec[6];
    int i = 0;

    DEBUGF(("tcp_closed_message(%ld):\r\n", (long)desc->inet.port));
    if (!(desc->tcp_add_flags & TCP_ADDF_CLOSE_SENT)) {
	desc->tcp_add_flags |= TCP_ADDF_CLOSE_SENT;

	i = LOAD_ATOM(spec, i, am_tcp_closed);
	i = LOAD_PORT(spec, i, desc->inet.dport);
	i = LOAD_TUPLE(spec, i, 2);
	ASSERT(i <= 6);
	return erl_drv_output_term(desc->inet.dport, spec, i);
    }
    return 0;
}

/*
** send active message {tcp_error, S, Error}
*/
static int tcp_error_message(tcp_descriptor* desc, int err)
{
    ErlDrvTermData spec[8];
    ErlDrvTermData am_err = error_atom(err);
    int i = 0;

    DEBUGF(("tcp_error_message(%ld): %d\r\n", (long)desc->inet.port, err));

    i = LOAD_ATOM(spec, i, am_tcp_error);
    i = LOAD_PORT(spec, i, desc->inet.dport);
    i = LOAD_ATOM(spec, i, am_err);
    i = LOAD_TUPLE(spec, i, 3);
    ASSERT(i <= 8);
    return erl_drv_output_term(desc->inet.dport, spec, i);
}

#ifdef HAVE_UDP
/*
** active mode message:
**    {udp,  S, IP, Port, [H1,...Hsz | Data]} or
**    {sctp, S, IP, Port, {[AncilData],  Event_or_Data}}
** where
** 	  [H1,...,HSz] are msg headers (without IP/Port, UDP only),
**    [AddrLen, H2,...,HSz] are msg headers for UDP AF_UNIX only
**	  Data  : List() | Binary()
*/
static int packet_binary_message(inet_descriptor* desc,
                                 ErlDrvBinary* bin, int offs, int len,
                                 void *mp)
{
    unsigned int hsz = desc->hsz;
    ErlDrvTermData spec [PACKET_ERL_DRV_TERM_DATA_LEN];
    int i = 0;
    int alen;
    char* data = bin->orig_bytes+offs;

    DEBUGF(("packet_binary_message(%ld): len = %d\r\n",
	   (long)desc->port, len));
#   ifdef HAVE_SCTP
    i = LOAD_ATOM(spec, i, IS_SCTP(desc) ? am_sctp : am_udp); /* UDP|SCTP */
#   else
    i = LOAD_ATOM(spec, i, am_udp );			      /* UDP only */
#   endif
    i = LOAD_PORT(spec, i, desc->dport);   		      /* S	  */

    alen = addrlen(data);
    i = load_address(spec, i, data);     /* IP,Port | Family,Addr */

    offs += alen;
    len  -= alen;

#   ifdef HAVE_SCTP
    if (!IS_SCTP(desc))
#   endif
    {
#ifndef __WIN32__
        if (mp) i = udp_parse_ancillary_data(spec, i, (struct msghdr*)mp);
#endif
        /* We got ancillary data from an UDP recvmsg.
         * Insert an additional tuple level {AncData,[F|AddrData]}
         */
	if ((desc->mode == INET_MODE_LIST) || (hsz > len))
	    /* INET_MODE_LIST, or only headers => [H1,H2,...Hn] */
	    i = LOAD_STRING(spec, i, bin->orig_bytes+offs, len);
	else {
	    /* INET_MODE_BINARY => [H1,H2,...HSz | Binary]	*/
	    int sz = len - hsz;

	    i = LOAD_BINARY(spec, i, bin, offs+hsz, sz);
	    if (hsz > 0)
		i = LOAD_STRING_CONS(spec, i, bin->orig_bytes+offs, hsz);
	}
        /* Close up the outer 5-or-6-tuple */
#ifndef __WIN32__
        if (mp) i = LOAD_TUPLE(spec, i, 6);
        else
#endif
            i = LOAD_TUPLE(spec, i, 5);
    }
#   ifdef HAVE_SCTP
    else
    {
        struct msghdr *mptr;

        mptr = mp;
	/* For SCTP we always have desc->hsz==0 (i.e., no application-level
	   headers are used): */
	ASSERT(hsz == 0);

	/* Put in the list (possibly empty) of Ancillary Data: */
	i = sctp_parse_ancillary_data (spec, i, mptr);

	/* Then: Data or Event (Notification)? */
	if (mptr->msg_flags & MSG_NOTIFICATION)
	    /* This is an Event, parse it. It may indicate a normal or an error
	       condition; in the latter case,  the initial 'sctp' atom is over-
	       ridden by 'sctp_error',   and the Event we receive contains the
	       error term: */
	    i = sctp_parse_async_event
		(spec, i, 0, am_sctp_error, desc, bin, offs, len);
        else
    	    /* This is SCTP data, not a notification event.   The data can be
	       returned as a List or as a Binary, similar to the generic case:
	    */
	    if (desc->mode == INET_MODE_LIST)
		/* INET_MODE_LIST   => [H1,H2,...Hn], addr and data together,
		   but the Addr has already been parsed, so start at offs:
		*/
		i = LOAD_STRING(spec, i, bin->orig_bytes+offs, len);
	    else
	    	/* INET_MODE_BINARY => Binary */
		i = LOAD_BINARY(spec, i, bin, offs, len);

	/* Close up the {[AncilData], Event_OR_Data} tuple: */
	i = LOAD_TUPLE (spec, i, 2);
        /* Close up the outer 5-tuple: */
        i = LOAD_TUPLE(spec, i, 5);
    }
#   endif /* HAVE_SCTP */

    ASSERT(i <= PACKET_ERL_DRV_TERM_DATA_LEN);
    return erl_drv_output_term(desc->dport, spec, i);
}
#endif

/*
** active mode message: send active-to-passive transition message
**        {tcp_passive, S} or
**        {udp_passive, S} or
**        {sctp_passive, S}
*/
 static int packet_passive_message(inet_descriptor* desc)
 {
     ErlDrvTermData spec[6];
     int i = 0;

     DEBUGF(("packet_passive_message(%ld):\r\n", (long)desc->port));

#if !defined(HAVE_UDP) && !defined(HAVE_SCTP)
     i = LOAD_ATOM(spec, i, am_tcp_passive);
#else
     if (desc->sprotocol == IPPROTO_TCP)
         i = LOAD_ATOM(spec, i, am_tcp_passive);
     else {
#ifdef HAVE_SCTP
         i = LOAD_ATOM(spec, i, IS_SCTP(desc) ? am_sctp_passive : am_udp_passive);
#else
         i = LOAD_ATOM(spec, i, am_udp_passive);
#endif
     }
#endif
     i = LOAD_PORT(spec, i, desc->dport);
     i = LOAD_TUPLE(spec, i, 2);
     ASSERT(i <= 6);
     return erl_drv_output_term(desc->dport, spec, i);
 }

#ifdef HAVE_UDP
/*
** send active message {udp_error|sctp_error, S, Error}
*/
static int packet_error_message(udp_descriptor* udesc, int err)
{
    inet_descriptor* desc = INETP(udesc);
    ErlDrvTermData spec[2*LOAD_ATOM_CNT + LOAD_PORT_CNT + LOAD_TUPLE_CNT];
    ErlDrvTermData am_err = error_atom(err);
    int i = 0;

    DEBUGF(("packet_error_message(%ld): %d\r\n",
	   (long)desc->port, err));

#   ifdef HAVE_SCTP
    if (IS_SCTP(desc) )
    	i = LOAD_ATOM(spec, i, am_sctp_error);
    else
#   endif
	i = LOAD_ATOM(spec, i, am_udp_error);

    i = LOAD_PORT(spec, i, desc->dport);
    i = LOAD_ATOM(spec, i, am_err);
    i = LOAD_TUPLE(spec, i, 3);
    ASSERT(i == sizeof(spec)/sizeof(*spec));
    return erl_drv_output_term(desc->dport, spec, i);
}
#endif

/*
** active=TRUE:
**  (NOTE! distribution MUST use active=TRUE, deliver=PORT)
**       deliver=PORT  {S, {data, [H1,..Hsz | Data]}}
**       deliver=TERM  {tcp, S, [H1..Hsz | Data]}
**
** active=FALSE:
**       {async, S, Ref, {ok,[H1,...Hsz | Data]}}
*/
static int tcp_reply_data(tcp_descriptor* desc, char* buf, int len)
{
    int code;
    const char* body = buf;
    int bodylen = len;

    packet_get_body(desc->inet.htype, &body, &bodylen);

    if (desc->inet.deliver == INET_DELIVER_PORT) {
        code = inet_port_data(INETP(desc), body, bodylen);
    }
    else if ((code=packet_parse(desc->inet.htype, buf, len,
                                &desc->http_state, &packet_callbacks,
                                desc)) == 0) {
        /* No body parsing, return raw binary */
        if (desc->inet.active == INET_PASSIVE)
            return inet_async_data(INETP(desc), body, bodylen);
        else
            code = tcp_message(INETP(desc), body, bodylen);
    }

    if (code < 0)
	return code;
    INET_CHECK_ACTIVE_TO_PASSIVE(INETP(desc));
    return code;
}

static int
tcp_reply_binary_data(tcp_descriptor* desc, ErlDrvBinary* bin, int offs, int len)
{
    int code;
    const char* buf = bin->orig_bytes + offs;
    const char* body = buf;
    int bodylen = len;

    packet_get_body(desc->inet.htype, &body, &bodylen);
    offs = body - bin->orig_bytes; /* body offset now */

    if (desc->inet.deliver == INET_DELIVER_PORT)
        code = inet_port_binary_data(INETP(desc), bin, offs, bodylen);
    else if ((code=packet_parse(desc->inet.htype, buf, len, &desc->http_state,
                                     &packet_callbacks,desc)) == 0) {
        /* No body parsing, return raw data */
        if (desc->inet.active == INET_PASSIVE)
            return inet_async_binary_data(INETP(desc), 0, bin, offs, bodylen, NULL);
        else
            code = tcp_binary_message(INETP(desc), bin, offs, bodylen);
    }
    if (code < 0)
	return code;
    INET_CHECK_ACTIVE_TO_PASSIVE(INETP(desc));
    return code;
}

#ifdef HAVE_UDP
static int
packet_reply_binary_data(inet_descriptor* desc, unsigned  int hsz,
			 ErlDrvBinary   * bin,  int offs, int len,
                         void *mp)
{
    int code;

    if (desc->active == INET_PASSIVE)
	/* "inet" is actually for both UDP and SCTP, as well as TCP! */
	return inet_async_binary_data(desc, hsz, bin, offs, len, mp);
    else
    {	/* INET_ACTIVE or INET_ONCE: */
	if (desc->deliver == INET_DELIVER_PORT)
	    code = inet_port_binary_data(desc, bin, offs, len);
	else
	    code = packet_binary_message(desc, bin, offs, len, mp);
	if (code < 0)
	    return code;
        INET_CHECK_ACTIVE_TO_PASSIVE(desc);
	return code;
    }
}
#endif

/* ----------------------------------------------------------------------------

   INET

---------------------------------------------------------------------------- */

static int
sock_init(void) /* May be called multiple times. */
{
#ifdef __WIN32__
    WORD wVersionRequested;
    WSADATA wsaData;
    static int res = -1; /* res < 0 == initialization never attempted */

    if (res >= 0)
	return res;

    wVersionRequested = MAKEWORD(2,0);
    if (WSAStartup(wVersionRequested, &wsaData) != 0)
	goto error;

    if ((LOBYTE(wsaData.wVersion) != 2) || (HIBYTE(wsaData.wVersion) != 0))
	goto error;

    find_dynamic_functions();

    return res = 1;

 error:

    WSACleanup();
    return res = 0;
#else
    return 1;
#endif
}

#ifdef HAVE_SCTP
static void inet_init_sctp(void) {
    INIT_ATOM(sctp);
    INIT_ATOM(sctp_passive);
    INIT_ATOM(sctp_error);
    INIT_ATOM(true);
    INIT_ATOM(false);
    INIT_ATOM(buffer);
    INIT_ATOM(mode);
    INIT_ATOM(list);
    INIT_ATOM(binary);
    INIT_ATOM(active);
    INIT_ATOM(once);
    INIT_ATOM(multi);
    INIT_ATOM(buffer);
    INIT_ATOM(linger);
    INIT_ATOM(recbuf);
    INIT_ATOM(sndbuf);
    INIT_ATOM(reuseaddr);
    INIT_ATOM(dontroute);
    INIT_ATOM(priority);
    INIT_ATOM(recvtos);
    INIT_ATOM(recvtclass);
    INIT_ATOM(recvttl);
    INIT_ATOM(ipv6_v6only);
    INIT_ATOM(netns);
    INIT_ATOM(bind_to_device);

    /* Option names */
    INIT_ATOM(sctp_rtoinfo);
    INIT_ATOM(sctp_associnfo);
    INIT_ATOM(sctp_initmsg);
    INIT_ATOM(sctp_autoclose);
    INIT_ATOM(sctp_nodelay);
    INIT_ATOM(sctp_disable_fragments);
    INIT_ATOM(sctp_i_want_mapped_v4_addr);
    INIT_ATOM(sctp_maxseg);
    INIT_ATOM(sctp_set_peer_primary_addr);
    INIT_ATOM(sctp_primary_addr);
    INIT_ATOM(sctp_adaptation_layer);
    INIT_ATOM(sctp_peer_addr_params);
    INIT_ATOM(sctp_default_send_param);
    INIT_ATOM(sctp_events);
    INIT_ATOM(sctp_delayed_ack_time);
    INIT_ATOM(sctp_status);
    INIT_ATOM(sctp_get_peer_addr_info);

    /* Record names */
    INIT_ATOM(sctp_sndrcvinfo);
    INIT_ATOM(sctp_assoc_change);
    INIT_ATOM(sctp_paddr_change);
    INIT_ATOM(sctp_remote_error);
    INIT_ATOM(sctp_send_failed);
    INIT_ATOM(sctp_shutdown_event);
    INIT_ATOM(sctp_adaptation_event);
    INIT_ATOM(sctp_pdapi_event);
    INIT_ATOM(sctp_assocparams);
    INIT_ATOM(sctp_prim);
    INIT_ATOM(sctp_setpeerprim);
    INIT_ATOM(sctp_setadaptation);
    INIT_ATOM(sctp_paddrparams);
    INIT_ATOM(sctp_event_subscribe);
    INIT_ATOM(sctp_assoc_value);
    INIT_ATOM(sctp_paddrinfo);

    /* For #sctp_sndrcvinfo{}: */
    INIT_ATOM(unordered);
    INIT_ATOM(addr_over);
    INIT_ATOM(abort);
    INIT_ATOM(eof);

    /* For #sctp_assoc_change{}: */
    INIT_ATOM(comm_up);
    INIT_ATOM(comm_lost);
    INIT_ATOM(restart);
    INIT_ATOM(shutdown_comp);
    INIT_ATOM(cant_assoc);

    /* For #sctp_paddr_change{}: */
    INIT_ATOM(addr_available);
    INIT_ATOM(addr_unreachable);
    INIT_ATOM(addr_removed);
    INIT_ATOM(addr_added);
    INIT_ATOM(addr_made_prim);
    INIT_ATOM(addr_confirmed);

    INIT_ATOM(short_recv);
    INIT_ATOM(wrong_anc_data);

    /* For #sctp_pdap_event{}: */
    INIT_ATOM(partial_delivery_aborted);

    /* For #sctp_paddrparams{}: */
    INIT_ATOM(hb_enable);
    INIT_ATOM(hb_disable);
    INIT_ATOM(hb_demand);
    INIT_ATOM(pmtud_enable);
    INIT_ATOM(pmtud_disable);
    INIT_ATOM(sackdelay_enable);
    INIT_ATOM(sackdelay_disable);

    /* For #sctp_paddrinfo{}: */
    INIT_ATOM(active);
    INIT_ATOM(inactive);
#    if HAVE_DECL_SCTP_UNCONFIRMED
    INIT_ATOM(unconfirmed);
#    endif

    /* For #sctp_status{}: */
#    if HAVE_DECL_SCTP_EMPTY
    INIT_ATOM(empty);
#    endif
#    if HAVE_DECL_SCTP_BOUND
    INIT_ATOM(bound);
#    endif
#    if HAVE_DECL_SCTP_LISTEN
    INIT_ATOM(listen);
#    endif
    INIT_ATOM(cookie_wait);
    INIT_ATOM(cookie_echoed);
    INIT_ATOM(established);
    INIT_ATOM(shutdown_pending);
    INIT_ATOM(shutdown_sent);
    INIT_ATOM(shutdown_received);
    INIT_ATOM(shutdown_ack_sent);
}
#endif /* HAVE_SCTP */

static int inet_init()
{
    if (!sock_init())
	goto error;

    if (0 != erl_drv_tsd_key_create("inet_buffer_stack_key", &buffer_stack_key))
	goto error;

    ERTS_CT_ASSERT(sizeof(struct in_addr) == 4);
#   if defined(HAVE_IN6) && defined(AF_INET6)
    ERTS_CT_ASSERT(sizeof(struct in6_addr) == 16);
#   endif

    INIT_ATOM(ok);
    INIT_ATOM(undefined);
    INIT_ATOM(unspec);
    INIT_ATOM(tcp);
#ifdef HAVE_UDP
    INIT_ATOM(udp);
#endif
    INIT_ATOM(error);
    INIT_ATOM(einval);
    INIT_ATOM(inet_async);
    INIT_ATOM(inet_reply);
    INIT_ATOM(timeout);
    INIT_ATOM(closed);
    INIT_ATOM(tcp_passive);
    INIT_ATOM(tcp_closed);
    INIT_ATOM(tcp_error);
#ifdef HAVE_UDP
    INIT_ATOM(udp_passive);
    INIT_ATOM(udp_error);
#endif
#ifdef HAVE_SYS_UN_H
    INIT_ATOM(local);
#endif
    INIT_ATOM(empty_out_q);
    INIT_ATOM(ssl_tls);
#ifndef __WIN32__
    INIT_ATOM(tos);
    INIT_ATOM(tclass);
    INIT_ATOM(ttl);
#endif

    INIT_ATOM(http_eoh);
    INIT_ATOM(http_header);
    INIT_ATOM(http_request);
    INIT_ATOM(http_response);
    INIT_ATOM(http_error);
    INIT_ATOM(abs_path);
    INIT_ATOM(absoluteURI);
    am_star = driver_mk_atom("*");
    INIT_ATOM(http);
    INIT_ATOM(https);
    INIT_ATOM(scheme);

#ifdef HAVE_SENDFILE
    INIT_ATOM(sendfile);
#endif

    /* add TCP, UDP and SCTP drivers */
    add_driver_entry(&tcp_inet_driver_entry);
#ifdef HAVE_UDP
    add_driver_entry(&udp_inet_driver_entry);
#endif

#ifdef HAVE_SCTP
    /* Check the size of SCTP AssocID -- currently both this driver and the
       Erlang part require 32 bit: */
    ERTS_CT_ASSERT(sizeof(sctp_assoc_t)==ASSOC_ID_LEN);
#   if defined(HAVE_SCTP_BINDX)
    p_sctp_bindx = sctp_bindx;
#     if defined(HAVE_SCTP_PEELOFF)
    p_sctp_peeloff = sctp_peeloff;
#     else
    p_sctp_peeloff = NULL;
#     endif
#     if defined(HAVE_SCTP_GETLADDRS) && defined(HAVE_SCTP_FREELADDRS)
    p_sctp_getladdrs = sctp_getladdrs;
    p_sctp_freeladdrs = sctp_freeladdrs;
#     else
    p_sctp_getladdrs = NULL;
    p_sctp_freeladdrs = NULL;
#     endif
#     if defined(HAVE_SCTP_GETPADDRS) && defined(HAVE_SCTP_FREEPADDRS)
    p_sctp_getpaddrs = sctp_getpaddrs;
    p_sctp_freepaddrs = sctp_freepaddrs;
#     else
    p_sctp_getpaddrs = NULL;
    p_sctp_freepaddrs = NULL;
#     endif
    inet_init_sctp();
    add_driver_entry(&sctp_inet_driver_entry);
#   else
#       ifndef LIBSCTP
#           error LIBSCTP not defined
#       endif
    {
	static void *h_libsctp = NULL;

	if (erts_sys_ddll_open_noext(STRINGIFY(LIBSCTP), &h_libsctp, NULL)
	    == 0) {
	    void *ptr;
	    if (erts_sys_ddll_sym(h_libsctp, "sctp_bindx", &ptr) == 0) {
		p_sctp_bindx = ptr;
		if (erts_sys_ddll_sym(h_libsctp, "sctp_peeloff", &ptr) == 0) {
		    p_sctp_peeloff = ptr;
		}
		else p_sctp_peeloff = NULL;
		if (erts_sys_ddll_sym(h_libsctp, "sctp_getladdrs", &ptr) == 0) {
		    p_sctp_getladdrs = ptr;
		}
		else p_sctp_getladdrs = NULL;
		if (erts_sys_ddll_sym(h_libsctp, "sctp_freeladdrs", &ptr) == 0) {
		    p_sctp_freeladdrs = ptr;
		}
		else {
		    p_sctp_freeladdrs = NULL;
		    p_sctp_getladdrs = NULL;
		}
		if (erts_sys_ddll_sym(h_libsctp, "sctp_getpaddrs", &ptr) == 0) {
		    p_sctp_getpaddrs = ptr;
		}
		else p_sctp_getpaddrs = NULL;
		if (erts_sys_ddll_sym(h_libsctp, "sctp_freepaddrs", &ptr) == 0) {
		    p_sctp_freepaddrs = ptr;
		}
		else {
		    p_sctp_freepaddrs = NULL;
		    p_sctp_getpaddrs = NULL;
		}
		inet_init_sctp();
		add_driver_entry(&sctp_inet_driver_entry);
	    }
	    else p_sctp_bindx = NULL;
	}
    }
#   endif
#endif

    /* remove the dummy inet driver */
    remove_driver_entry(&inet_driver_entry);
    return 0;

 error:
    remove_driver_entry(&inet_driver_entry);
    return -1;
}


/*
** Set an inaddr structure:
**  *src = [P1,P0,X1,X2,.....]
**  dst points to a structure large enugh to keep any kind
**  of inaddr.
** *len is set to length of src on call
** and is set to actual length of dst on return
** return NULL if ok or ptr to errno string for error
*/
static char* inet_set_address(int family, inet_address* dst,
			      char* *src, ErlDrvSizeT* len)
{
    short port;

    switch (family) {
    case AF_INET: {
        if (*len < 2+4) return str_einval;
	sys_memzero((char*)dst, sizeof(struct sockaddr_in));
	port = get_int16(*src);
#ifndef NO_SA_LEN
	dst->sai.sin_len    = sizeof(struct sockaddr_in);
#endif
	dst->sai.sin_family = family;
	dst->sai.sin_port   = sock_htons(port);
	sys_memcpy(&dst->sai.sin_addr, (*src)+2, 4);
	*len = sizeof(struct sockaddr_in);
	*src += 2 + 4;
	return NULL;
    }
#if defined(HAVE_IN6) && defined(AF_INET6)
    case AF_INET6: {
        if (*len < 2+16) return str_einval;
	sys_memzero((char*)dst, sizeof(struct sockaddr_in6));
	port = get_int16(*src);
#ifndef NO_SA_LEN
	dst->sai6.sin6_len    = sizeof(struct sockaddr_in6);
#endif
	dst->sai6.sin6_family = family;
	dst->sai6.sin6_port   = sock_htons(port);
	dst->sai6.sin6_flowinfo = 0;   /* XXX this may be set as well ?? */
	sys_memcpy(&dst->sai6.sin6_addr, (*src)+2, 16);
	*len = sizeof(struct sockaddr_in6);
	*src += 2 + 16;
	return NULL;
    }
#endif
#ifdef HAVE_SYS_UN_H
    case AF_UNIX: {
        int n;
        if (*len == 0) return str_einval;
	n = *((unsigned char*)(*src)); /* Length field */
	if (*len < 1+n) return str_einval;
	if (n +
#ifdef __linux__
            /* Make sure the address gets zero terminated
             * except when the first byte is \0 because then it is
             * sort of zero terminated although the zero termination
             * comes before the address...
             * This fix handles Linux's nonportable
             * abstract socket address extension.
             */
            ((*len) > 1 && (*src)[1] == '\0' ? 0 : 1)
#else
            1
#endif
            > sizeof(dst->sal.sun_path)) {
	    return str_einval;
	}
	sys_memzero((char*)dst, sizeof(struct sockaddr_un));
	dst->sal.sun_family = family;
	sys_memcpy(dst->sal.sun_path, (*src)+1, n);
	*len = offsetof(struct sockaddr_un, sun_path) + n;
#ifndef NO_SA_LEN
        dst->sal.sun_len = *len;
#endif
	*src += 1 + n;
	return NULL;
    }
#endif
    }
    return str_eafnosupport;
}

/*
** Set an inaddr structure, address family comes from source data,
** or from argument if source data specifies constant address.
**
** *src = [TAG,P1,P0]
**            when TAG = INET_AF_ANY  | INET_AF_LOOPBACK
** *src = [TAG,P1,P0,X1,X2,...]
**            when TAG = INET_AF_INET | INET_AF_INET6 | INET_AF_LOCAL
** *src = [TAG,Len,...]
**            when TAG = INET_AF_LOCAL
*/
static char *inet_set_faddress(int family, inet_address* dst,
			       char* *src, ErlDrvSizeT* len) {
    int tag;

    if (*len < 1) return str_einval;
    (*len) --;
    tag = *((*src) ++);
    switch (tag) {
    case INET_AF_INET:
	family = AF_INET;
	break;
#   if defined(HAVE_IN6) && defined(AF_INET6)
    case INET_AF_INET6:
	family = AF_INET6;
	break;
#   endif
#   ifdef HAVE_SYS_UN_H
    case INET_AF_LOCAL: {
        family = AF_UNIX;
        break;
    }
#   endif
    case INET_AF_ANY:
    case INET_AF_LOOPBACK: {
	int port;

	if (*len < 2) return str_einval;
	port = get_int16(*src);
	switch (family) {
	case AF_INET: {
	    struct in_addr addr;
	    switch (tag) {
	    case INET_AF_ANY:
		addr.s_addr = sock_htonl(INADDR_ANY);
		break;
	    case INET_AF_LOOPBACK:
		addr.s_addr = sock_htonl(INADDR_LOOPBACK);
		break;
	    default:
		return str_einval;
	    }
	    sys_memzero((char*)dst, sizeof(struct sockaddr_in));
#ifndef NO_SA_LEN
	    dst->sai.sin_len         = sizeof(struct sockaddr_in6);
#endif
	    dst->sai.sin_family      = family;
	    dst->sai.sin_port        = sock_htons(port);
	    dst->sai.sin_addr.s_addr = addr.s_addr;
	    *len = sizeof(struct sockaddr_in);
	}   break;
#       if defined(HAVE_IN6) && defined(AF_INET6)
	case AF_INET6: {
	    const struct in6_addr* paddr;
	    switch (tag) {
	    case INET_AF_ANY:
		paddr = &in6addr_any;
		break;
	    case INET_AF_LOOPBACK:
		paddr = &in6addr_loopback;
		break;
	    default:
		return str_einval;
	    }
	    sys_memzero((char*)dst, sizeof(struct sockaddr_in6));
#ifndef NO_SA_LEN
	    dst->sai6.sin6_len    = sizeof(struct sockaddr_in6);
#endif
	    dst->sai6.sin6_family = family;
	    dst->sai6.sin6_port   = sock_htons(port);
	    dst->sai6.sin6_flowinfo = 0;   /* XXX this may be set as well ?? */
	    dst->sai6.sin6_addr = *paddr;
	    *len = sizeof(struct sockaddr_in6);
	}   break;
#       endif
	default:
	    return str_einval;
	}
	*src += 2;
	return NULL;
    }   break;
    default:
	return str_eafnosupport;
    }
    return inet_set_address(family, dst, src, len);
}

/* Get a inaddr structure
** src = inaddr structure
** dst is filled with [F,P1,P0,X1,....]
** *len is the length of structure
** where F is the family code (coded)
** and *len is the length of dst on return
** (suitable to deliver to erlang)
*/
static int inet_get_address(char* dst, inet_address* src, unsigned int* len)
{
    /* Compare the code with inet_address_to_erlang() */
    int family;
    short port;

    family = src->sa.sa_family;
    if ((family == AF_INET) && (*len >= sizeof(struct sockaddr_in))) {
	dst[0] = INET_AF_INET;
	port = sock_ntohs(src->sai.sin_port);
	put_int16(port, dst+1);
	sys_memcpy(dst+3, (char*)&src->sai.sin_addr, sizeof(struct in_addr));
	*len = 3 + sizeof(struct in_addr);
	return 0;
    }
#if defined(HAVE_IN6) && defined(AF_INET6)
    else if ((family == AF_INET6) && (*len >= sizeof(struct sockaddr_in6))) {
	dst[0] = INET_AF_INET6;
	port = sock_ntohs(src->sai6.sin6_port);
	put_int16(port, dst+1);
	sys_memcpy(dst+3, (char*)&src->sai6.sin6_addr,sizeof(struct in6_addr));
	*len = 3 + sizeof(struct in6_addr);
	return 0;
    }
#endif
#ifdef HAVE_SYS_UN_H
    else if (family == AF_UNIX) {
        size_t n, m;
        if (*len < offsetof(struct sockaddr_un, sun_path)) return -1;
        n = *len - offsetof(struct sockaddr_un, sun_path);
        if (255 < n) return -1;
        m = my_strnlen(src->sal.sun_path, n);
#ifdef __linux__
	/* Assume that the address is a zero terminated string,
         * except when the first byte is \0 i.e the string length is 0,
         * then use the reported length instead.
	 * This fix handles Linux's nonportable
         * abstract socket address extension.
	 */
	if (m == 0)  m = n;
#endif
        dst[0] = INET_AF_LOCAL;
        dst[1] = (char) ((unsigned char) m);
        sys_memcpy(dst+2, src->sal.sun_path, m);
        *len = 1 + 1 + m;
        return 0;
      }
#endif
    else if (family == AF_UNSPEC) {
        dst[0] = INET_AF_UNSPEC;
	*len = 1;
    }
    else {
        dst[0] = INET_AF_UNDEFINED;
	*len = 1;
    }
    return -1;
}

/* Same as the above, but take family from the address structure,
** and advance the address pointer to the next address
** according to the size of the current,
** and return the resulting encoded size
*/
static int
inet_address_to_erlang(char *dst, inet_address **src, SOCKLEN_T sz) {
    /* Compare the code with inet_get_address() */
    short port;

    switch ((*src)->sa.sa_family) {
    case AF_INET:
	if (dst) {
	    dst[0] = INET_AF_INET;
	    port = sock_ntohs((*src)->sai.sin_port);
	    put_int16(port, dst+1);
	    sys_memcpy(dst+1+2, (char *) &(*src)->sai.sin_addr, 4);
	}
	(*src) = (inet_address *) (&(*src)->sai + 1);
	return 1 + 2 + 4;
#if defined(HAVE_IN6) && defined(AF_INET6)
    case AF_INET6:
	if (dst) {
	    dst[0] = INET_AF_INET6;
	    port = sock_ntohs((*src)->sai6.sin6_port);
	    put_int16(port, dst+1);
            VALGRIND_MAKE_MEM_DEFINED(&(*src)->sai6.sin6_addr,16); /* false undefs from syscall sctp_get[lp]addrs */
	    sys_memcpy(dst+1+2, (char *) &(*src)->sai6.sin6_addr, 16);
	}
	(*src) = (inet_address *) (&(*src)->sai6 + 1);
	return 1 + 2 + 16;
#endif
#ifdef HAVE_SYS_UN_H
    case AF_UNIX: {
        size_t n, m;
	if (sz < offsetof(struct sockaddr_un, sun_path)) return -1;
	n = sz - offsetof(struct sockaddr_un, sun_path);
	if (255 < n) return -1;
        m = my_strnlen((*src)->sal.sun_path, n);
#ifdef __linux__
	/* Assume that the address is a zero terminated string,
         * except when the first byte is \0 i.e the string length is 0,
         * Then use the reported length instead.
	 * This fix handles Linux's nonportable
         * abstract socket address extension.
	 */
	if (m == 0)  m = n;
#endif
	if (dst) {
	    dst[0] = INET_AF_LOCAL;
	    dst[1] = (char) ((unsigned char) m);
            sys_memcpy(dst+2, (*src)->sal.sun_path, m);
	}
	(*src) = (inet_address *) (&(*src)->sal + 1);
	return 1 + 1 + m;
    }
#endif
    default:
	return -1;
    }
}

/* Encode n encoded addresses from addrs in the result buffer
*/
static ErlDrvSizeT reply_inet_addrs
(int n, inet_address *addrs, char **rbuf, ErlDrvSizeT rsize, SOCKLEN_T sz) {
    inet_address *ia;
    int i, s;
    ErlDrvSizeT rlen;

    if (IS_SOCKET_ERROR(n)) return ctl_error(sock_errno(), rbuf, rsize);
    if (n == 0) return ctl_reply(INET_REP_OK, NULL, 0, rbuf, rsize);

    /* The sz argument is only used when we have got an actual size
     * of addrs[0] from e.g getsockname() and then n == 1
     * so we will loop over 1 element below.  Otherwise sz
     * would be expected to differ between addresses but that
     * can only happen for AF_UNIX and we will only be called with
     * n > 1 for SCTP and that will never (?) happen with AF_UNIX
     */

    /* Calculate result length */
    rlen = 1;
    ia = addrs;
    for (i = 0;  i < n;  i++) {
        s = inet_address_to_erlang(NULL, &ia, sz);
	if (s < 0) break;
	rlen += s;
    }

    if (rlen > rsize) (*rbuf) = ALLOC(rlen);

    (*rbuf)[0] = INET_REP_OK;
    rlen = 1;
    ia = addrs;
    for (i = 0;  i < n;  i++) {
        s = inet_address_to_erlang((*rbuf)+rlen, &ia, sz);
	if (s < 0) break;
	rlen += s;
    }

    return rlen;
}

static void desc_close(inet_descriptor* desc)
{
    if (desc->s != INVALID_SOCKET) {
#ifdef __WIN32__
	winsock_event_select(desc, FD_READ|FD_WRITE|FD_CLOSE, 0);
	sock_close(desc->s);
	desc->forced_events = 0;
	desc->send_would_block = 0;
#endif
	/*
	 * We should close the fd here, but the other driver might still
	 * be selecting on it.
	 */
	if (!INET_IGNORED(desc))
	    driver_select(desc->port,(ErlDrvEvent)(long)desc->event,
			  ERL_DRV_USE, 0);
	else
	  inet_stop_select((ErlDrvEvent)(long)desc->event,NULL);
	desc->event = INVALID_EVENT; /* closed by stop_select callback */
	desc->s = INVALID_SOCKET;
	desc->event_mask = 0;

	/* mark as disconnected in case when socket is left lingering due to
	 * {exit_on_close, false} option in gen_tcp socket creation. Next
	 * write to socket should produce {error, enotconn} and send a
	 * message {tcp_error,#Port<>,econnreset} */
	desc->state &= ~INET_STATE_CONNECTED;
    }
}

static void desc_close_read(inet_descriptor* desc)
{
    if (desc->s != INVALID_SOCKET) {
#ifdef __WIN32__
	/* This call can not be right???
	 * We want to turn off read events but keep any write events.
	 * But on windows driver_select(...,READ,1) is only used as a
	 * way to hook into the pollset. sock_select is used to control
	 * which events to wait for.
	 * It seems we used to disabled all events for the socket here.
	 *
	driver_select(desc->port, desc->event, DO_READ, 0); REMOVED */
#endif
	sock_select(desc, FD_READ | FD_CLOSE, 0);
    }
}


static int erl_inet_close(inet_descriptor* desc)
{
    free_subscribers(&desc->empty_out_q_subs);
    if ((desc->prebound == 0) && (desc->state & INET_F_OPEN)) {
	desc_close(desc);
	desc->state = INET_STATE_CLOSED;
    } else if (desc->prebound && (desc->s != INVALID_SOCKET)) {
	sock_select(desc, FD_READ | FD_WRITE | FD_CLOSE | ERL_DRV_USE_NO_CALLBACK, 0);
	desc->event_mask = 0;
#ifdef __WIN32__
	desc->forced_events = 0;
	desc->send_would_block = 0;
#endif
    }
    return 0;
}

static ErlDrvSSizeT inet_ctl_open(inet_descriptor* desc, int domain, int type,
				  char** rbuf, ErlDrvSizeT rsize)
{
    int save_errno;
    int protocol;
#ifdef HAVE_SETNS
    int current_ns, new_ns;
    current_ns = new_ns = 0;
#endif
    save_errno = 0;

    if (desc->state != INET_STATE_CLOSED)
	return ctl_xerror(EXBADSEQ, rbuf, rsize);

#ifdef HAVE_SETNS
    if (desc->netns != NULL) {
	/* Temporarily change network namespace for this thread
	 * while creating the socket
	 */
	current_ns = open("/proc/self/ns/net", O_RDONLY);
	if (current_ns == INVALID_SOCKET)
	    return ctl_error(sock_errno(), rbuf, rsize);
	new_ns = open(desc->netns, O_RDONLY);
	if (new_ns == INVALID_SOCKET) {
	    save_errno = sock_errno();
	    while (close(current_ns) == INVALID_SOCKET &&
		   sock_errno() == EINTR);
	    return ctl_error(save_errno, rbuf, rsize);
	}
	if (setns(new_ns, CLONE_NEWNET) != 0) {
	    save_errno = sock_errno();
	    while (close(new_ns) == INVALID_SOCKET &&
		   sock_errno() == EINTR);
	    while (close(current_ns) == INVALID_SOCKET &&
		   sock_errno() == EINTR);
	    return ctl_error(save_errno, rbuf, rsize);
	}
	else {
	    while (close(new_ns) == INVALID_SOCKET &&
		   sock_errno() == EINTR);
	}
    }
#endif
    protocol = desc->sprotocol;
#ifdef HAVE_SYS_UN_H
    if (domain == AF_UNIX) protocol = 0;
#endif
    if ((desc->s = sock_open(domain, type, protocol)) == INVALID_SOCKET)
	save_errno = sock_errno();
#ifdef HAVE_SETNS
    if (desc->netns != NULL) {
	/* Restore network namespace */
	if (setns(current_ns, CLONE_NEWNET) != 0) {
	    /* XXX Failed to restore network namespace.
	     * What to do? Tidy up and return an error...
	     * Note that the thread now might still be in the namespace.
	     * Can this even happen? Should the emulator be aborted?
	     */
	    if (desc->s != INVALID_SOCKET)
		save_errno = sock_errno();
	    while (close(desc->s) == INVALID_SOCKET &&
		   sock_errno() == EINTR);
	    desc->s = INVALID_SOCKET;
	    while (close(current_ns) == INVALID_SOCKET &&
		   sock_errno() == EINTR);
	    return ctl_error(save_errno, rbuf, rsize);
	}
	else {
	    while (close(current_ns) == INVALID_SOCKET &&
		   sock_errno() == EINTR);
	}
    }
#endif
    if (desc->s == INVALID_SOCKET)
	return ctl_error(save_errno, rbuf, rsize);

    if ((desc->event = sock_create_event(desc)) == INVALID_EVENT) {
	save_errno = sock_errno();
	while (close(desc->s) == INVALID_SOCKET &&
	       sock_errno() == EINTR);
	desc->s = INVALID_SOCKET;
	return ctl_error(save_errno, rbuf, rsize);
    }
    SET_NONBLOCKING(desc->s);
#ifdef __WIN32__
    driver_select(desc->port, desc->event, ERL_DRV_READ, 1);
#endif

    desc->state = INET_STATE_OPEN;
    desc->stype = type;
    desc->sfamily = domain;
    return ctl_reply(INET_REP_OK, NULL, 0, rbuf, rsize);
}


/* as inet_open but pass in an open socket (MUST BE OF RIGHT TYPE) */
static ErlDrvSSizeT inet_ctl_fdopen(inet_descriptor* desc, int domain, int type,
				    SOCKET s, Uint32 bound,
                                    char** rbuf, ErlDrvSizeT rsize)
{
    inet_address name;
    unsigned int sz;

    if (bound) {
        /* check that it is a socket and that the socket is bound */
        sz = sizeof(name);
	sys_memzero((char *) &name, sz);
        if (IS_SOCKET_ERROR(sock_name(s, (struct sockaddr*) &name, &sz)))
            return ctl_error(sock_errno(), rbuf, rsize);
        if (name.sa.sa_family != domain)
            return ctl_error(EINVAL, rbuf, rsize);
    }
    desc->s = s;

    if ((desc->event = sock_create_event(desc)) == INVALID_EVENT)
	return ctl_error(sock_errno(), rbuf, rsize);
    SET_NONBLOCKING(desc->s);
#ifdef __WIN32__
    driver_select(desc->port, desc->event, ERL_DRV_READ, 1);
#endif

    desc->state = INET_STATE_OPEN;

    if (type == SOCK_STREAM) { /* check if connected */
	sz = sizeof(name);
	if (!IS_SOCKET_ERROR(sock_peer(s, (struct sockaddr*) &name, &sz))) {
	    desc->state = INET_STATE_CONNECTED;
        }
    }

    desc->prebound = 1; /* used to prevent a real close since
			 * the fd probably comes from an
			 * external wrapper program, so it is
			 * not certain that we can open it again */
    desc->stype = type;
    desc->sfamily = domain;
    return ctl_reply(INET_REP_OK, NULL, 0, rbuf, rsize);
}

/*
**  store interface info as: (bytes)
**  [Len] Name(Len) Flags(1) addr(4) baddr(4) mask(4) bw(4)
*/
struct addr_if {
    char name[INET_IFNAMSIZ];
    long           flags;        /* coded flags */
    struct in_addr addr;         /* interface address */
    struct in_addr baddr;        /* broadcast address */
    struct in_addr mask;         /* netmask */
};


#ifndef SIOCGIFNETMASK
static struct in_addr net_mask(struct in_addr in)
{
    register u_long i = sock_ntohl(in.s_addr);

    if (IN_CLASSA(i))
	in.s_addr = sock_htonl(IN_CLASSA_NET);
    else if (IN_CLASSB(i))
	in.s_addr = sock_htonl(IN_CLASSB_NET);
    else
	in.s_addr = sock_htonl(IN_CLASSC_NET);
    return in;
}
#endif

#if defined(__WIN32__) && defined(SIO_GET_INTERFACE_LIST)

/* format address in dot notation */
static char* fmt_addr(unsigned long x, char* ptr)
{
    int i;
    for (i = 0; i < 4; i++) {
	int nb[3];
	int y = (x >> 24) & 0xff;
	x <<= 8;
	nb[0] = y % 10; y /= 10;
	nb[1] = y % 10; y /= 10;
	nb[2] = y % 10; y /= 10;
	switch((nb[2] ? 3 : (nb[1] ? 2 : 1))) {
	case 3:  *ptr++ = nb[2] + '0';
	case 2:  *ptr++ = nb[1] + '0';
	case 1:  *ptr++ = nb[0] + '0';
	}
	*ptr++ = '.';
    }
    *(ptr-1) = '\0';
    return ptr;
}

static int parse_addr(char* ptr, int n, long* x)
{
    long addr = 0;
    int  dots = 0;
    int  digs = 0;
    int  v  = 0;

    while(n--) {
	switch(*ptr) {
	case '0': case '1': case '2':case '3':case '4':case '5':
	case '6': case '7': case '8':case '9':
	    v = v*10 + *ptr - '0';
	    if (++digs > 3) return -1;
	    break;
	case '.':
	    if ((dots>2) || (digs==0) || (digs > 3) || (v > 0xff)) return -1;
	    dots++;
	    digs = 0;
	    addr = (addr << 8) | v;
	    v = 0;
	    break;
	default:
	    return -1;
	}
	ptr++;
    }
    if ((dots!=3) || (digs==0) || (digs > 3) || (v > 0xff)) return -1;
    addr = (addr << 8) | v;
    *x = addr;
    return 0;
}

#endif

#define buf_check(ptr, end, n) \
do { if ((end)-(ptr) < (n)) goto error; } while(0)

static char* sockaddr_to_buf(struct sockaddr* addr, char* ptr, char* end)
{
    if (addr->sa_family == AF_INET || addr->sa_family == 0) {
	struct in_addr *p = &(((struct sockaddr_in*) addr)->sin_addr);
	buf_check(ptr, end, 1 + sizeof(struct in_addr));
	*ptr = INET_AF_INET;
	sys_memcpy(ptr+1, (char*)p, sizeof(struct in_addr));
	return ptr + 1 + sizeof(struct in_addr);
    }
#if defined(HAVE_IN6) && defined(AF_INET6)
    else if (addr->sa_family == AF_INET6) {
	struct in6_addr *p = &(((struct sockaddr_in6*) addr)->sin6_addr);
	buf_check(ptr, end, 1 + sizeof(struct in6_addr));
	*ptr = INET_AF_INET6;
	sys_memcpy(ptr+1, (char*)p, sizeof(struct in6_addr));
	return ptr + 1 + sizeof(struct in6_addr);
    }
#endif
#if defined(AF_LINK)
    else if (addr->sa_family == AF_LINK) {
	struct sockaddr_dl *sdl_p = (struct sockaddr_dl*) addr;
	buf_check(ptr, end, 2 + sdl_p->sdl_alen);
	put_int16(sdl_p->sdl_alen, ptr); ptr += 2;
	sys_memcpy(ptr, sdl_p->sdl_data + sdl_p->sdl_nlen, sdl_p->sdl_alen);
	return ptr + sdl_p->sdl_alen;
    }
#endif
#if defined(AF_PACKET) && defined(HAVE_NETPACKET_PACKET_H)
    else if(addr->sa_family == AF_PACKET) {
	struct sockaddr_ll *sll_p = (struct sockaddr_ll*) addr;
	buf_check(ptr, end, 2 + sll_p->sll_halen);
	put_int16(sll_p->sll_halen, ptr); ptr += 2;
	sys_memcpy(ptr, sll_p->sll_addr, sll_p->sll_halen);
	return ptr + sll_p->sll_halen;
    }
#endif
    return ptr;
 error:
    return NULL;
}

/* sockaddr_bufsz_need
 * Returns the number of bytes needed to store the information
 * through sockaddr_to_buf
 */

static size_t sockaddr_bufsz_need(struct sockaddr* addr)
{
    if (addr->sa_family == AF_INET || addr->sa_family == 0) {
	return 1 + sizeof(struct in_addr);
    }
#if defined(HAVE_IN6) && defined(AF_INET6)
    else if (addr->sa_family == AF_INET6) {
	return 1 + sizeof(struct in6_addr);
    }
#endif
#if defined(AF_LINK)
    if (addr->sa_family == AF_LINK) {
	struct sockaddr_dl *sdl_p = (struct sockaddr_dl*) addr;
	return 2 + sdl_p->sdl_alen;
    }
#endif
#if defined(AF_PACKET) && defined(HAVE_NETPACKET_PACKET_H)
    else if(addr->sa_family == AF_PACKET) {
	struct sockaddr_ll *sll_p = (struct sockaddr_ll*) addr;
	return 2 + sll_p->sll_halen;
    }
#endif
    return 0;
}

static char* buf_to_sockaddr(char* ptr, char* end, struct sockaddr* addr)
{
    buf_check(ptr,end,1);
    switch (*ptr++) {
    case INET_AF_INET: {
	struct in_addr *p = &((struct sockaddr_in*)addr)->sin_addr;
	buf_check(ptr,end,sizeof(struct in_addr));
	sys_memcpy((char*) p, ptr, sizeof(struct in_addr));
	addr->sa_family = AF_INET;
	return ptr + sizeof(struct in_addr);
    }
#if defined(HAVE_IN6) && defined(AF_INET6)
    case INET_AF_INET6: {
	struct in6_addr *p = &((struct sockaddr_in6*)addr)->sin6_addr;
	buf_check(ptr,end,sizeof(struct in6_addr));
	sys_memcpy((char*) p, ptr, sizeof(struct in6_addr));
	addr->sa_family = AF_INET6;
	return ptr + sizeof(struct in6_addr);
    }
#endif
    }
 error:
    return NULL;
}


#if defined (IFF_POINTOPOINT)
#define IFGET_FLAGS(cflags) IFGET_FLAGS_P2P(cflags, IFF_POINTOPOINT)
#elif defined IFF_POINTTOPOINT
#define IFGET_FLAGS(cflags) IFGET_FLAGS_P2P(cflags, IFF_POINTTOPOINT)
#endif

#define IFGET_FLAGS_P2P(cflags, iff_ptp)				\
    ((((cflags) & IFF_UP) ? INET_IFF_UP : 0) |				\
     (((cflags) & IFF_BROADCAST) ? INET_IFF_BROADCAST : 0) |		\
     (((cflags) & IFF_LOOPBACK) ? INET_IFF_LOOPBACK : 0) |		\
     (((cflags) & iff_ptp) ? INET_IFF_POINTTOPOINT : 0) |		\
     (((cflags) & IFF_UP) ? INET_IFF_RUNNING : 0) |  /* emulate running ? */ \
     (((cflags) & IFF_MULTICAST) ? INET_IFF_MULTICAST : 0))

#if defined(__WIN32__) && defined(SIO_GET_INTERFACE_LIST)

static ErlDrvSSizeT inet_ctl_getiflist(inet_descriptor* desc,
				       char** rbuf, ErlDrvSizeT rsize)
{
    char ifbuf[BUFSIZ];
    char sbuf[BUFSIZ];
    char* sptr;
    INTERFACE_INFO* ifp;
    DWORD len;
    ErlDrvSizeT n;
    int err;

    ifp = (INTERFACE_INFO*) ifbuf;
    len = 0;
    err = WSAIoctl(desc->s, SIO_GET_INTERFACE_LIST, NULL, 0,
		   (LPVOID) ifp, BUFSIZ, (LPDWORD) &len,
		   NULL, NULL);

    if (err == SOCKET_ERROR)
	return ctl_error(sock_errno(), rbuf, rsize);

    n = (len + sizeof(INTERFACE_INFO) - 1) / sizeof(INTERFACE_INFO);
    sptr = sbuf;

    while(n--) {
	if (((struct sockaddr*)&ifp->iiAddress)->sa_family == desc->sfamily) {
	    struct in_addr sina = ((struct sockaddr_in*)&ifp->iiAddress)->sin_addr;
	    /* discard INADDR_ANY interface address */
	    if (sina.s_addr != INADDR_ANY)
		sptr = fmt_addr(sock_ntohl(sina.s_addr), sptr);
	}
	ifp++;
    }
    return ctl_reply(INET_REP_OK, sbuf, sptr - sbuf, rbuf, rsize);
}

/* input is an ip-address in string format i.e A.B.C.D
** scan the INTERFACE_LIST to get the options
*/
static ErlDrvSSizeT inet_ctl_ifget(inet_descriptor* desc, char* buf,
				   ErlDrvSizeT len, char** rbuf, ErlDrvSizeT rsize)
{
    char ifbuf[BUFSIZ];
    int  n;
    char sbuf[BUFSIZ];
    char* sptr;
    char* s_end = sbuf + BUFSIZ;
    int namlen;
    int   err;
    INTERFACE_INFO* ifp;
    long namaddr;

    if ((len == 0) || ((namlen = get_int8(buf)) > len))
	goto error;
    if (parse_addr(buf+1, namlen, &namaddr) < 0)
	goto error;
    namaddr = sock_ntohl(namaddr);
    buf += (namlen+1);
    len -= (namlen+1);

    ifp = (INTERFACE_INFO*) ifbuf;
    err = WSAIoctl(desc->s, SIO_GET_INTERFACE_LIST, NULL, 0,
			      (LPVOID) ifp, BUFSIZ, (LPDWORD) &n,
			      NULL, NULL);
    if (err == SOCKET_ERROR) {
	return ctl_error(sock_errno(), rbuf, rsize);
    }

    n = (n + sizeof(INTERFACE_INFO) - 1) / sizeof(INTERFACE_INFO);

    /* find interface */
    while(n) {
	if (((struct sockaddr_in*)&ifp->iiAddress)->sin_addr.s_addr == namaddr)
	    break;
	ifp++;
	n--;
    }
    if (n == 0)
	goto error;

    sptr = sbuf;

    while (len--) {
	switch(*buf++) {
	case INET_IFOPT_ADDR:
	    buf_check(sptr, s_end, 1);
	    *sptr++ = INET_IFOPT_ADDR;
	    if ((sptr = sockaddr_to_buf((struct sockaddr *)&ifp->iiAddress,
					sptr, s_end)) == NULL)
		goto error;
	    break;

	case INET_IFOPT_HWADDR:
	    break;

	case INET_IFOPT_BROADADDR:
#ifdef SIOCGIFBRDADDR
	    buf_check(sptr, s_end, 1);
	    *sptr++ = INET_IFOPT_BROADADDR;
	    if ((sptr=sockaddr_to_buf((struct sockaddr *)
				      &ifp->iiBroadcastAddress,sptr,s_end))
		== NULL)
		goto error;
#endif
	    break;

	case INET_IFOPT_DSTADDR:
	    break;

	case INET_IFOPT_NETMASK:
	    buf_check(sptr, s_end, 1);
	    *sptr++ = INET_IFOPT_NETMASK;
	    if ((sptr = sockaddr_to_buf((struct sockaddr *)
					&ifp->iiNetmask,sptr,s_end)) == NULL)
		goto error;
	    break;

	case INET_IFOPT_MTU:
	    break;

	case INET_IFOPT_FLAGS: {
	    int flags = ifp->iiFlags;
	    /* just enumerate the interfaces (no names) */

	    buf_check(sptr, s_end, 5);
	    *sptr++ = INET_IFOPT_FLAGS;
	    put_int32(IFGET_FLAGS(flags), sptr);
	    sptr += 4;
	    break;
	}
	default:
	    goto error;
	}
    }
    return ctl_reply(INET_REP_OK, sbuf, sptr - sbuf, rbuf, rsize);

 error:
    return ctl_error(EINVAL, rbuf, rsize);
}

/* not supported */
static ErlDrvSSizeT inet_ctl_ifset(inet_descriptor* desc,
				   char* buf, ErlDrvSizeT len,
				   char** rbuf, ErlDrvSizeT rsize)
{
    return ctl_reply(INET_REP_OK, NULL, 0, rbuf, rsize);
}

#elif defined(SIOCGIFCONF) && defined(SIOCSIFFLAGS)
/* cygwin has SIOCGIFCONF but not SIOCSIFFLAGS (Nov 2002) */

#define VOIDP(x) ((void*)(x))
#if defined(AF_LINK) && !defined(NO_SA_LEN)
#define SIZEA(p) (((p).sa_len > sizeof(p)) ? (p).sa_len : sizeof(p))
#else
#define SIZEA(p) (sizeof (p))
#endif

static int get_ifconf(SOCKET s, struct ifconf *ifcp) {
    int ifc_len = 0;
    int buflen = 100 * sizeof(struct ifreq);
    char *buf = ALLOC(buflen);

    for (;;) {
	ifcp->ifc_len = buflen;
	ifcp->ifc_buf = buf;
	if (ioctl(s, SIOCGIFCONF, (char *)ifcp) < 0) {
	    int res = sock_errno();
	    if (res != EINVAL || ifc_len) {
		FREE(buf);
		return -1;
	    }
	} else {
	    if (ifcp->ifc_len == ifc_len) break; /* buf large enough */
	    ifc_len = ifcp->ifc_len;
	}
	buflen += 10 * sizeof(struct ifreq);
	buf = (char *)REALLOC(buf, buflen);
    }
    return 0;
}

static void free_ifconf(struct ifconf *ifcp) {
    FREE(ifcp->ifc_buf);
}

static ErlDrvSSizeT inet_ctl_getiflist(inet_descriptor* desc,
				       char** rbuf, ErlDrvSizeT rsize)
{
    struct ifconf ifc;
    struct ifreq *ifrp;
    char *sbuf, *sp;
    ErlDrvSizeT i;

    /* Courtesy of Per Bergqvist and W. Richard Stevens */

    if (get_ifconf(desc->s, &ifc) < 0) {
	return ctl_error(sock_errno(), rbuf, rsize);
    }

    sp = sbuf = ALLOC(ifc.ifc_len+1);
    *sp++ = INET_REP_OK;
    i = 0;
    for (;;) {
	ErlDrvSizeT n;

	ifrp = (struct ifreq *) VOIDP(ifc.ifc_buf + i);
	n = sizeof(ifrp->ifr_name) + SIZEA(ifrp->ifr_addr);
	if (n < sizeof(*ifrp)) n = sizeof(*ifrp);
	if (i+n > ifc.ifc_len) break;
	i += n;

	switch (ifrp->ifr_addr.sa_family) {
#if defined(HAVE_IN6) && defined(AF_INET6)
	case AF_INET6:
#endif
	case AF_INET:
	    ASSERT(sp+IFNAMSIZ+1 < sbuf+ifc.ifc_len+1);
	    strncpy(sp, ifrp->ifr_name, IFNAMSIZ);
	    sp[IFNAMSIZ] = '\0';
	    sp += strlen(sp), ++sp;
	}

	if (i >= ifc.ifc_len) break;
    }
    free_ifconf(&ifc);
    *rbuf = sbuf;
    return sp - sbuf;
}

#ifdef HAVE_LIBDLPI_H
#include <libdlpi.h>
static int hwaddr_libdlpi_lookup(const char *ifnm,
                                 uchar_t *addr, size_t *alen)
{
    dlpi_handle_t handle;
    dlpi_info_t linkinfo;
    int ret = -1;

    if (dlpi_open(ifnm, &handle, 0) != DLPI_SUCCESS) {
        return -1;
    }

    if (dlpi_get_physaddr(handle, DL_CURR_PHYS_ADDR,
                          addr, alen) == DLPI_SUCCESS &&
        dlpi_info(handle, &linkinfo, 0) == DLPI_SUCCESS)
    {
        ret = 0;
    }

    dlpi_close(handle);
    return ret;
}
#endif

#ifdef HAVE_GETIFADDRS
/* Returns 0 for success and errno() for failure */
static int call_getifaddrs(inet_descriptor* desc_p, struct ifaddrs **ifa_pp)
{
    int result, save_errno;
#ifdef HAVE_SETNS
    int current_ns;

    current_ns = 0;
    if (desc_p->netns != NULL) {
        int new_ns;
	/* Temporarily change network namespace for this thread
	 * over the getifaddrs() call
	 */
	current_ns = open("/proc/self/ns/net", O_RDONLY);
	if (current_ns == INVALID_SOCKET)
	    return sock_errno();
	new_ns = open(desc_p->netns, O_RDONLY);
	if (new_ns == INVALID_SOCKET) {
	    save_errno = sock_errno();
	    while (close(current_ns) == INVALID_SOCKET &&
		   sock_errno() == EINTR);
	    return save_errno;
	}
	if (setns(new_ns, CLONE_NEWNET) != 0) {
	    save_errno = sock_errno();
	    while (close(new_ns) == INVALID_SOCKET &&
		   sock_errno() == EINTR);
	    while (close(current_ns) == INVALID_SOCKET &&
		   sock_errno() == EINTR);
	    return save_errno;
	}
	else {
	    while (close(new_ns) == INVALID_SOCKET &&
		   sock_errno() == EINTR);
	}
    }
#endif
    save_errno = 0;
    result = getifaddrs(ifa_pp);
    if (result  < 0)
        save_errno = sock_errno();
#ifdef HAVE_SETNS
    if (desc_p->netns != NULL) {
        /* Restore network namespace */
        if (setns(current_ns, CLONE_NEWNET) != 0) {
            /* XXX Failed to restore network namespace.
             * What to do? Tidy up and return an error...
             * Note that the thread now might still be in the set namespace.
	     * Can this even happen? Should the emulator be aborted?
	     */
            if (result >= 0) {
                /* We got a result but have to waste it */
                save_errno = sock_errno();
                freeifaddrs(*ifa_pp);
            }
	}
        while (close(current_ns) == INVALID_SOCKET &&
               sock_errno() == EINTR);
    }
#endif
    return save_errno;
}
#endif /* #ifdef HAVE_GETIFADDRS */

/* FIXME: temporary hack */
#ifndef IFHWADDRLEN
#define IFHWADDRLEN 6
#endif

static ErlDrvSSizeT inet_ctl_ifget(inet_descriptor* desc,
				   char* buf, ErlDrvSizeT len,
				   char** rbuf, ErlDrvSizeT rsize)
{
    char sbuf[BUFSIZ];
    char* sptr;
    char* s_end = sbuf + BUFSIZ;
    struct ifreq ifreq;
    int namlen;

    if ((len == 0) || ((namlen = get_int8(buf)) > len))
	goto error;
    sys_memset(ifreq.ifr_name, '\0', IFNAMSIZ);
    sys_memcpy(ifreq.ifr_name, buf+1,
	       (namlen >= IFNAMSIZ) ? IFNAMSIZ-1 : namlen);
    buf += (namlen+1);
    len -= (namlen+1);
    sptr = sbuf;

    while (len--) {
	switch(*buf++) {
	case INET_IFOPT_ADDR:
	    if (ioctl(desc->s, SIOCGIFADDR, (char *)&ifreq) < 0)
		break;
	    buf_check(sptr, s_end, 1);
	    *sptr++ = INET_IFOPT_ADDR;
	    if ((sptr = sockaddr_to_buf(&ifreq.ifr_addr, sptr, s_end)) == NULL)
		goto error;
	    break;

	case INET_IFOPT_HWADDR: {
#ifdef HAVE_LIBDLPI_H
	    /*
	    ** OpenSolaris have SIGCGIFHWADDR, but no ifr_hwaddr member..
	    ** The proper way to get the mac address would be to
	    ** use libdlpi...
	    */
	    uchar_t addr[DLPI_PHYSADDR_MAX];
	    size_t alen = sizeof(addr);

	    if (hwaddr_libdlpi_lookup(ifreq.ifr_name, addr, &alen) == 0) {
		buf_check(sptr, s_end, 1+2+alen);
		*sptr++ = INET_IFOPT_HWADDR;
		put_int16(alen, sptr);
                    sptr += 2;
                    sys_memcpy(sptr, addr, alen);
                    sptr += alen;
	    }
#elif defined(SIOCGIFHWADDR) && defined(HAVE_STRUCT_IFREQ_IFR_HWADDR)
	    if (ioctl(desc->s, SIOCGIFHWADDR, (char *)&ifreq) < 0)
		break;
	    buf_check(sptr, s_end, 1+2+IFHWADDRLEN);
	    *sptr++ = INET_IFOPT_HWADDR;
	    put_int16(IFHWADDRLEN, sptr); sptr += 2;
	    /* raw memcpy (fix include autoconf later) */
	    sys_memcpy(sptr, (char*)(&ifreq.ifr_hwaddr.sa_data), IFHWADDRLEN);
	    sptr += IFHWADDRLEN;
#elif defined(SIOCGENADDR) && defined(HAVE_STRUCT_IFREQ_IFR_ENADDR)
	    if (ioctl(desc->s, SIOCGENADDR, (char *)&ifreq) < 0)
		break;
	    buf_check(sptr, s_end, 1+2+sizeof(ifreq.ifr_enaddr));
	    *sptr++ = INET_IFOPT_HWADDR;
	    put_int16(sizeof(ifreq.ifr_enaddr), sptr); sptr += 2;
	    /* raw memcpy (fix include autoconf later) */
	    sys_memcpy(sptr, (char*)(&ifreq.ifr_enaddr),
		       sizeof(ifreq.ifr_enaddr));
	    sptr += sizeof(ifreq.ifr_enaddr);
#elif defined(HAVE_GETIFADDRS) && defined(AF_LINK)
	    struct ifaddrs *ifa, *ifp;
	    struct sockaddr_dl *sdlp;
	    int found = 0;

            if (call_getifaddrs(desc, &ifa) != 0)
                goto error;

	    for (ifp = ifa; ifp; ifp = ifp->ifa_next) {
		if ((ifp->ifa_addr->sa_family == AF_LINK) &&
		    (sys_strcmp(ifp->ifa_name, ifreq.ifr_name) == 0)) {
		    found = 1;
		    break;
		}
	    }

	    if (found == 0) {
		freeifaddrs(ifa);
		break;
	    }
	    sdlp = (struct sockaddr_dl *)ifp->ifa_addr;

	    buf_check(sptr, s_end, 1+2+sdlp->sdl_alen);
	    *sptr++ = INET_IFOPT_HWADDR;
	    put_int16(sdlp->sdl_alen, sptr); sptr += 2;
	    sys_memcpy(sptr,
		       sdlp->sdl_data + sdlp->sdl_nlen,
		       sdlp->sdl_alen);
	    sptr += sdlp->sdl_alen;
	    freeifaddrs(ifa);
#endif
	    break;
	}


	case INET_IFOPT_BROADADDR:
#ifdef SIOCGIFBRDADDR
	    if (ioctl(desc->s, SIOCGIFBRDADDR, (char *)&ifreq) < 0)
		break;
	    buf_check(sptr, s_end, 1);
	    *sptr++ = INET_IFOPT_BROADADDR;
	    if ((sptr=sockaddr_to_buf(&ifreq.ifr_broadaddr,sptr,s_end)) == NULL)
		goto error;
#endif
	    break;

	case INET_IFOPT_DSTADDR:
#ifdef SIOCGIFDSTADDR
	    if (ioctl(desc->s, SIOCGIFDSTADDR, (char *)&ifreq) < 0)
		break;
	    buf_check(sptr, s_end, 1);
	    *sptr++ = INET_IFOPT_DSTADDR;
	    if ((sptr = sockaddr_to_buf(&ifreq.ifr_dstaddr,sptr,s_end)) == NULL)
		goto error;
#endif
	    break;

	case INET_IFOPT_NETMASK:
#if defined(SIOCGIFNETMASK)
	    if (ioctl(desc->s, SIOCGIFNETMASK, (char *)&ifreq) < 0)
		break;
	    buf_check(sptr, s_end, 1);
	    *sptr++ = INET_IFOPT_NETMASK;
#if defined(ifr_netmask)
	    sptr = sockaddr_to_buf(&ifreq.ifr_netmask,sptr,s_end);
#else
	    /* SIOCGNETMASK exist but not macro ??? */
	    sptr = sockaddr_to_buf(&ifreq.ifr_addr,sptr,s_end);
#endif
	    if (sptr == NULL)
		goto error;
#else
	    if (ioctl(desc->s, SIOCGIFADDR, (char *)&ifreq) < 0)
		break;
	    else {
		struct sockadd_in* ap;
		/* emulate netmask,
		 * (wasted stuff since noone uses classes)
		 */
		buf_check(sptr, s_end, 1);
		*sptr++ = INET_IFOPT_NETMASK;
		ap = (struct sockaddr_in*) VOIDP(&ifreq.ifr_addr);
		ap->sin_addr = net_mask(ap->sin_addr);
		if ((sptr = sockaddr_to_buf(&ifreq.ifr_addr,sptr,s_end)) == NULL)
		    goto error;
	    }
#endif
	    break;

	case INET_IFOPT_MTU: {
#if defined(SIOCGIFMTU) && defined(ifr_mtu)
	    int n;

	    if (ioctl(desc->s, SIOCGIFMTU, (char *)&ifreq) < 0)
		break;
	    buf_check(sptr, s_end, 5);
	    *sptr++ = INET_IFOPT_MTU;
	    n = ifreq.ifr_mtu;
	    put_int32(n, sptr);
	    sptr += 4;
#endif
	    break;
	}

	case INET_IFOPT_FLAGS: {
	    int flags;

	    if (ioctl(desc->s, SIOCGIFFLAGS, (char*)&ifreq) < 0)
		flags = 0;
	    else
		flags = ifreq.ifr_flags;

	    buf_check(sptr, s_end, 5);
	    *sptr++ = INET_IFOPT_FLAGS;
	    put_int32(IFGET_FLAGS(flags), sptr);
	    sptr += 4;
	    break;
	}
	default:
	    goto error;
	}
    }
    return ctl_reply(INET_REP_OK, sbuf, sptr - sbuf, rbuf, rsize);

 error:
    return ctl_error(EINVAL, rbuf, rsize);
}


static ErlDrvSSizeT inet_ctl_ifset(inet_descriptor* desc,
				   char* buf, ErlDrvSizeT len,
				   char** rbuf, ErlDrvSizeT rsize)
{
    struct ifreq ifreq;
    int namlen;
    char* b_end = buf + len;

    if ((len == 0) || ((namlen = get_int8(buf)) > len))
	goto error;
    sys_memset(ifreq.ifr_name, '\0', IFNAMSIZ);
    sys_memcpy(ifreq.ifr_name, buf+1,
	       (namlen >= IFNAMSIZ) ? IFNAMSIZ-1 : namlen);
    buf += (namlen+1);
    len -= (namlen+1);

    while(buf < b_end) {
	switch(*buf++) {
	case INET_IFOPT_ADDR:
	    if ((buf = buf_to_sockaddr(buf, b_end, &ifreq.ifr_addr)) == NULL)
		goto error;
	    (void) ioctl(desc->s, SIOCSIFADDR, (char*)&ifreq);
	    break;

	case INET_IFOPT_HWADDR: {
	    unsigned int hwalen;
	    buf_check(buf, b_end, 2);
	    hwalen = get_int16(buf); buf += 2;
	    buf_check(buf, b_end, hwalen);
#ifdef SIOCSIFHWADDR
	    /* raw memcpy (fix include autoconf later) */
	    sys_memset((char*)(&ifreq.ifr_hwaddr.sa_data),
		       '\0', sizeof(ifreq.ifr_hwaddr.sa_data));
	    sys_memcpy((char*)(&ifreq.ifr_hwaddr.sa_data), buf, hwalen);

	    (void) ioctl(desc->s, SIOCSIFHWADDR, (char *)&ifreq);
#endif
	    buf += hwalen;
	    break;
	}

	case INET_IFOPT_BROADADDR:
#ifdef SIOCSIFBRDADDR
	    if ((buf = buf_to_sockaddr(buf, b_end, &ifreq.ifr_broadaddr)) == NULL)
		goto error;
	    (void) ioctl(desc->s, SIOCSIFBRDADDR, (char *)&ifreq);
#endif
	    break;

	case INET_IFOPT_DSTADDR:
#ifdef SIOCSIFDSTADDR
	    if ((buf = buf_to_sockaddr(buf, b_end, &ifreq.ifr_dstaddr)) == NULL)
		goto error;
	    (void) ioctl(desc->s, SIOCSIFDSTADDR, (char *)&ifreq);
#endif
	    break;

	case INET_IFOPT_NETMASK:
#ifdef SIOCSIFNETMASK

#if defined(ifr_netmask)
	    buf = buf_to_sockaddr(buf,b_end, &ifreq.ifr_netmask);
#else
	    buf = buf_to_sockaddr(buf,b_end, &ifreq.ifr_addr);
#endif
	    if (buf == NULL)
		goto error;
	    (void) ioctl(desc->s, SIOCSIFNETMASK, (char *)&ifreq);
#endif
	    break;

	case INET_IFOPT_MTU:
	    buf_check(buf, b_end, 4);
#if defined(SIOCSIFMTU) && defined(ifr_mtu)
	    ifreq.ifr_mtu = get_int32(buf);
	    (void) ioctl(desc->s, SIOCSIFMTU, (char *)&ifreq);
#endif
	    buf += 4;
	    break;

	case INET_IFOPT_FLAGS: {
	    int flags0;
	    int flags;
	    int eflags;

	    buf_check(buf, b_end, 4);
	    eflags = get_int32(buf);

	    /* read current flags */
	    if (ioctl(desc->s, SIOCGIFFLAGS, (char*)&ifreq) < 0)
		flags0 = flags = 0;
	    else
		flags0 = flags = ifreq.ifr_flags;

	    /* update flags */
	    if (eflags & INET_IFF_UP)            flags |= IFF_UP;
	    if (eflags & INET_IFF_DOWN)          flags &= ~IFF_UP;
	    if (eflags & INET_IFF_BROADCAST)     flags |= IFF_BROADCAST;
	    if (eflags & INET_IFF_NBROADCAST)    flags &= ~IFF_BROADCAST;
	    if (eflags & INET_IFF_POINTTOPOINT)  flags |= IFF_POINTOPOINT;
	    if (eflags & INET_IFF_NPOINTTOPOINT) flags &= ~IFF_POINTOPOINT;

	    if (flags != flags0) {
		ifreq.ifr_flags = flags;
		(void) ioctl(desc->s, SIOCSIFFLAGS, (char*)&ifreq);
	    }
	    buf += 4;
	    break;
	}

	default:
	    goto error;
	}
    }
    return ctl_reply(INET_REP_OK, NULL, 0, rbuf, rsize);

 error:
    return ctl_error(EINVAL, rbuf, rsize);
}

#else


static ErlDrvSSizeT inet_ctl_getiflist(inet_descriptor* desc,
				       char** rbuf, ErlDrvSizeT rsize)
{
    return ctl_reply(INET_REP_OK, NULL, 0, rbuf, rsize);
}


static ErlDrvSSizeT inet_ctl_ifget(inet_descriptor* desc,
				   char* buf, ErlDrvSizeT len,
				   char** rbuf, ErlDrvSizeT rsize)
{
    return ctl_reply(INET_REP_OK, NULL, 0, rbuf, rsize);
}


static ErlDrvSSizeT inet_ctl_ifset(inet_descriptor* desc,
				   char* buf, ErlDrvSizeT len,
				   char** rbuf, ErlDrvSizeT rsize)
{
    return ctl_reply(INET_REP_OK, NULL, 0, rbuf, rsize);
}

#endif



#if defined(__WIN32__) || defined(HAVE_GETIFADDRS)
/* Latin-1 to utf8 */

static int utf8_len(const char *c, int m) {
    int l;
    for (l = 0;  m;  c++, l++, m--) {
	if (*c == '\0') break;
	if ((*c & 0x7f) != *c) l++;
    }
    return l;
}

static void utf8_encode(const char *c, int m, char *p) {
    for (;  m;  c++, m--) {
	if (*c == '\0') break;
	if ((*c & 0x7f) != *c) {
	    *p++ = (char) (0xC0 | (0x03 & (*c >> 6)));
	    *p++ = (char) (0x80 | (0x3F & *c));
	} else {
	    *p++ = (char) *c;
	}
    }
}
#endif

#if defined(__WIN32__)

static void set_netmask_bytes(char *c, int len, int pref_len) {
    int i, m;
    for (i = 0, m = pref_len >> 3;  i < m && i < len;  i++) c[i] = '\xFF';
    if (i < len) c[i++] = 0xFF << (8 - (pref_len & 7));
    for (;  i < len;  i++) c[i] = '\0';
}


int eq_masked_bytes(char *a, char *b, int pref_len) {
    int i, m;
    for (i = 0, m = pref_len >> 3;  i < m;  i++) {
	if (a[i] != b[i]) return 0;
    }
    m = pref_len & 7;
    if (m) {
	m = 0xFF & (0xFF << (8 - m));
	if ((a[i] & m) != (b[i] & m)) return 0;
    }
    return !0;
}

static ErlDrvSSizeT inet_ctl_getifaddrs(inet_descriptor* desc_p,
					char **rbuf_pp, ErlDrvSizeT rsize)
{
    int i;
    DWORD ret, n;
    IP_INTERFACE_INFO *info_p;
    MIB_IPADDRTABLE *ip_addrs_p;
    IP_ADAPTER_ADDRESSES *ip_adaddrs_p, *ia_p;

    char *buf_p;
    char *buf_alloc_p;
    ErlDrvSizeT buf_size = 512;
#   define BUF_ENSURE(Size)						\
    do {								\
	int NEED_, GOT_ = buf_p - buf_alloc_p;				\
	NEED_ = GOT_ + (Size);						\
	if (NEED_ > buf_size) {						\
	    buf_size = NEED_ + 512;					\
	    buf_alloc_p = REALLOC(buf_alloc_p, buf_size);		\
	    buf_p = buf_alloc_p + GOT_;					\
	}								\
    } while(0)
#   define SOCKADDR_TO_BUF(opt, sa)					\
    do {								\
	if (sa) {							\
	    char *P_;							\
	    *buf_p++ = (opt);						\
	    while (! (P_ = sockaddr_to_buf((sa), buf_p,			\
					   buf_alloc_p+buf_size))) {    \
		int GOT_ = buf_p - buf_alloc_p;				\
		buf_size += 512;					\
		buf_alloc_p = REALLOC(buf_alloc_p, buf_size);		\
		buf_p = buf_alloc_p + GOT_;				\
	    }								\
	    if (P_ == buf_p) {						\
		buf_p--;						\
	    } else {							\
		buf_p = P_;						\
	    }								\
	}								\
    } while (0)

    {
	/* Try GetAdaptersAddresses, if it is available */
	unsigned long ip_adaddrs_size = 16 * 1024;
	ULONG family = AF_UNSPEC;
	ULONG flags =
	    GAA_FLAG_INCLUDE_PREFIX | GAA_FLAG_SKIP_ANYCAST |
	    GAA_FLAG_SKIP_DNS_SERVER | GAA_FLAG_SKIP_FRIENDLY_NAME |
	    GAA_FLAG_SKIP_MULTICAST;
	ULONG (WINAPI *fpGetAdaptersAddresses)
	    (ULONG, ULONG, PVOID, PIP_ADAPTER_ADDRESSES, PULONG);
	HMODULE iphlpapi = GetModuleHandle("iphlpapi");
	fpGetAdaptersAddresses = (void *)
	    (iphlpapi ?
		GetProcAddress(iphlpapi, "GetAdaptersAddresses") :
		NULL);
	if (fpGetAdaptersAddresses) {
	    ip_adaddrs_p = ALLOC(ip_adaddrs_size);
	    for (i = 17;  i;  i--) {
		ret = fpGetAdaptersAddresses(
		    family, flags, NULL, ip_adaddrs_p, &ip_adaddrs_size);
		ip_adaddrs_p = REALLOC(ip_adaddrs_p, ip_adaddrs_size);
		if (ret == NO_ERROR) break;
		if (ret == ERROR_BUFFER_OVERFLOW) continue;
		i = 0;
	    }
	    if (! i) {
		FREE(ip_adaddrs_p);
		ip_adaddrs_p = NULL;
	    }
	} else ip_adaddrs_p = NULL;
    }

    {
	/* Load the IP_INTERFACE_INFO table (only IPv4 interfaces),
	 * reliable source of interface names on XP
	 */
	unsigned long info_size = 4 * 1024;
	info_p = ALLOC(info_size);
	for (i = 17;  i;  i--) {
	    ret = GetInterfaceInfo(info_p, &info_size);
	    info_p = REALLOC(info_p, info_size);
	    if (ret == NO_ERROR) break;
	    if (ret == ERROR_INSUFFICIENT_BUFFER) continue;
	    i = 0;
	}
	if (! i) {
	    FREE(info_p);
	    info_p = NULL;
	}
    }

    if (! ip_adaddrs_p) {
	/* If GetAdaptersAddresses gave nothing we fall back to
	 * MIB_IPADDRTABLE (only IPv4 interfaces)
	 */
	unsigned long ip_addrs_size = 16 * sizeof(*ip_addrs_p);
	ip_addrs_p = ALLOC(ip_addrs_size);
	for (i = 17;  i;  i--) {
	    ret = GetIpAddrTable(ip_addrs_p, &ip_addrs_size, FALSE);
	    ip_addrs_p = REALLOC(ip_addrs_p, ip_addrs_size);
	    if (ret == NO_ERROR) break;
	    if (ret == ERROR_INSUFFICIENT_BUFFER) continue;
	    i = 0;
	}
	if (! i) {
	    if (info_p) FREE(info_p);
	    FREE(ip_addrs_p);
	    return ctl_reply(INET_REP_OK, NULL, 0, rbuf_pp, rsize);
	}
    } else ip_addrs_p = NULL;

    buf_p = buf_alloc_p = ALLOC(buf_size);
    *buf_p++ = INET_REP_OK;

    /* Iterate over MIB_IPADDRTABLE or IP_ADAPTER_ADDRESSES */
    for (ia_p = NULL, ip_addrs_p ? ((void *)(i = 0)) : (ia_p = ip_adaddrs_p);
	 ip_addrs_p ? (i < ip_addrs_p->dwNumEntries) : (ia_p != NULL);
	 ip_addrs_p ? ((void *)(i++)) : (ia_p = ia_p->Next)) {
	MIB_IPADDRROW *ipaddrrow_p = NULL;
	DWORD flags = INET_IFF_MULTICAST;
	DWORD index = 0;
	WCHAR *wname_p = NULL;
	MIB_IFROW ifrow;

	if (ip_addrs_p) {
	    ipaddrrow_p = ip_addrs_p->table + i;
	    index = ipaddrrow_p->dwIndex;
	} else {
	    index = ia_p->IfIndex;
	    if (ia_p->Flags & IP_ADAPTER_NO_MULTICAST) {
		flags &= ~INET_IFF_MULTICAST;
	    }
	}
index:
	if (! index) goto done;
	sys_memzero(&ifrow, sizeof(ifrow));
	ifrow.dwIndex = index;
	if (GetIfEntry(&ifrow) != NO_ERROR) break;
	/* Find the interface name - first try MIB_IFROW.wzname */
	if (ifrow.wszName[0] != 0) {
	    wname_p = ifrow.wszName;
	} else {
	    /* Then try IP_ADAPTER_INDEX_MAP.Name (only IPv4 adapters) */
	    int j;
	    for (j = 0;  j < info_p->NumAdapters;  j++) {
		if (info_p->Adapter[j].Index == (ULONG) ifrow.dwIndex) {
		    if (info_p->Adapter[j].Name[0] != 0) {
			wname_p = info_p->Adapter[j].Name;
		    }
		    break;
		}
	    }
	}
	if (wname_p) {
	    int len;
	    /* Convert interface name to UTF-8 */
	    len =
		WideCharToMultiByte(
		    CP_UTF8, 0, wname_p, -1, NULL, 0, NULL, NULL);
	    if (! len) break;
	    BUF_ENSURE(len);
	    WideCharToMultiByte(
		CP_UTF8, 0, wname_p, -1, buf_p, len, NULL, NULL);
	    buf_p += len;
	} else {
	    /* Found no name -
	    * use "MIB_IFROW.dwIndex: MIB_IFROW.bDescr" as name instead */
	    int l;
	    l = utf8_len(ifrow.bDescr, ifrow.dwDescrLen);
	    BUF_ENSURE(9 + l+1);
	    buf_p +=
		erts_sprintf(
		    buf_p, "%lu: ", (unsigned long) ifrow.dwIndex);
	    utf8_encode(ifrow.bDescr, ifrow.dwDescrLen, buf_p);
	    buf_p += l;
	    *buf_p++ = '\0';
	}
	/* Interface flags, often make up broadcast and multicast flags */
	switch (ifrow.dwType) {
	case IF_TYPE_ETHERNET_CSMACD:
	    flags |= INET_IFF_BROADCAST;
	    break;
	case IF_TYPE_SOFTWARE_LOOPBACK:
	    flags |= INET_IFF_LOOPBACK;
	    flags &= ~INET_IFF_MULTICAST;
	    break;
	default:
	    flags &= ~INET_IFF_MULTICAST;
	    break;
	}
	if (ifrow.dwAdminStatus) {
	    flags |= INET_IFF_UP;
	    switch (ifrow.dwOperStatus) {
	    case IF_OPER_STATUS_CONNECTING:
		flags |= INET_IFF_POINTTOPOINT;
		break;
	    case IF_OPER_STATUS_CONNECTED:
		flags |= INET_IFF_RUNNING | INET_IFF_POINTTOPOINT;
		break;
	    case IF_OPER_STATUS_OPERATIONAL:
		flags |= INET_IFF_RUNNING;
		break;
	    }
	}
	BUF_ENSURE(1 + 4);
	*buf_p++ = INET_IFOPT_FLAGS;
	put_int32(flags, buf_p); buf_p += 4;
	if (ipaddrrow_p) {
	    /* Legacy implementation through GetIpAddrTable */
	    struct sockaddr_in sin;
	    /* IP Address */
	    sys_memzero(&sin, sizeof(sin));
	    sin.sin_family = AF_INET;
	    sin.sin_addr.s_addr = ipaddrrow_p->dwAddr;
	    BUF_ENSURE(1);
	    /* Netmask */
	    SOCKADDR_TO_BUF(INET_IFOPT_ADDR, (struct sockaddr *) &sin);
	    sin.sin_addr.s_addr = ipaddrrow_p->dwMask;
	    BUF_ENSURE(1);
	    SOCKADDR_TO_BUF(INET_IFOPT_NETMASK, (struct sockaddr *) &sin);
	    if (flags & INET_IFF_BROADCAST) {
		/* Broadcast address - fake it*/
		sin.sin_addr.s_addr = ipaddrrow_p->dwAddr;
		sin.sin_addr.s_addr |= ~ipaddrrow_p->dwMask;
		BUF_ENSURE(1);
		SOCKADDR_TO_BUF(
		    INET_IFOPT_BROADADDR, (struct sockaddr *) &sin);
	    }
	} else {
	    IP_ADAPTER_UNICAST_ADDRESS *p;
	    /* IP Address(es) */
	    for (p = ia_p->FirstUnicastAddress;
		p;
		p = p->Next)
	    {
		IP_ADAPTER_PREFIX *q;
		ULONG shortest_length;
		struct sockaddr *shortest_p, *sa_p = p->Address.lpSockaddr;
		BUF_ENSURE(1);
		SOCKADDR_TO_BUF(INET_IFOPT_ADDR, sa_p);
		shortest_p = NULL;
		shortest_length = 0;
		for (q = ia_p->FirstPrefix;
		     q;
		     q = q->Next) {
		    struct sockaddr *sp_p = q->Address.lpSockaddr;
		    if (sa_p->sa_family != sp_p->sa_family) continue;
		    switch (sa_p->sa_family) {
		    case AF_INET: {
			struct sockaddr_in sin;
			DWORD sa, sp, mask;
			sa = ntohl((DWORD)
				   ((struct sockaddr_in *)
				    sa_p)->sin_addr.s_addr);
			sp = ntohl((DWORD)
				   ((struct sockaddr_in *)
				    sp_p)->sin_addr.s_addr);
			mask = 0xFFFFFFFF << (32 - q->PrefixLength);
			if ((sa & mask) != (sp & mask)) continue;
			if ((! shortest_p)
			    || q->PrefixLength < shortest_length) {
			    shortest_p = sp_p;
			    shortest_length = q->PrefixLength;
			}
		    }   break;
		    case AF_INET6: {
			struct sockaddr_in6 sin6;
			if (!eq_masked_bytes((char *)
					     &((struct sockaddr_in6 *)
					       sa_p)->sin6_addr,
					     (char *)
					     &((struct sockaddr_in6 *)
					       sp_p)->sin6_addr,
					     q->PrefixLength)) {
			    continue;
			}
			if ((! shortest_p)
			    || q->PrefixLength < shortest_length) {
			    shortest_p = sp_p;
			    shortest_length = q->PrefixLength;
			}
		    }   break;
		    }
		}
		if (! shortest_p) {
		    /* Found no shortest prefix */
		    shortest_p = sa_p;
		    switch (shortest_p->sa_family) {
		    case AF_INET: {
			/* Fall back to old classfull network addresses */
			DWORD addr = ntohl(((struct sockaddr_in *)shortest_p)
					   ->sin_addr.s_addr);
			if (! (addr & 0x800000)) {
			    /* Class A */
			    shortest_length = 8;
			} else if (! (addr & 0x400000)) {
			    /* Class B */
			    shortest_length = 16;
			} else if (! (addr & 0x200000)) {
			    /* Class C */
			    shortest_length = 24;
			} else {
			    shortest_length = 32;
			}
		    }   break;
		    case AF_INET6: {
			/* Just play it safe */
			shortest_length = 128;
		    }   break;
		    }
		}
		switch (shortest_p->sa_family) {
		case AF_INET: {
		    struct sockaddr_in sin;
		    DWORD mask = 0xFFFFFFFF << (32 - shortest_length);
		    sys_memzero(&sin, sizeof(sin));
		    sin.sin_family = shortest_p->sa_family;
		    sin.sin_addr.s_addr = htonl(mask);
		    BUF_ENSURE(1);
		    SOCKADDR_TO_BUF(INET_IFOPT_NETMASK,
				    (struct sockaddr *) &sin);
		    if (flags & INET_IFF_BROADCAST) {
			DWORD sp =
			    ntohl((DWORD)
				  ((struct sockaddr_in *)shortest_p)
				  -> sin_addr.s_addr);
			sin.sin_addr.s_addr = htonl(sp | ~mask);
			BUF_ENSURE(1);
			SOCKADDR_TO_BUF(INET_IFOPT_BROADADDR,
					(struct sockaddr *) &sin);
		    }
		}   break;
		case AF_INET6: {
		    struct sockaddr_in6 sin6;
		    sys_memzero(&sin6, sizeof(sin6));
		    sin6.sin6_family = shortest_p->sa_family;
		    set_netmask_bytes((char *) &sin6.sin6_addr,
				      16,
				      shortest_length);
		    BUF_ENSURE(1);
		    SOCKADDR_TO_BUF(INET_IFOPT_NETMASK,
				    (struct sockaddr *) &sin6);
		}   break;
		}
	    }
	}
	if (ifrow.dwPhysAddrLen) {
	    /* Hardware Address */
	    BUF_ENSURE(1 + 2 + ifrow.dwPhysAddrLen);
	    *buf_p++ = INET_IFOPT_HWADDR;
	    put_int16(ifrow.dwPhysAddrLen, buf_p); buf_p += 2;
	    sys_memcpy(buf_p, ifrow.bPhysAddr, ifrow.dwPhysAddrLen);
	    buf_p += ifrow.dwPhysAddrLen;
	}

done:
	/* That is all for this interface */
	BUF_ENSURE(1);
	*buf_p++ = '\0';
	if (ia_p &&
	    ia_p->Ipv6IfIndex &&
	    ia_p->Ipv6IfIndex != index)
	{
	    /* Oops, there was another interface for IPv6. Possible? XXX */
	    index = ia_p->Ipv6IfIndex;
	    goto index;
	}
    }

    if (ip_adaddrs_p) FREE(ip_adaddrs_p);
    if (info_p) FREE(info_p);
    if (ip_addrs_p) FREE(ip_addrs_p);

    buf_size = buf_p - buf_alloc_p;
    buf_alloc_p = REALLOC(buf_alloc_p, buf_size);
    /* buf_p is now unreliable */
    *rbuf_pp = buf_alloc_p;
    return buf_size;
#   undef BUF_ENSURE
}

#elif defined(HAVE_GETIFADDRS)
#ifdef  DEBUG
#define GETIFADDRS_BUFSZ (1)
#else
#define GETIFADDRS_BUFSZ (512)
#endif

static ErlDrvSSizeT inet_ctl_getifaddrs(inet_descriptor* desc_p,
					char **rbuf_pp, ErlDrvSizeT rsize)
{
    struct ifaddrs *ifa_p, *ifa_free_p;

    ErlDrvSizeT buf_size;
    char *buf_p;
    char *buf_alloc_p;
    int save_errno;

    buf_size = GETIFADDRS_BUFSZ;
    buf_alloc_p = ALLOC(GETIFADDRS_BUFSZ);
    buf_p = buf_alloc_p;
#   define BUF_ENSURE(Size)						\
    do {								\
	int NEED_, GOT_ = buf_p - buf_alloc_p;				\
	NEED_ = GOT_ + (Size);						\
	if (NEED_ > buf_size) {						\
	    buf_size = NEED_ + GETIFADDRS_BUFSZ;			\
	    buf_alloc_p = REALLOC(buf_alloc_p, buf_size);		\
	    buf_p = buf_alloc_p + GOT_;					\
	}								\
    } while (0)
#   define SOCKADDR_TO_BUF(opt, sa)				        \
    do {    						                \
	if (sa) {                                                       \
	    char *P_;							\
	    *buf_p++ = (opt);						\
	    while (! (P_ = sockaddr_to_buf((sa), buf_p,			\
					   buf_alloc_p+buf_size))) {	\
		int GOT_ = buf_p - buf_alloc_p;				\
		buf_size += GETIFADDRS_BUFSZ;				\
		buf_alloc_p = REALLOC(buf_alloc_p, buf_size);		\
		buf_p = buf_alloc_p + GOT_;				\
	    }								\
	    if (P_ == buf_p) {						\
		buf_p--;						\
	    } else {							\
		buf_p = P_;						\
	    }								\
	}                                                               \
    } while (0)

    if ((save_errno = call_getifaddrs(desc_p, &ifa_p)) != 0)
        return ctl_error(save_errno, rbuf_pp, rsize);

    ifa_free_p = ifa_p;
    *buf_p++ = INET_REP_OK;
    for (;  ifa_p;  ifa_p = ifa_p->ifa_next) {
	int len = utf8_len(ifa_p->ifa_name, -1);
	BUF_ENSURE(len+1 + 1+4 + 1);
	utf8_encode(ifa_p->ifa_name, -1, buf_p);
	buf_p += len;
	*buf_p++ = '\0';
	*buf_p++ = INET_IFOPT_FLAGS;
	put_int32(IFGET_FLAGS(ifa_p->ifa_flags), buf_p); buf_p += 4;
	if (ifa_p->ifa_addr) {
	    if (ifa_p->ifa_addr->sa_family == AF_INET
#if defined(AF_INET6)
		|| ifa_p->ifa_addr->sa_family == AF_INET6
#endif
		) {
		SOCKADDR_TO_BUF(INET_IFOPT_ADDR, ifa_p->ifa_addr);
		if (ifa_p->ifa_netmask) {
		    BUF_ENSURE(1);
		    SOCKADDR_TO_BUF(INET_IFOPT_NETMASK, ifa_p->ifa_netmask);
		}
		if (ifa_p->ifa_dstaddr &&
		    (ifa_p->ifa_flags & IFF_POINTOPOINT)) {
		    BUF_ENSURE(1);
		    SOCKADDR_TO_BUF(INET_IFOPT_DSTADDR, ifa_p->ifa_dstaddr);
		} else if (ifa_p->ifa_broadaddr &&
			   (ifa_p->ifa_flags & IFF_BROADCAST)) {
		    BUF_ENSURE(1);
		    SOCKADDR_TO_BUF(INET_IFOPT_BROADADDR, ifa_p->ifa_broadaddr);
		}
	    }
#if defined(AF_LINK) || defined(AF_PACKET)
	    else if (
#if defined(AF_LINK)
		     ifa_p->ifa_addr->sa_family == AF_LINK
#else
		     0
#endif
#if defined(AF_PACKET)
		     || ifa_p->ifa_addr->sa_family == AF_PACKET
#endif
		     ) {
		size_t need = sockaddr_bufsz_need(ifa_p->ifa_addr);
		if (need > 3) {
		    BUF_ENSURE(1 + need);
		    SOCKADDR_TO_BUF(INET_IFOPT_HWADDR, ifa_p->ifa_addr);
		}
	    }
#endif
	}
	BUF_ENSURE(1);
	*buf_p++ = '\0';
    }
    buf_size = buf_p - buf_alloc_p;
    buf_alloc_p = REALLOC(buf_alloc_p, buf_size);
    /* buf_p is now unreliable */
    freeifaddrs(ifa_free_p);
    *rbuf_pp = buf_alloc_p;
    return buf_size;
#   undef BUF_ENSURE
}
#undef GETIFADDRS_BUFSZ

#else

static ErlDrvSSizeT inet_ctl_getifaddrs(inet_descriptor* desc_p,
					char **rbuf_pp, ErlDrvSizeT rsize)
{
    return ctl_error(ENOTSUP, rbuf_pp, rsize);
}

#endif

/* Per H @ Tail-f: The original code here had problems that possibly
   only occur if you abuse it for non-INET sockets, but anyway:
   a) If the getsockopt for SO_PRIORITY or IP_TOS failed, the actual
      requested setsockopt was never even attempted.
   b) If {get,set}sockopt for one of IP_TOS and SO_PRIORITY failed,
      but ditto for the other worked and that was actually the requested
      option, failure was still reported to erlang.                  */

#if  defined(IP_TOS) && defined(IPPROTO_IP)             \
    && defined(SO_PRIORITY) && !defined(__WIN32__)
static int setopt_prio_tos_trick
	(int fd, int proto, int type, char* arg_ptr, int arg_sz, int propagate)
{
    /* The relations between SO_PRIORITY, TOS and other options
       is not what you (or at least I) would expect...:
       If TOS is set after priority, priority is zeroed.
       If any other option is set after tos, tos might be zeroed.
       Therefore, save tos and priority. If something else is set,
       restore both after setting, if  tos is set, restore only
       prio and if prio is set restore none... All to keep the
       user feeling socket options are independent. /PaN */
    int          tmp_ival_prio;
    int          tmp_ival_tos;
    int          res;
    int          res_prio;
    int          res_tos;
    SOCKLEN_T    tmp_arg_sz_prio = sizeof(tmp_ival_prio);
    SOCKLEN_T    tmp_arg_sz_tos  = sizeof(tmp_ival_tos);

    res_prio = sock_getopt(fd, SOL_SOCKET, SO_PRIORITY,
		      (char *) &tmp_ival_prio, &tmp_arg_sz_prio);
    res_tos = sock_getopt(fd, IPPROTO_IP, IP_TOS,
		      (char *) &tmp_ival_tos, &tmp_arg_sz_tos);
	    res = sock_setopt(fd, proto, type, arg_ptr, arg_sz);
	    if (res == 0) {
		if (type != SO_PRIORITY) {
	    if (type != IP_TOS && res_tos == 0) {
		res_tos = sock_setopt(fd,
					  IPPROTO_IP,
					  IP_TOS,
					  (char *) &tmp_ival_tos,
					  tmp_arg_sz_tos);
		if (propagate)
		    res = res_tos;
		    }
	    if (res == 0 && res_prio == 0) {
		res_prio = sock_setopt(fd,
					   SOL_SOCKET,
					   SO_PRIORITY,
					   (char *) &tmp_ival_prio,
					   tmp_arg_sz_prio);
		if (propagate) {
		    /* Some kernels set a SO_PRIORITY by default that you are not permitted to reset,
		       silently ignore this error condition */
		    if (res_prio != 0 && sock_errno() == EPERM) {
			res = 0;
		    } else {
			res = res_prio;
		    }
		}
	    }
	}
    }
    return (res);
}
#endif

/* set socket options:
** return -1 on error
**         0 if ok
**         1 if ok force deliver of queued data
*/
#ifdef HAVE_SCTP
static int sctp_set_opts(inet_descriptor* desc, char* ptr, int len);
#endif

static int inet_set_opts(inet_descriptor* desc, char* ptr, int len)
{
    int type;
    int proto;
    int opt;
    struct linger li_val;
#if defined(HAVE_MULTICAST_SUPPORT) && defined(IPPROTO_IP)
    struct ip_mreq mreq_val;
#endif
    int ival;
    char* arg_ptr;
    int arg_sz;
#ifdef SO_BINDTODEVICE
    char ifname[IFNAMSIZ];
#endif
    enum PacketParseType old_htype = desc->htype;
    int old_active = desc->active;
    int propagate; /* Set to 1 if failure to set this option
		      should be propagated to erlang (not all
		      errors can be propagated for BC reasons) */
    int res;
#ifdef HAVE_SCTP
    /* SCTP sockets are treated completely separately: */
    if (IS_SCTP(desc))
	return sctp_set_opts(desc, ptr, len);
#endif
    /* XXX { int i; for(i=0;i<len;++i) fprintf(stderr,"0x%02X, ", (unsigned) ptr[i]); fprintf(stderr,"\r\n");} */

    while(len >= 5) {
        int recv_cmsgflags;

	opt = *ptr++;
	ival = get_int32(ptr);
	ptr += 4;
	len -= 5;
	arg_ptr = (char*) &ival;
	arg_sz = sizeof(ival);
	proto = SOL_SOCKET;
	propagate = 0;
        recv_cmsgflags = desc->recv_cmsgflags;

	switch(opt) {
	case INET_LOPT_HEADER:
	    DEBUGF(("inet_set_opts(%ld): s=%d, HEADER=%d\r\n",
		    (long)desc->port, desc->s,ival));
	    desc->hsz = ival;
	    continue;

	case INET_LOPT_MODE:
	    /* List or Binary: */
	    DEBUGF(("inet_set_opts(%ld): s=%d, MODE=%d\r\n",
		    (long)desc->port, desc->s, ival));
	    desc->mode = ival;
	    continue;

	case INET_LOPT_DELIVER:
	    DEBUGF(("inet_set_opts(%ld): s=%d, DELIVER=%d\r\n",
		    (long)desc->port, desc->s, ival));
	    desc->deliver = ival;
	    continue;

	case INET_LOPT_BUFFER:
	    DEBUGF(("inet_set_opts(%ld): s=%d, BUFFER=%d\r\n",
		    (long)desc->port, desc->s, ival));
	    if (ival < INET_MIN_BUFFER) ival = INET_MIN_BUFFER;
	    desc->bufsz = ival;
            desc->flags |= INET_FLG_BUFFER_SET;
	    continue;

	case INET_LOPT_ACTIVE:
	    DEBUGF(("inet_set_opts(%ld): s=%d, ACTIVE=%d\r\n",
		    (long)desc->port, desc->s, ival));
	    desc->active = ival;
            if (desc->active == INET_MULTI) {
                long ac = desc->active_count;
                Sint16 nval = get_int16(ptr);
                ptr += 2;
                len -= 2;
                ac += nval;
                if (ac > INT16_MAX || ac < INT16_MIN)
                    return -1;
                desc->active_count += nval;
                if (desc->active_count < 0)
                    desc->active_count = 0;
                if (desc->active_count == 0) {
                    desc->active = INET_PASSIVE;
                    packet_passive_message(desc);
                }
            } else
                desc->active_count = 0;
	    if ((desc->stype == SOCK_STREAM) && (desc->active != INET_PASSIVE) &&
		(desc->state == INET_STATE_CLOSED)) {
		tcp_descriptor *tdesc = (tcp_descriptor *) desc;
		if (tdesc->tcp_add_flags & TCP_ADDF_DELAYED_ECONNRESET) {
		    tdesc->tcp_add_flags &= ~TCP_ADDF_DELAYED_ECONNRESET;
		    tcp_error_message(tdesc, ECONNRESET);
		}
		tcp_closed_message(tdesc);
		if (desc->exitf) {
		    driver_exit(desc->port, 0);
		    return 0; /* Give up on this socket, descriptor lost */
		} else {
		    desc_close_read(desc);
		}
	    }
	    continue;

	case INET_LOPT_PACKET:
	    DEBUGF(("inet_set_opts(%ld): s=%d, PACKET=%d\r\n",
		    (long)desc->port, desc->s, ival));
	    desc->htype = ival;
	    continue;

	case INET_LOPT_PACKET_SIZE:
	    DEBUGF(("inet_set_opts(%ld): s=%d, PACKET_SIZE=%d\r\n",
		    (long)desc->port, desc->s, ival));
	    desc->psize = (unsigned int)ival;
	    continue;

	case INET_LOPT_EXITONCLOSE:
	    DEBUGF(("inet_set_opts(%ld): s=%d, EXITONCLOSE=%d\r\n",
		    (long)desc->port, desc->s, ival));
	    desc->exitf = ival;
	    continue;

	case INET_LOPT_TCP_HIWTRMRK:
	    if (desc->stype == SOCK_STREAM) {
		tcp_descriptor* tdesc = (tcp_descriptor*) desc;
		if (ival < 0) ival = 0;
		if (tdesc->low > ival)
		    tdesc->low = ival;
		tdesc->high = ival;
	    }
	    continue;

	case INET_LOPT_TCP_LOWTRMRK:
	    if (desc->stype == SOCK_STREAM) {
		tcp_descriptor* tdesc = (tcp_descriptor*) desc;
		if (ival < 0) ival = 0;
		if (tdesc->high < ival)
		    tdesc->high = ival;
		tdesc->low = ival;
	    }
	    continue;

	case INET_LOPT_MSGQ_HIWTRMRK: {
	    ErlDrvSizeT high;
	    if (ival < ERL_DRV_BUSY_MSGQ_LIM_MIN
		|| ERL_DRV_BUSY_MSGQ_LIM_MAX < ival)
		return -1;
	    high = (ErlDrvSizeT) ival;
	    erl_drv_busy_msgq_limits(desc->port, NULL, &high);
	    continue;
	}

	case INET_LOPT_MSGQ_LOWTRMRK: {
	    ErlDrvSizeT low;
	    if (ival < ERL_DRV_BUSY_MSGQ_LIM_MIN
		|| ERL_DRV_BUSY_MSGQ_LIM_MAX < ival)
		return -1;
	    low = (ErlDrvSizeT) ival;
	    erl_drv_busy_msgq_limits(desc->port, &low, NULL);
	    continue;
	}

	case INET_LOPT_TCP_SEND_TIMEOUT:
	    if (desc->stype == SOCK_STREAM) {
		tcp_descriptor* tdesc = (tcp_descriptor*) desc;
		tdesc->send_timeout = ival;
	    }
	    continue;

	case INET_LOPT_TCP_SEND_TIMEOUT_CLOSE:
	    if (desc->stype == SOCK_STREAM) {
		tcp_descriptor* tdesc = (tcp_descriptor*) desc;
		tdesc->send_timeout_close = ival;
	    }
	    continue;


	case INET_LOPT_TCP_DELAY_SEND:
	    if (desc->stype == SOCK_STREAM) {
		tcp_descriptor* tdesc = (tcp_descriptor*) desc;
		if (ival)
		    tdesc->tcp_add_flags |= TCP_ADDF_DELAY_SEND;
		else
		    tdesc->tcp_add_flags &= ~TCP_ADDF_DELAY_SEND;
	    }
	    continue;

#ifdef HAVE_UDP
	case INET_LOPT_UDP_READ_PACKETS:
	    if (desc->stype == SOCK_DGRAM) {
		udp_descriptor* udesc = (udp_descriptor*) desc;
		if (ival <= 0) return -1;
		udesc->read_packets = ival;
	    }
	    continue;
#endif

#ifdef HAVE_SETNS
	case INET_LOPT_NETNS:
	    /* It is annoying that ival and len are both (signed) int */
	    if (ival < 0) return -1;
	    if (len < ival) return -1;
	    if (desc->netns != NULL) FREE(desc->netns);
	    desc->netns = ALLOC(((unsigned int) ival) + 1);
	    memcpy(desc->netns, ptr, ival);
	    desc->netns[ival] = '\0';
	    ptr += ival;
	    len -= ival;
	    continue;
#endif

	case INET_LOPT_TCP_SHOW_ECONNRESET:
	    if (desc->sprotocol == IPPROTO_TCP) {
		tcp_descriptor* tdesc = (tcp_descriptor*) desc;
		if (ival)
		    tdesc->tcp_add_flags |= TCP_ADDF_SHOW_ECONNRESET;
		else
		    tdesc->tcp_add_flags &= ~TCP_ADDF_SHOW_ECONNRESET;
	    }
	    continue;

	case INET_LOPT_LINE_DELIM:
	    DEBUGF(("inet_set_opts(%ld): s=%d, LINE_DELIM=%d\r\n",
		    (long)desc->port, desc->s, ival));
	    desc->delimiter = (char)ival;
	    continue;

	case INET_OPT_REUSEADDR:
#ifdef __WIN32__
	    continue;  /* Bjorn says */
#else
	    type = SO_REUSEADDR;
	    DEBUGF(("inet_set_opts(%ld): s=%d, SO_REUSEADDR=%d\r\n",
		    (long)desc->port, desc->s,ival));
	    break;
#endif
	case INET_OPT_KEEPALIVE: type = SO_KEEPALIVE;
	    DEBUGF(("inet_set_opts(%ld): s=%d, SO_KEEPALIVE=%d\r\n",
		    (long)desc->port, desc->s, ival));
	    break;
	case INET_OPT_DONTROUTE: type = SO_DONTROUTE;
	    DEBUGF(("inet_set_opts(%ld): s=%d, SO_DONTROUTE=%d\r\n",
		    (long)desc->port, desc->s, ival));
	    break;
	case INET_OPT_BROADCAST: type = SO_BROADCAST;
	    DEBUGF(("inet_set_opts(%ld): s=%d, SO_BROADCAST=%d\r\n",
		    (long)desc->port, desc->s,ival));
	    break;
	case INET_OPT_OOBINLINE: type = SO_OOBINLINE;
	    DEBUGF(("inet_set_opts(%ld): s=%d, SO_OOBINLINE=%d\r\n",
		    (long)desc->port, desc->s, ival));
	    break;
	case INET_OPT_SNDBUF:    type = SO_SNDBUF;
	    DEBUGF(("inet_set_opts(%ld): s=%d, SO_SNDBUF=%d\r\n",
		    (long)desc->port, desc->s, ival));
	    break;
	case INET_OPT_RCVBUF:    type = SO_RCVBUF;
	    DEBUGF(("inet_set_opts(%ld): s=%d, SO_RCVBUF=%d\r\n",
		    (long)desc->port, desc->s, ival));
            if (!(desc->flags & INET_FLG_BUFFER_SET)) {
                /* make sure we have desc->bufsz >= SO_RCVBUF */
                if (ival > (1 << 16) && desc->stype == SOCK_DGRAM && !IS_SCTP(desc))
                    /* For UDP we don't want to automatically
                       set the buffer size to be larger than
                       the theoretical max MTU */
                    desc->bufsz = 1 << 16;
                else if (ival > desc->bufsz)
                    desc->bufsz = ival;
            }
	    break;
	case INET_OPT_LINGER:    type = SO_LINGER;
	    if (len < 4)
		return -1;
	    li_val.l_onoff = ival;
	    li_val.l_linger = get_int32(ptr);
	    ptr += 4;
	    len -= 4;
	    arg_ptr = (char*) &li_val;
	    arg_sz = sizeof(li_val);
	    DEBUGF(("inet_set_opts(%ld): s=%d, SO_LINGER=%d,%d",
		    (long)desc->port, desc->s, li_val.l_onoff,li_val.l_linger));
	    if (desc->sprotocol == IPPROTO_TCP) {
		tcp_descriptor* tdesc = (tcp_descriptor*) desc;
		if (li_val.l_onoff && li_val.l_linger == 0)
		    tdesc->tcp_add_flags |= TCP_ADDF_LINGER_ZERO;
		else
		    tdesc->tcp_add_flags &= ~TCP_ADDF_LINGER_ZERO;
	    }
	    break;

	case INET_OPT_PRIORITY:
#ifdef SO_PRIORITY
	    type = SO_PRIORITY;
	    propagate = 1; /* We do want to know if this fails */
	    DEBUGF(("inet_set_opts(%ld): s=%d, SO_PRIORITY=%d\r\n",
		    (long)desc->port, desc->s, ival));
	    break;
#else
            /* inet_fill_opts always returns a value for this option,
             * so we need to ignore it if not implemented */
	    continue;
#endif
	case INET_OPT_TOS:
#if defined(IP_TOS) && defined(IPPROTO_IP)
	    proto = IPPROTO_IP;
	    type = IP_TOS;
	    propagate = 1;
	    DEBUGF(("inet_set_opts(%ld): s=%d, IP_TOS=%d\r\n",
		    (long)desc->port, desc->s, ival));
	    break;
#else
            /* inet_fill_opts always returns a value for this option,
             * so we need to ignore it if not implemented. */
	    continue;
#endif
#if defined(IPV6_TCLASS) && defined(IPPROTO_IPV6)
	case INET_OPT_TCLASS:
	    proto = IPPROTO_IPV6;
	    type = IPV6_TCLASS;
	    propagate = 1;
	    DEBUGF(("inet_set_opts(%ld): s=%d, IPV6_TCLASS=%d\r\n",
		    (long)desc->port, desc->s, ival));
	    break;
#endif
#if defined(IP_TTL) && defined(IPPROTO_IP)
	case INET_OPT_TTL:
	    proto = IPPROTO_IP;
	    type = IP_TTL;
	    propagate = 1;
	    DEBUGF(("inet_set_opts(%ld): s=%d, IP_TTL=%d\r\n",
		    (long)desc->port, desc->s, ival));
	    break;
#endif
#if defined(IP_RECVTOS) && defined(IPPROTO_IP)
	case INET_OPT_RECVTOS:
	    proto = IPPROTO_IP;
	    type = IP_RECVTOS;
	    propagate = 1;
            recv_cmsgflags =
                ival ?
                (desc->recv_cmsgflags | INET_CMSG_RECVTOS) :
                (desc->recv_cmsgflags & ~INET_CMSG_RECVTOS);
	    DEBUGF(("inet_set_opts(%ld): s=%d, IP_RECVTOS=%d\r\n",
		    (long)desc->port, desc->s, ival));
	    break;
#endif
#if defined(IPV6_RECVTCLASS) && defined(IPPROTO_IPV6)
	case INET_OPT_RECVTCLASS:
	    proto = IPPROTO_IPV6;
	    type = IPV6_RECVTCLASS;
	    propagate = 1;
            recv_cmsgflags =
                ival ?
                (desc->recv_cmsgflags | INET_CMSG_RECVTCLASS) :
                (desc->recv_cmsgflags & ~INET_CMSG_RECVTCLASS);
	    DEBUGF(("inet_set_opts(%ld): s=%d, IPV6_RECVTCLASS=%d\r\n",
		    (long)desc->port, desc->s, ival));
	    break;
#endif
#if defined(IP_RECVTTL) && defined(IPPROTO_IP)
	case INET_OPT_RECVTTL:
	    proto = IPPROTO_IP;
	    type = IP_RECVTTL;
	    propagate = 1;
            recv_cmsgflags =
                ival ?
                (desc->recv_cmsgflags | INET_CMSG_RECVTTL) :
                (desc->recv_cmsgflags & ~INET_CMSG_RECVTTL);
	    DEBUGF(("inet_set_opts(%ld): s=%d, IP_RECVTTL=%d\r\n",
		    (long)desc->port, desc->s, ival));
	    break;
#endif

	case TCP_OPT_NODELAY:
	    proto = IPPROTO_TCP;
	    type = TCP_NODELAY;
	    DEBUGF(("inet_set_opts(%ld): s=%d, TCP_NODELAY=%d\r\n",
		    (long)desc->port, desc->s, ival));
	    break;

	case TCP_OPT_NOPUSH:
#if defined(INET_TCP_NOPUSH)
	    proto = IPPROTO_TCP;
	    type = INET_TCP_NOPUSH;
	    DEBUGF(("inet_set_opts(%ld): s=%d, t=%d TCP_NOPUSH=%d\r\n",
	            (long)desc->port, desc->s, type, ival));
	    break;
#else
	    /* inet_fill_opts always returns a value for this option,
	     * so we need to ignore it if not implemented, just in case */
	    continue;
#endif

#if defined(HAVE_MULTICAST_SUPPORT) && defined(IPPROTO_IP)

	case UDP_OPT_MULTICAST_TTL:
	    proto = IPPROTO_IP;
	    type = IP_MULTICAST_TTL;
	    DEBUGF(("inet_set_opts(%ld): s=%d, IP_MULTICAST_TTL=%d\r\n",
		    (long)desc->port,desc->s,ival));
	    break;

	case UDP_OPT_MULTICAST_LOOP:
	    proto = IPPROTO_IP;
	    type = IP_MULTICAST_LOOP;
	    DEBUGF(("inet_set_opts(%ld): s=%d, IP_MULTICAST_LOOP=%d\r\n",
		    (long)desc->port,desc->s,ival));
	    break;

	case UDP_OPT_MULTICAST_IF:
	    proto = IPPROTO_IP;
	    type = IP_MULTICAST_IF;
	    DEBUGF(("inet_set_opts(%ld): s=%d, IP_MULTICAST_IF=%x\r\n",
		    (long)desc->port, desc->s, ival));
	    ival = sock_htonl(ival);
	    break;

	case UDP_OPT_ADD_MEMBERSHIP:
	    proto = IPPROTO_IP;
	    type = IP_ADD_MEMBERSHIP;
	    DEBUGF(("inet_set_opts(%ld): s=%d, IP_ADD_MEMBERSHIP=%d\r\n",
		    (long)desc->port, desc->s,ival));
	    goto L_set_mreq;

	case UDP_OPT_DROP_MEMBERSHIP:
	    proto = IPPROTO_IP;
	    type = IP_DROP_MEMBERSHIP;
	    DEBUGF(("inet_set_opts(%ld): s=%d, IP_DROP_MEMBERSHIP=%x\r\n",
		    (long)desc->port, desc->s, ival));
	L_set_mreq:
	    mreq_val.imr_multiaddr.s_addr = sock_htonl(ival);
	    ival = get_int32(ptr);
	    mreq_val.imr_interface.s_addr = sock_htonl(ival);
	    ptr += 4;
	    len -= 4;
	    arg_ptr = (char*)&mreq_val;
	    arg_sz = sizeof(mreq_val);
	    break;

#endif /* defined(HAVE_MULTICAST_SUPPORT) && defined(IPPROTO_IP) */

	case INET_OPT_IPV6_V6ONLY:
#if HAVE_DECL_IPV6_V6ONLY && defined(IPPROTO_IPV6)
	    proto = IPPROTO_IPV6;
	    type = IPV6_V6ONLY;
	    propagate = 1;
	    DEBUGF(("inet_set_opts(%ld): s=%d, IPV6_V6ONLY=%d\r\n",
		    (long)desc->port, desc->s, ival));
	    break;
#elif defined(__WIN32__) && defined(HAVE_IN6) && defined(AF_INET6)
	    /* Fake a'la OpenBSD; set to 'true' is fine but 'false' invalid. */
	    if (ival != 0) continue;
	    else return -1;
	    break;
#else
	    continue;
#endif

	case INET_OPT_RAW:
	    if (len < 8) {
		return -1;
	    }
	    proto = ival;
	    type = get_int32(ptr);
	    ptr += 4;
	    arg_sz = get_int32(ptr);
	    ptr += 4;
	    len -= 8;
	    if (len < arg_sz) {
		return -1;
	    }
	    arg_ptr = ptr;
	    ptr += arg_sz;
	    len -= arg_sz;
	    break;

#ifdef SO_BINDTODEVICE
	case INET_OPT_BIND_TO_DEVICE:
	    if (ival < 0) return -1;
	    if (len < ival) return -1;
	    if (ival > sizeof(ifname)) {
		return -1;
	    }
	    memcpy(ifname, ptr, ival);
	    ifname[ival] = '\0';
	    ptr += ival;
	    len -= ival;

	    proto = SOL_SOCKET;
	    type = SO_BINDTODEVICE;
	    arg_ptr = (char*)&ifname;
	    arg_sz = sizeof(ifname);
	    propagate = 1; /* We do want to know if this fails */

	    DEBUGF(("inet_set_opts(%ld): s=%d, SO_BINDTODEVICE=%s\r\n",
		    (long)desc->port, desc->s, ifname));
	    break;
#endif

	default:
	    return -1;
	}
#if  defined(IP_TOS) && defined(IPPROTO_IP) \
    && defined(SO_PRIORITY) && !defined(__WIN32__)
	res = setopt_prio_tos_trick (desc->s, proto, type, arg_ptr, arg_sz, propagate);
#else
	res = sock_setopt	    (desc->s, proto, type, arg_ptr, arg_sz);
#endif
        if (res == 0) desc->recv_cmsgflags = recv_cmsgflags;
	if (propagate && res != 0) {
	    return -1;
	}
	DEBUGF(("inet_set_opts(%ld): s=%d returned %d\r\n",
		(long)desc->port, desc->s, res));
    }

    if ( ((desc->stype == SOCK_STREAM) && IS_CONNECTED(desc)) ||
	((desc->stype == SOCK_DGRAM) && IS_OPEN(desc))) {

	if (desc->active != old_active) {
            /* Need to cancel the read_packet timer if we go from active to passive. */
            if (desc->active == INET_PASSIVE && desc->stype == SOCK_DGRAM)
                driver_cancel_timer(desc->port);
	    sock_select(desc, (FD_READ|FD_CLOSE), (desc->active>0));
        }

	/* XXX: UDP sockets could also trigger immediate read here NIY */
	if ((desc->stype==SOCK_STREAM) && desc->active) {
	    if (!old_active || (desc->htype != old_htype)) {
		/* passive => active change OR header type change in active mode */
		/* Return > 1 if only active changed to INET_ONCE -> direct read if
		   header type is unchanged. */
		/* XXX fprintf(stderr,"desc->htype == %d, old_htype == %d,
		   desc->active == %d, old_active == %d\r\n",(int)desc->htype,
		   (int) old_htype, (int) desc->active, (int) old_active );*/
		return 1+(desc->htype == old_htype &&
                          (desc->active == INET_ONCE || desc->active == INET_MULTI));
	    }
	    return 0;
	}
    }
    return 0;
}

#ifdef HAVE_SCTP

/*  "sctp_get_initmsg":
**  Used by both "send*" and "setsockopt". Gets the 4 fields of "sctp_initmsg"
**  from the input buffer:
*/
#define SCTP_GET_INITMSG_LEN (4*2)
static char* sctp_get_initmsg(struct sctp_initmsg* ini, char* curr)
{
    ini->sinit_num_ostreams   = get_int16 (curr);	curr += 2;
    ini->sinit_max_instreams  = get_int16 (curr);	curr += 2;
    ini->sinit_max_attempts   = get_int16 (curr);	curr += 2;
    ini->sinit_max_init_timeo = get_int16 (curr);	curr += 2;
    return curr;
}

/*  "sctp_get_sendparams":
**  Parses (from the command buffer) the 6 user-sprcified parms of
**  "sctp_sndrcvinfo":
**	stream(u16),      flags(u16), ppid(u32), context(u32),
**	timetoleave(u32), assoc_id
**  Is used by both "send*" and "setsockopt":
*/
#define SCTP_GET_SENDPARAMS_LEN (2*2 + 3*4 + ASSOC_ID_LEN)
static char* sctp_get_sendparams (struct sctp_sndrcvinfo* sri, char* curr)
{
    int eflags;
    int cflags;

    sri->sinfo_stream       = get_int16(curr);		curr += 2;
    sri->sinfo_ssn	    = 0;

    /* The "flags" are already ORed at the Erlang side, here we
       reconstruct the real SCTP flags:
    */
    eflags		    = get_int16(curr);		curr += 2;
    cflags		    = 0;
    if (eflags & SCTP_FLAG_UNORDERED) cflags |= SCTP_UNORDERED;
    if (eflags & SCTP_FLAG_ADDR_OVER) cflags |= SCTP_ADDR_OVER;
    if (eflags & SCTP_FLAG_ABORT)     cflags |= SCTP_ABORT;
    if (eflags & SCTP_FLAG_EOF)	      cflags |= SCTP_EOF;

    sri->sinfo_flags	    = cflags;
    sri->sinfo_ppid         = sock_htonl(get_int32(curr));
							curr += 4;
    sri->sinfo_context      = get_int32(curr);		curr += 4;
    sri->sinfo_timetolive   = get_int32(curr);		curr += 4;
    sri->sinfo_tsn	    = 0;
    sri->sinfo_cumtsn	    = 0;
    sri->sinfo_assoc_id	    = GET_ASSOC_ID  (curr);	curr += ASSOC_ID_LEN;

    return curr;
}

/* Set SCTP options:
** return -1 on error
**         0 if ok
** NB: unlike inet_set_opts(), we don't have an active mode here, so there is no
** mode change which could force data delivery on setting an option.
** Arg: "ptr": [(erlang_encoded_opt(u8), value(...)), ...];  thus, multiple opts
** can be set at a time.
*/
static int sctp_set_opts(inet_descriptor* desc, char* ptr, int len)
{
#   define CHKLEN(Ptr, Len)                        \
    do {                                           \
	if ((Ptr) + (Len) > ptr + len) return -1; \
    } while (0)

    char * curr = ptr;
    int    proto, type, res;

    /* The following union is used to hold any arg to "setsockopt": */
    union  opts_union
    {
	int			    ival;
	struct sctp_rtoinfo	    rtoi;
	struct sctp_assocparams	    ap;
	struct sctp_initmsg	    im;
	struct linger		    lin;
	struct sctp_setpeerprim	    prim;
	struct sctp_setadaptation   ad;
	struct sctp_paddrparams	    pap;
	struct sctp_sndrcvinfo	    sri;
	struct sctp_event_subscribe es;
#	ifdef SCTP_DELAYED_ACK_TIME
	struct sctp_assoc_value     av; /* Not in SOLARIS10 */
#	endif
#	ifdef SO_BINDTODEVICE
	char ifname[IFNAMSIZ];
#	endif
    }
    arg;

    char * arg_ptr = NULL;
    int    arg_sz  = 0;
    int    old_active = desc->active;

    while (curr < ptr + len)
    {
        int recv_cmsgflags;
	/* Get the Erlang-encoded option type -- always 1 byte: */
	int eopt;

        eopt = *curr;
	curr++;

        recv_cmsgflags = desc->recv_cmsgflags;
	/* Get the option value.  XXX: The condition  (curr < ptr + len)
	   does not preclude us from reading from beyond the buffer end,
	   if the Erlang part of the driver specifies its input wrongly!
	*/
	CHKLEN(curr, 4); /* All options need at least 4 bytes */
	switch(eopt)
	{
	/* Local INET options: */

	case INET_LOPT_BUFFER:
	    desc->bufsz  = get_int32(curr);		curr += 4;

	    if (desc->bufsz < INET_MIN_BUFFER)
		desc->bufsz = INET_MIN_BUFFER;
            desc->flags |= INET_FLG_BUFFER_SET;
	    res = 0;	  /* This does not affect the kernel buffer size */
	    continue;

	case INET_LOPT_MODE:
	    desc->mode   = get_int32(curr);		curr += 4;
	    res = 0;
	    continue;

	case INET_LOPT_ACTIVE:
	    desc->active = get_int32(curr);		curr += 4;
            if (desc->active == INET_MULTI) {
                long ac = desc->active_count;
                Sint16 nval = get_int16(curr);          curr += 2;
		ac += nval;
                if (ac > INT16_MAX || ac < INT16_MIN)
                    return -1;
                desc->active_count += nval;
                if (desc->active_count < 0)
                    desc->active_count = 0;
                if (desc->active_count == 0) {
                    desc->active = INET_PASSIVE;
                    packet_passive_message(desc);
                }
            } else
                desc->active_count = 0;
	    res = 0;
	    continue;

#ifdef HAVE_SETNS
	case INET_LOPT_NETNS:
	{
	    size_t ns_len;
	    ns_len = get_int32(curr);                   curr += 4;
	    CHKLEN(curr, ns_len);
	    if (desc->netns != NULL) FREE(desc->netns);
	    desc->netns = ALLOC(ns_len + 1);
	    memcpy(desc->netns, curr, ns_len);
	    desc->netns[ns_len] = '\0';
	    curr += ns_len;
	}
	    continue;
#endif

	/* SCTP options and applicable generic INET options: */

	case SCTP_OPT_RTOINFO:
	{
	    CHKLEN(curr, ASSOC_ID_LEN + 3*4);
	    arg.rtoi.srto_assoc_id = GET_ASSOC_ID(curr);  curr += ASSOC_ID_LEN;
	    arg.rtoi.srto_initial  = get_int32   (curr);  curr += 4;
	    arg.rtoi.srto_max      = get_int32   (curr);  curr += 4;
	    arg.rtoi.srto_min      = get_int32   (curr);  curr += 4;

	    proto   = IPPROTO_SCTP;
	    type    = SCTP_RTOINFO;
	    arg_ptr = (char*) (&arg.rtoi);
	    arg_sz  = sizeof  ( arg.rtoi);
	    break;
	}
	case SCTP_OPT_ASSOCINFO:
	{
	    CHKLEN(curr, ASSOC_ID_LEN + 2*2 + 3*4);

	    arg.ap.sasoc_assoc_id    = GET_ASSOC_ID(curr); curr += ASSOC_ID_LEN;
	    arg.ap.sasoc_asocmaxrxt  = get_int16   (curr); curr += 2;
	    arg.ap.sasoc_number_peer_destinations =
				       get_int16   (curr); curr += 2;
	    arg.ap.sasoc_peer_rwnd   = get_int32   (curr); curr += 4;
	    arg.ap.sasoc_local_rwnd  = get_int32   (curr); curr += 4;
	    arg.ap.sasoc_cookie_life = get_int32   (curr); curr += 4;

	    proto   = IPPROTO_SCTP;
	    type    = SCTP_ASSOCINFO;
	    arg_ptr = (char*) (&arg.ap);
	    arg_sz  = sizeof  ( arg.ap);
	    break;
	}
	case SCTP_OPT_INITMSG:
	{
	    CHKLEN(curr, SCTP_GET_INITMSG_LEN);
	    curr  = sctp_get_initmsg (&arg.im, curr);

	    proto   = IPPROTO_SCTP;
	    type    = SCTP_INITMSG;
	    arg_ptr = (char*) (&arg.im);
	    arg_sz  = sizeof  ( arg.im);
	    break;
	}
	case INET_OPT_LINGER:
	{
	    CHKLEN(curr, 2*4);
	    arg.lin.l_onoff  = get_int32 (curr);  curr += 4;
	    arg.lin.l_linger = get_int32 (curr);  curr += 4;

	    proto   = SOL_SOCKET;
	    type    = SO_LINGER;
	    arg_ptr = (char*) (&arg.lin);
	    arg_sz  = sizeof  ( arg.lin);
	    break;
	}
	case SCTP_OPT_NODELAY:
	{
	    arg.ival= get_int32 (curr);	  curr += 4;
	    proto   = IPPROTO_SCTP;
	    type    = SCTP_NODELAY;
	    arg_ptr = (char*) (&arg.ival);
	    arg_sz  = sizeof  ( arg.ival);
	    break;
	}
	case INET_OPT_RCVBUF:
	{
	    arg.ival= get_int32 (curr);	  curr += 4;
	    proto   = SOL_SOCKET;
	    type    = SO_RCVBUF;
	    arg_ptr = (char*) (&arg.ival);
	    arg_sz  = sizeof  ( arg.ival);

	    /* Adjust the size of the user-level recv buffer, so it's not
	       smaller than the kernel one: */
	    if (desc->bufsz <= arg.ival)
		desc->bufsz  = arg.ival;
            desc->flags |= INET_FLG_BUFFER_SET;
	    break;
	}
	case INET_OPT_SNDBUF:
	{
	    arg.ival= get_int32 (curr);	  curr += 4;
	    proto   = SOL_SOCKET;
	    type    = SO_SNDBUF;
	    arg_ptr = (char*) (&arg.ival);
	    arg_sz  = sizeof  ( arg.ival);

	    break;
	}
	case INET_OPT_REUSEADDR:
	{
	    arg.ival= get_int32 (curr);	  curr += 4;
	    proto   = SOL_SOCKET;
	    type    = SO_REUSEADDR;
	    arg_ptr = (char*) (&arg.ival);
	    arg_sz  = sizeof  ( arg.ival);
	    break;
	}
	case INET_OPT_DONTROUTE:
	{
	    arg.ival= get_int32 (curr);	  curr += 4;
	    proto   = SOL_SOCKET;
	    type    = SO_DONTROUTE;
	    arg_ptr = (char*) (&arg.ival);
	    arg_sz  = sizeof  ( arg.ival);
	    break;
	}
	case INET_OPT_PRIORITY:
#	ifdef SO_PRIORITY
	{
	    arg.ival= get_int32 (curr);	  curr += 4;
	    proto   = SOL_SOCKET;
	    type    = SO_PRIORITY;
	    arg_ptr = (char*) (&arg.ival);
	    arg_sz  = sizeof  ( arg.ival);
	    break;
	}
#	else
        /* inet_fill_opts always returns a value for this option,
         * so we need to ignore it if not implemented, just in case */
	    continue;
#	endif

	case INET_OPT_TOS:
#	if defined(IP_TOS) && defined(IPPROTO_IP)
	{
	    arg.ival= get_int32 (curr);	  curr += 4;
	    proto   = IPPROTO_IP;
	    type    = IP_TOS;
	    arg_ptr = (char*) (&arg.ival);
	    arg_sz  = sizeof  ( arg.ival);
	    break;
	}
#	else
        /* inet_fill_opts always returns a value for this option,
         * so we need to ignore it if not implemented, just in case */
	    continue;
#	endif

#       if defined(IPV6_TCLASS) && defined(IPPROTO_IPV6)
	case INET_OPT_TCLASS:
	{
	    arg.ival= get_int32 (curr);	  curr += 4;
	    proto   = IPPROTO_IPV6;
	    type    = IPV6_TCLASS;
	    arg_ptr = (char*) (&arg.ival);
	    arg_sz  = sizeof  ( arg.ival);
	    break;
	}
#	endif

#       if defined(IP_TTL) && defined(IPPROTO_IP)
	case INET_OPT_TTL:
	{
	    arg.ival= get_int32 (curr);	  curr += 4;
	    proto   = IPPROTO_IP;
	    type    = IP_TTL;
	    arg_ptr = (char*) (&arg.ival);
	    arg_sz  = sizeof  ( arg.ival);
	    break;
	}
#	endif

#	if defined(IP_RECVTOS) && defined(IPPROTO_IP)
	case INET_OPT_RECVTOS:
	{
	    arg.ival= get_int32 (curr);	  curr += 4;
	    proto   = IPPROTO_IP;
	    type    = IP_RECVTOS;
	    arg_ptr = (char*) (&arg.ival);
	    arg_sz  = sizeof  ( arg.ival);
            recv_cmsgflags =
                arg.ival ?
                (desc->recv_cmsgflags | INET_CMSG_RECVTOS) :
                (desc->recv_cmsgflags & ~INET_CMSG_RECVTOS);
	    break;
	}
#	endif

#       if defined(IPV6_RECVTCLASS) && defined(IPPROTO_IPV6)
	case INET_OPT_RECVTCLASS:
	{
	    arg.ival= get_int32 (curr);	  curr += 4;
	    proto   = IPPROTO_IPV6;
	    type    = IPV6_RECVTCLASS;
	    arg_ptr = (char*) (&arg.ival);
	    arg_sz  = sizeof  ( arg.ival);
            recv_cmsgflags =
                arg.ival ?
                (desc->recv_cmsgflags | INET_CMSG_RECVTCLASS) :
                (desc->recv_cmsgflags & ~INET_CMSG_RECVTCLASS);
	    break;
	}
#	endif

#	if defined(IP_RECVTTL) && defined(IPPROTO_IP)
	case INET_OPT_RECVTTL:
	{
	    arg.ival= get_int32 (curr);	  curr += 4;
	    proto   = IPPROTO_IP;
	    type    = IP_RECVTTL;
	    arg_ptr = (char*) (&arg.ival);
	    arg_sz  = sizeof  ( arg.ival);
            recv_cmsgflags =
                arg.ival ?
                (desc->recv_cmsgflags | INET_CMSG_RECVTTL) :
                (desc->recv_cmsgflags & ~INET_CMSG_RECVTTL);
	    break;
	}
#	endif


	case INET_OPT_IPV6_V6ONLY:
#       if HAVE_DECL_IPV6_V6ONLY && defined(IPPROTO_IPV6)
	{
	    arg.ival= get_int32 (curr);   curr += 4;
	    proto   = IPPROTO_IPV6;
	    type    = IPV6_V6ONLY;
	    arg_ptr = (char*) (&arg.ival);
	    arg_sz  = sizeof  ( arg.ival);
	    break;
	}
#       elif defined(__WIN32__) && defined(HAVE_IN6) && defined(AF_INET6)
#           error Here is a fix for Win IPv6 SCTP missing
#       else
	    continue; /* Option not supported -- ignore it */
#       endif

#ifdef SO_BINDTODEVICE
	case INET_OPT_BIND_TO_DEVICE:
	    arg_sz = get_int32(curr);			curr += 4;
	    CHKLEN(curr, arg_sz);
	    if (arg_sz >= sizeof(arg.ifname))
		return -1;
	    memcpy(arg.ifname, curr, arg_sz);
	    arg.ifname[arg_sz] = '\0';
	    curr += arg_sz;

	    proto   = SOL_SOCKET;
	    type    = SO_BINDTODEVICE;
	    arg_ptr = (char*) (&arg.ifname);
	    arg_sz  = sizeof  ( arg.ifname);
	    break;
#endif

	case SCTP_OPT_AUTOCLOSE:
	{
	    arg.ival= get_int32 (curr);	  curr += 4;
	    proto   = IPPROTO_SCTP;
	    type    = SCTP_AUTOCLOSE;
	    arg_ptr = (char*) (&arg.ival);
	    arg_sz  = sizeof  ( arg.ival);
	    break;
	}
	case SCTP_OPT_DISABLE_FRAGMENTS:
	{
	    arg.ival= get_int32 (curr);	  curr += 4;
	    proto   = IPPROTO_SCTP;
	    type    = SCTP_DISABLE_FRAGMENTS;
	    arg_ptr = (char*) (&arg.ival);
	    arg_sz  = sizeof  ( arg.ival);
	    break;
	}
	case SCTP_OPT_I_WANT_MAPPED_V4_ADDR:
	{
	    arg.ival= get_int32 (curr);	  curr += 4;
	    proto   = IPPROTO_SCTP;
	    type    = SCTP_I_WANT_MAPPED_V4_ADDR;
	    arg_ptr = (char*) (&arg.ival);
	    arg_sz  = sizeof  ( arg.ival);
	    break;
	}
	case SCTP_OPT_MAXSEG:
	{
	    arg.ival= get_int32 (curr);	  curr += 4;
	    proto   = IPPROTO_SCTP;
	    type    = SCTP_MAXSEG;
	    arg_ptr = (char*) (&arg.ival);
	    arg_sz  = sizeof  ( arg.ival);
	    break;
	}
	case SCTP_OPT_PRIMARY_ADDR:
	case SCTP_OPT_SET_PEER_PRIMARY_ADDR:
	{
	    ErlDrvSizeT alen;

	    CHKLEN(curr, ASSOC_ID_LEN);
	    /* XXX: These 2 opts have isomorphic value data structures,
	       "sctp_setpeerprim" and "sctp_prim" (in Solaris 10, the latter
	       is called "sctp_setprim"),  so we grouped them together:
	    */
	    arg.prim.sspp_assoc_id = GET_ASSOC_ID(curr); curr += ASSOC_ID_LEN;

	    /* Fill in "arg.prim.sspp_addr": */
	    alen  = ptr + len - curr;
	    if (inet_set_faddress
		(desc->sfamily, (inet_address*) (&arg.prim.sspp_addr),
		 &curr,  &alen) != NULL) return -1;

	    proto = IPPROTO_SCTP;
	    if (eopt == SCTP_OPT_PRIMARY_ADDR)
		type =  SCTP_PRIMARY_ADDR;
	    else
		type =  SCTP_SET_PEER_PRIMARY_ADDR;

	    arg_ptr  =  (char*) (&arg.prim);
	    arg_sz   =  sizeof  ( arg.prim);
	    break;
	}
	case SCTP_OPT_ADAPTATION_LAYER:
	{
	    /* XXX: do we need to convert the Ind into network byte order??? */
	    arg.ad.ssb_adaptation_ind = sock_htonl (get_int32(curr));  curr += 4;

	    proto   = IPPROTO_SCTP;
	    type    = SCTP_ADAPTATION_LAYER;
	    arg_ptr = (char*) (&arg.ad);
	    arg_sz  = sizeof  ( arg.ad);
	    break;
	}
	case SCTP_OPT_PEER_ADDR_PARAMS:
	{
	    ErlDrvSizeT alen;
#	    ifdef HAVE_STRUCT_SCTP_PADDRPARAMS_SPP_FLAGS
	    int eflags, cflags, hb_enable, hb_disable,
		pmtud_enable, pmtud_disable;
#	    ifdef HAVE_STRUCT_SCTP_PADDRPARAMS_SPP_SACKDELAY
	    int
		sackdelay_enable, sackdelay_disable;
#           endif
#           endif

	    CHKLEN(curr, ASSOC_ID_LEN);
	    arg.pap.spp_assoc_id = GET_ASSOC_ID(curr);	curr += ASSOC_ID_LEN;

	    /* Fill in "pap.spp_address": */
	    alen  = ptr + len - curr;
	    if (inet_set_faddress
		(desc->sfamily, (inet_address*) (&arg.pap.spp_address),
		 &curr,  &alen) != NULL) return -1;

	    CHKLEN(curr, 4 + 2 + 3*4);

	    arg.pap.spp_hbinterval = get_int32(curr);	curr += 4;
	    arg.pap.spp_pathmaxrxt = get_int16(curr);	curr += 2;

	    /* The following are missing in Solaris 10: */
#	    ifdef HAVE_STRUCT_SCTP_PADDRPARAMS_SPP_PATHMTU
	    arg.pap.spp_pathmtu    = get_int32(curr);
#           endif
	    curr += 4;
#	    ifdef HAVE_STRUCT_SCTP_PADDRPARAMS_SPP_SACKDELAY
	    arg.pap.spp_sackdelay  = get_int32(curr);
#           endif
	    curr += 4;

#	    ifdef HAVE_STRUCT_SCTP_PADDRPARAMS_SPP_FLAGS
	    /* Now re-construct the flags: */
	    eflags	       = get_int32(curr);
	    cflags	       = 0;

	    hb_enable      = eflags & SCTP_FLAG_HB_ENABLE;
	    hb_disable     = eflags & SCTP_FLAG_HB_DISABLE;
	    if (hb_enable && hb_disable)
		return -1;
	    if (hb_enable)	 		cflags |= SPP_HB_ENABLE;
	    if (hb_disable)	 		cflags |= SPP_HB_DISABLE;
	    if (eflags & SCTP_FLAG_HB_DEMAND)	cflags |= SPP_HB_DEMAND;

	    pmtud_enable   = eflags & SCTP_FLAG_PMTUD_ENABLE;
	    pmtud_disable  = eflags & SCTP_FLAG_PMTUD_DISABLE;
	    if (pmtud_enable && pmtud_disable)
		return -1;
	    if (pmtud_enable)			cflags |= SPP_PMTUD_ENABLE;
	    if (pmtud_disable)			cflags |= SPP_PMTUD_DISABLE;

#	    ifdef HAVE_STRUCT_SCTP_PADDRPARAMS_SPP_SACKDELAY
	    /* The followings are missing in FreeBSD 7.1 */
	    sackdelay_enable =eflags& SCTP_FLAG_SACDELAY_ENABLE;
	    sackdelay_disable=eflags& SCTP_FLAG_SACDELAY_DISABLE;
	    if (sackdelay_enable && sackdelay_disable)
		return -1;
	    if (sackdelay_enable)		cflags |= SPP_SACKDELAY_ENABLE;
	    if (sackdelay_disable)		cflags |= SPP_SACKDELAY_DISABLE;
#           endif

	    arg.pap.spp_flags  = cflags;
#	    endif
	    curr += 4;

	    proto   = IPPROTO_SCTP;
	    type    = SCTP_PEER_ADDR_PARAMS;
	    arg_ptr = (char*) (&arg.pap);
	    arg_sz  = sizeof  ( arg.pap);
	    break;
	}
	case SCTP_OPT_DEFAULT_SEND_PARAM:
	{
	    CHKLEN(curr, SCTP_GET_SENDPARAMS_LEN);
	    curr = sctp_get_sendparams (&arg.sri, curr);

	    proto   = IPPROTO_SCTP;
	    type    = SCTP_DEFAULT_SEND_PARAM;
	    arg_ptr = (char*) (&arg.sri);
	    arg_sz  = sizeof  ( arg.sri);
	    VALGRIND_MAKE_MEM_DEFINED(arg_ptr, arg_sz); /*suppress "uninitialised bytes"*/
	    break;
	}
	case SCTP_OPT_EVENTS:
	{
	    CHKLEN(curr, 9);
	    /* We do not support "sctp_authentication_event" -- it is not
	       implemented in Linux Kernel SCTP anyway.   Just in case if
	       the above structure has more fields than we support,  zero
	       it out -- the extraneous events will NOT be used:
	    */
	    memset (&arg.es, 0, sizeof(arg.es));

	    /* The input "buf" must contain the full definition of all the
	       supported event fields, 1 byte per each,   as each event is
	       either explicitly subscribed or cleared:
	    */
	    arg.es.sctp_data_io_event          = get_int8(curr);   curr++;
	    arg.es.sctp_association_event      = get_int8(curr);   curr++;
	    arg.es.sctp_address_event	       = get_int8(curr);   curr++;
	    arg.es.sctp_send_failure_event     = get_int8(curr);   curr++;
	    arg.es.sctp_peer_error_event       = get_int8(curr);   curr++;
	    arg.es.sctp_shutdown_event	       = get_int8(curr);   curr++;
	    arg.es.sctp_partial_delivery_event = get_int8(curr);   curr++;
	    arg.es.sctp_adaptation_layer_event = get_int8(curr);   curr++;
	    /* sctp_authentication_event not implemented */ curr++;

	    proto   = IPPROTO_SCTP;
	    type    = SCTP_EVENTS;
	    arg_ptr = (char*) (&arg.es);
	    arg_sz  = sizeof  ( arg.es);
	    break;
	}
	/* The following is not available on Solaris 10: */
#	ifdef SCTP_DELAYED_ACK_TIME
	case SCTP_OPT_DELAYED_ACK_TIME:
	{
	    CHKLEN(curr, ASSOC_ID_LEN + 4);
	    arg.av.assoc_id    = GET_ASSOC_ID(curr);	curr += ASSOC_ID_LEN;
	    arg.av.assoc_value = get_int32(curr);	curr += 4;

	    proto   = IPPROTO_SCTP;
	    type    = SCTP_DELAYED_ACK_TIME;
	    arg_ptr = (char*) (&arg.av);
	    arg_sz  = sizeof  ( arg.av);
	    break;
	}
#	endif
	default:
	    /* XXX: No more supported SCTP options. In particular, authentica-
	       tion options (SCTP_AUTH_CHUNK, SCTP_AUTH_KEY, SCTP_PEER_AUTH_
               CHUNKS, SCTP_LOCAL_AUTH_CHUNKS, SCTP_AUTH_SETKEY_ACTIVE)  are
	       not yet implemented in the Linux kernel,  hence not supported
	       here.  Also not supported are SCTP_HMAC_IDENT, as well as any
	       "generic" options except "INET_LOPT_MODE".    Raise an error:
	    */
	    return -1;
	}
#if  defined(IP_TOS) && defined(IPPROTO_IP)             \
    && defined(SO_PRIORITY) && !defined(__WIN32__)
	res = setopt_prio_tos_trick (desc->s, proto, type, arg_ptr, arg_sz, 1);
#else
	res = sock_setopt	    (desc->s, proto, type, arg_ptr, arg_sz);
#endif
	/* The return values of "sock_setopt" can only be 0 or -1: */
	ASSERT(res == 0 || res == -1);
        if (res == 0) desc->recv_cmsgflags = recv_cmsgflags;
	if (res == -1)
	{  /* Got an error, DO NOT continue with other options. However, on
	      Solaris 10, we DO allow SO_SNDBUF and SO_RCVBUF to fail, assu-
	      min that the default kernel versions are good enough:
	   */
#	   ifdef SOLARIS10
	   if (type != SO_SNDBUF && type != SO_RCVBUF)
#	   endif
	   return res;
	}
    }
    /* If we got here, all "sock_setopt"s above were successful:   */
    if (IS_OPEN(desc) && desc->active != old_active) {
	sock_select(desc, (FD_READ|FD_CLOSE), (desc->active > 0));
    }
    return 0;
#   undef CHKLEN
}
#endif /* HAVE_SCTP */

#ifndef __WIN32__
static void put_cmsg_int32(struct cmsghdr *cmsg, char *ptr) {
    union u {
        byte uint8;
        Uint16 uint16;
        Uint32 uint32;
        Uint64 uint64;
    } *p;
    p = (union u*) CMSG_DATA(cmsg);
    switch (LEN_CMSG_DATA(cmsg) * CHAR_BIT) {
    case 8:
        put_int32((Uint32) p->uint8, ptr);
        break;
    case 16:
        put_int32((Uint32) p->uint16, ptr);
        break;
    case 32:
        put_int32(p->uint32, ptr);
        break;
    case 64:
        put_int32((Uint32) p->uint64, ptr);
        break;
    default:
        put_int32(0, ptr);
    }
    return;
}
#endif

/* load all option values into the buf and reply
** return total length of reply filled into ptr
** ptr should point to a buffer with 9*len +1 to be safe!!
*/

static ErlDrvSSizeT inet_fill_opts(inet_descriptor* desc,
				   char* buf, ErlDrvSizeT len,
				   char** dest, ErlDrvSizeT destlen)
{
    int type;
    int proto;
    int opt;
    struct linger li_val;
    int ival;
    char* arg_ptr;
    unsigned int arg_sz;
    char *ptr = NULL;
    ErlDrvSizeT dest_used = 0;
    ErlDrvSizeT dest_allocated = destlen;
    char *orig_dest = *dest;
#ifdef SO_BINDTODEVICE
    char ifname[IFNAMSIZ];
#endif

    /* Ptr is a name parameter */
#define RETURN_ERROR()				\
    do {					\
	if (dest_allocated > destlen) {		\
	    FREE(*dest);			\
	    *dest = orig_dest;			\
	}					\
	return -1;				\
    } while(0)

#define PLACE_FOR(Size,Ptr)						   \
    do {								   \
	ErlDrvSizeT need = dest_used + (Size);					   \
	if (need > INET_MAX_OPT_BUFFER) {				   \
	    RETURN_ERROR();						   \
	}								   \
	if (need > dest_allocated) {					   \
	    char *new_buffer;						   \
	    if (dest_allocated == destlen) {				   \
		new_buffer = ALLOC((dest_allocated = need + 10));	   \
		memcpy(new_buffer,*dest,dest_used);			   \
	    } else {							   \
		new_buffer = REALLOC(*dest, (dest_allocated = need + 10)); \
	    }								   \
	    *dest = new_buffer;						   \
	}								   \
	(Ptr) = (*dest) + dest_used;					   \
	dest_used = need;						   \
    } while (0)

    /* Ptr is a name parameter */
#define TRUNCATE_TO(Size,Ptr)				\
    do {						\
	ErlDrvSizeT new_need = ((Ptr) - (*dest)) + (Size);	\
	if (new_need > dest_used) {			\
	    erts_exit(ERTS_ERROR_EXIT,"Internal error in inet_drv, "	\
		     "miscalculated buffer size");	\
	}						\
	dest_used = new_need;				\
    } while(0)


    PLACE_FOR(1,ptr);
    *ptr = INET_REP_OK;

    while(len--) {
	opt = *buf++;
	proto = SOL_SOCKET;
	ival = 0; /* Windows Vista needs this (only writes part of it) */
	arg_sz = sizeof(ival);
	arg_ptr = (char*) &ival;

	PLACE_FOR(5,ptr);

	switch(opt) {
	case INET_LOPT_BUFFER:
	    *ptr++ = opt;
	    put_int32(desc->bufsz, ptr);
	    continue;
	case INET_LOPT_HEADER:
	    *ptr++ = opt;
	    put_int32(desc->hsz, ptr);
	    continue;
	case INET_LOPT_MODE:
	    *ptr++ = opt;
	    put_int32(desc->mode, ptr);
	    continue;
	case INET_LOPT_DELIVER:
	    *ptr++ = opt;
	    put_int32(desc->deliver, ptr);
	    continue;
	case INET_LOPT_ACTIVE:
	    *ptr++ = opt;
	    put_int32(desc->active, ptr);
            if (desc->active == INET_MULTI) {
                PLACE_FOR(2,ptr);
                put_int16(desc->active_count, ptr);
                ptr += 2;
            }
	    continue;
	case INET_LOPT_PACKET:
	    *ptr++ = opt;
	    put_int32(desc->htype, ptr);
	    continue;
	case INET_LOPT_PACKET_SIZE:
	    *ptr++ = opt;
	    put_int32(desc->psize, ptr);
	    continue;
	case INET_LOPT_EXITONCLOSE:
	    *ptr++ = opt;
	    put_int32(desc->exitf, ptr);
	    continue;

	case INET_LOPT_TCP_HIWTRMRK:
	    if (desc->stype == SOCK_STREAM) {
		*ptr++ = opt;
		ival = ((tcp_descriptor*)desc)->high;
		put_int32(ival, ptr);
	    } else {
		TRUNCATE_TO(0,ptr);
	    }
	    continue;

	case INET_LOPT_TCP_LOWTRMRK:
	    if (desc->stype == SOCK_STREAM) {
		*ptr++ = opt;
		ival = ((tcp_descriptor*)desc)->low;
		put_int32(ival, ptr);
	    } else {
		TRUNCATE_TO(0,ptr);
	    }
	    continue;

	case INET_LOPT_MSGQ_HIWTRMRK: {
	    ErlDrvSizeT high = ERL_DRV_BUSY_MSGQ_READ_ONLY;
	    *ptr++ = opt;
	    erl_drv_busy_msgq_limits(desc->port, NULL, &high);
	    ival = high > INT_MAX ? INT_MAX : (int) high;
	    put_int32(ival, ptr);
	    continue;
	}

	case INET_LOPT_MSGQ_LOWTRMRK: {
	    ErlDrvSizeT low = ERL_DRV_BUSY_MSGQ_READ_ONLY;
	    *ptr++ = opt;
	    erl_drv_busy_msgq_limits(desc->port, &low, NULL);
	    ival = low > INT_MAX ? INT_MAX : (int) low;
	    put_int32(ival, ptr);
	    continue;
	}

	case INET_LOPT_TCP_SEND_TIMEOUT:
	    if (desc->stype == SOCK_STREAM) {
		*ptr++ = opt;
		ival = ((tcp_descriptor*)desc)->send_timeout;
		put_int32(ival, ptr);
	    } else {
		TRUNCATE_TO(0,ptr);
	    }
	    continue;

	case INET_LOPT_TCP_SEND_TIMEOUT_CLOSE:
	    if (desc->stype == SOCK_STREAM) {
		*ptr++ = opt;
		ival = ((tcp_descriptor*)desc)->send_timeout_close;
		put_int32(ival, ptr);
	    } else {
		TRUNCATE_TO(0,ptr);
	    }
	    continue;

	case INET_LOPT_TCP_DELAY_SEND:
	    if (desc->stype == SOCK_STREAM) {
		*ptr++ = opt;
		ival = !!(((tcp_descriptor*)desc)->tcp_add_flags & TCP_ADDF_DELAY_SEND);
		put_int32(ival, ptr);
	    } else {
		TRUNCATE_TO(0,ptr);
	    }
	    continue;

#ifdef HAVE_UDP
	case INET_LOPT_UDP_READ_PACKETS:
	    if (desc->stype == SOCK_DGRAM) {
		*ptr++ = opt;
		ival = ((udp_descriptor*)desc)->read_packets;
		put_int32(ival, ptr);
	    } else {
		TRUNCATE_TO(0,ptr);
	    }
	    continue;
#endif

#ifdef HAVE_SETNS
	case INET_LOPT_NETNS:
	    if (desc->netns != NULL) {
		size_t netns_len;
		netns_len = strlen(desc->netns);
		*ptr++ = opt;
		put_int32(netns_len, ptr);
		PLACE_FOR(netns_len, ptr);
		memcpy(ptr, desc->netns, netns_len);
		ptr += netns_len;
	    } else {
		TRUNCATE_TO(0,ptr);
	    }
	    continue;
#endif

	case INET_LOPT_TCP_SHOW_ECONNRESET:
	    if (desc->sprotocol == IPPROTO_TCP) {
		tcp_descriptor* tdesc = (tcp_descriptor*) desc;
		*ptr++ = opt;
		ival = !!(tdesc->tcp_add_flags & TCP_ADDF_SHOW_ECONNRESET);
		put_int32(ival, ptr);
	    } else {
		TRUNCATE_TO(0,ptr);
	    }
	    continue;

	case INET_OPT_PRIORITY:
#ifdef SO_PRIORITY
	    type = SO_PRIORITY;
	    break;
#else
	    *ptr++ = opt;
	    put_int32(0, ptr);
	    continue;
#endif
	case INET_OPT_TOS:
#if defined(IP_TOS) && defined(IPPROTO_IP)
	    proto = IPPROTO_IP;
	    type = IP_TOS;
	    break;
#else
	    *ptr++ = opt;
	    put_int32(0, ptr);
	    continue;
#endif
	case INET_OPT_TCLASS:
#if defined(IPV6_TCLASS) && defined(IPPROTO_IPV6)
	    proto = IPPROTO_IPV6;
	    type = IPV6_TCLASS;
	    break;
#else
	    TRUNCATE_TO(0,ptr);
	    continue;
#endif
	case INET_OPT_TTL:
#if defined(IP_TTL) && defined(IPPROTO_IP)
	    proto = IPPROTO_IP;
	    type = IP_TTL;
	    break;
#else
	    TRUNCATE_TO(0,ptr);
	    continue;
#endif
	case INET_OPT_RECVTOS:
#if defined(IP_RECVTOS) && defined(IPPROTO_IP)
	    proto = IPPROTO_IP;
	    type = IP_RECVTOS;
	    break;
#else
	    TRUNCATE_TO(0,ptr);
	    continue;
#endif
	case INET_OPT_RECVTCLASS:
#if defined(IPV6_RECVTCLASS) && defined(IPPROTO_IPV6)
	    proto = IPPROTO_IPV6;
	    type = IPV6_RECVTCLASS;
	    break;
#else
	    TRUNCATE_TO(0,ptr);
	    continue;
#endif
	case INET_OPT_RECVTTL:
#if defined(IP_RECVTTL) && defined(IPPROTO_IP)
	    proto = IPPROTO_IP;
	    type = IP_RECVTTL;
	    break;
#else
	    TRUNCATE_TO(0,ptr);
	    continue;
#endif
	case INET_OPT_REUSEADDR:
	    type = SO_REUSEADDR;
	    break;
	case INET_OPT_KEEPALIVE:
	    type = SO_KEEPALIVE;
	    break;
	case INET_OPT_DONTROUTE:
	    type = SO_DONTROUTE;
	    break;
	case INET_OPT_BROADCAST:
	    type = SO_BROADCAST;
	    break;
	case INET_OPT_OOBINLINE:
	    type = SO_OOBINLINE;
	    break;
	case INET_OPT_SNDBUF:
	    type = SO_SNDBUF;
	    break;
	case INET_OPT_RCVBUF:
	    type = SO_RCVBUF;
	    break;
	case TCP_OPT_NODELAY:
	    proto = IPPROTO_TCP;
	    type = TCP_NODELAY;
	    break;
	case TCP_OPT_NOPUSH:
#if defined(INET_TCP_NOPUSH)
	    proto = IPPROTO_TCP;
	    type = INET_TCP_NOPUSH;
	    break;
#else
	    *ptr++ = opt;
	    put_int32(0, ptr);
	    continue;
#endif

#if defined(HAVE_MULTICAST_SUPPORT) && defined(IPPROTO_IP)
	case UDP_OPT_MULTICAST_TTL:
	    proto = IPPROTO_IP;
	    type = IP_MULTICAST_TTL;
	    break;
	case UDP_OPT_MULTICAST_LOOP:
	    proto = IPPROTO_IP;
	    type = IP_MULTICAST_LOOP;
	    break;
	case UDP_OPT_MULTICAST_IF:
	    proto = IPPROTO_IP;
	    type = IP_MULTICAST_IF;
	    break;
	case INET_OPT_LINGER:
	    arg_sz = sizeof(li_val);
	    sys_memzero((void *) &li_val, sizeof(li_val));
	    arg_ptr = (char*) &li_val;
	    type = SO_LINGER;
	    break;
#endif /* defined(HAVE_MULTICAST_SUPPORT) && defined(IPPROTO_IP) */

	case INET_OPT_IPV6_V6ONLY:
#if HAVE_DECL_IPV6_V6ONLY && defined(IPPROTO_IPV6)
	    proto = IPPROTO_IPV6;
	    type = IPV6_V6ONLY;
	    break;
#elif defined(__WIN32__) && defined(HAVE_IN6) && defined(AF_INET6)
	    /* Fake reading 'true' */
	    *ptr++ = opt;
	    put_int32(1, ptr);
	    ptr += 4;
	    continue;
#else
	    TRUNCATE_TO(0,ptr);
	    continue; /* skip - no result */
#endif

	case INET_OPT_RAW:
	    {
		int data_provided;
		/* Raw options are icky, handle directly... */
		if (len < 13) {
		    RETURN_ERROR();
		}
		len -= 13;
		proto = get_int32(buf);
		buf += 4;
		type = get_int32(buf);
		buf += 4;
		data_provided = (int) *buf++;
		arg_sz = get_int32(buf);
		if (arg_sz > INET_MAX_OPT_BUFFER) {
		    RETURN_ERROR();
		}
		buf += 4;
		TRUNCATE_TO(0,ptr);
		PLACE_FOR(13 + arg_sz,ptr);
		arg_ptr = ptr + 13;
		if (data_provided) {
		    if (len < arg_sz) {
			RETURN_ERROR();
		    }
		    memcpy(arg_ptr,buf,arg_sz);
		    buf += arg_sz;
		    len -= arg_sz;
		}
		if (IS_SOCKET_ERROR(sock_getopt(desc->s,proto,type,
						arg_ptr,&arg_sz))) {
		    TRUNCATE_TO(0,ptr);
		    continue;
		}
		TRUNCATE_TO(arg_sz + 13,ptr);
		*ptr++ = opt;
		put_int32(proto,ptr);
		ptr += 4;
		put_int32(type,ptr);
		ptr += 4;
		put_int32(arg_sz,ptr);
		continue;
	    }

#ifdef SO_BINDTODEVICE
	case INET_OPT_BIND_TO_DEVICE:
	    arg_sz = sizeof(ifname);
	    TRUNCATE_TO(0,ptr);
	    PLACE_FOR(5 + arg_sz,ptr);
	    arg_ptr = ptr + 5;
	    if (IS_SOCKET_ERROR(sock_getopt(desc->s,SOL_SOCKET,SO_BINDTODEVICE,
						arg_ptr,&arg_sz))) {
		    TRUNCATE_TO(0,ptr);
		    continue;
		}
	    arg_sz = my_strnlen(arg_ptr, arg_sz);
	    TRUNCATE_TO(arg_sz + 5,ptr);
	    *ptr++ = opt;
	    put_int32(arg_sz,ptr);
	    ptr += arg_sz;
	    continue;
#endif

#ifndef __WIN32__
            /* Winsock does not have struct cmsghdr */
        case INET_OPT_PKTOPTIONS: {
            struct cmsghdr *cmsg, *cmsg_top;
            SOCKLEN_T cmsg_sz;
            union {
                /* Ensure alignment */
                struct cmsghdr hdr;
                /* Room for (IP_TOS | IPV6_TCLASS) + IP_TTL */
                char buf[2*CMSG_SPACE(sizeof(int))];
            } cmsgbuf;
            /* Select between IPv4 or IPv6 PKTOPTIONS
             * depending on the socket protocol family
             */
            switch (desc->sfamily) {
#if defined(IPPROTO_IP) && defined(IP_PKTOPTIONS)
            case AF_INET: {
                proto = IPPROTO_IP;
                type = IP_PKTOPTIONS;
            }
                break;
#endif
#if defined(IPPROTO_IPV6) && defined(IPV6_PKTOPTIONS) && defined(AF_INET6)
            case AF_INET6: {
                proto = IPPROTO_IPV6;
                type = IPV6_PKTOPTIONS;
            }
                break;
#endif
            default: {
                RETURN_ERROR();
            }
            } /* switch */
            TRUNCATE_TO(0, ptr);
            /* Fetch a cmsg buffer from the socket */
            cmsg_sz = sizeof(cmsgbuf.buf);
            if (IS_SOCKET_ERROR(sock_getopt(desc->s, proto, type,
                                            cmsgbuf.buf, &cmsg_sz))) {
                continue;
            }
            /* Reply with Opt/8, Length/32, [COpt/8, Value/32]*
             * i.e opt, total length and then all returned
             * cmsg options and values
             */
            PLACE_FOR(1+4, ptr);
            *ptr++ = opt;
            arg_ptr = ptr; /* Where to put total length */
            arg_sz = 0; /* Total length */
            for (cmsg_top = (struct cmsghdr*)(cmsgbuf.buf + cmsg_sz),
                     cmsg = (struct cmsghdr*)cmsgbuf.buf;
                 cmsg < cmsg_top;
                 cmsg = NXT_CMSG_HDR(cmsg)) {
#define PUT_CMSG_INT32(CMSG_LEVEL, CMSG_TYPE, OPT)      \
                if ((cmsg->cmsg_level == CMSG_LEVEL) && \
                    (cmsg->cmsg_type == CMSG_TYPE)) {   \
                    PLACE_FOR(1+4, ptr);                \
                    *ptr++ = OPT;                       \
                    put_cmsg_int32(cmsg, ptr);          \
                    arg_sz += 1+4;                      \
                    continue;                           \
                }
#if defined(IPPROTO_IP) && defined(IP_TOS)
                PUT_CMSG_INT32(IPPROTO_IP, IP_TOS, INET_OPT_TOS);
#endif
#if defined(IPPROTO_IPV6) && defined(IPV6_TCLASS)
                PUT_CMSG_INT32(IPPROTO_IPV6, IPV6_TCLASS, INET_OPT_TCLASS);
#endif
#if defined(IPPROTO_IP) && defined(IP_TTL)
                PUT_CMSG_INT32(IPPROTO_IP, IP_TTL, INET_OPT_TTL);
#endif
                /* BSD uses the RECV* names in CMSG fields */
#if defined(IPPROTO_IP) && defined(IP_RECVTOS)
                PUT_CMSG_INT32(IPPROTO_IP, IP_RECVTOS, INET_OPT_TOS);
#endif
#if defined(IPPROTO_IPV6) && defined(IPV6_RECVTCLASS)
                PUT_CMSG_INT32(IPPROTO_IPV6, IPV6_RECVTCLASS, INET_OPT_TCLASS);
#endif
#if defined(IPPROTO_IP) && defined(IP_RECVTTL)
                PUT_CMSG_INT32(IPPROTO_IP, IP_RECVTTL, INET_OPT_TTL);
#endif
#undef PUT_CMSG_INT32
            }
            put_int32(arg_sz, arg_ptr); /* Put total length */
            continue;
        }
#endif /* #ifdef __WIN32__ */

	default:
	    RETURN_ERROR();
	}
	/* We have 5 bytes allocated to ptr */
	if (IS_SOCKET_ERROR(sock_getopt(desc->s,proto,type,arg_ptr,&arg_sz))) {
	    TRUNCATE_TO(0,ptr);
	    continue;
	}
	*ptr++ = opt;
	if (arg_ptr == (char*)&ival) {
	    put_int32(ival, ptr);
	}
	else {
	    put_int32(((Uint32) li_val.l_onoff), ptr);
	    PLACE_FOR(4,ptr);
	    put_int32(((Uint32) li_val.l_linger), ptr);
	}
    }
    return (dest_used);
#undef PLACE_FOR
#undef TRUNCATE_TO
#undef RETURN_ERROR
}

#ifdef HAVE_SCTP
#define LOAD_PADDRINFO_CNT                                            \
        (2*LOAD_ATOM_CNT + LOAD_ASSOC_ID_CNT + LOAD_INET_GET_ADDRESS_CNT + \
	 4*LOAD_INT_CNT + LOAD_TUPLE_CNT)
static int load_paddrinfo (ErlDrvTermData * spec, int i,
			   inet_descriptor* desc, struct sctp_paddrinfo* pai)
{
    i = LOAD_ATOM	(spec, i, am_sctp_paddrinfo);
    i = LOAD_ASSOC_ID	(spec, i, pai->spinfo_assoc_id);
    i = load_inet_get_address(spec, i, desc, &pai->spinfo_address);
    switch(pai->spinfo_state)
    {
    case SCTP_ACTIVE:
	i = LOAD_ATOM	(spec, i, am_active);
	break;
    case SCTP_INACTIVE:
	i = LOAD_ATOM	(spec, i, am_inactive);
	break;
#   if HAVE_DECL_SCTP_UNCONFIRMED
    case SCTP_UNCONFIRMED:
      i = LOAD_ATOM	(spec, i, am_unconfirmed);
      break;
#   endif
    default:
      i = LOAD_ATOM	(spec, i, am_undefined);
    }
    i = LOAD_INT	(spec, i, pai->spinfo_cwnd);
    i = LOAD_INT	(spec, i, pai->spinfo_srtt);
    i = LOAD_INT	(spec, i, pai->spinfo_rto );
    i = LOAD_INT	(spec, i, pai->spinfo_mtu );
    /* Close up the record: */
    i = LOAD_TUPLE	(spec, i, 8);
    return i;
}


/*
**  "sctp_fill_opts":   Returns {ok, Results}, or an error:
*/
static ErlDrvSSizeT sctp_fill_opts(inet_descriptor* desc,
				   char* buf, ErlDrvSizeT buflen,
				   char** dest, ErlDrvSizeT destlen)
{
    /* In contrast to the generic "inet_fill_opts", the output here is
       represented by tuples/records, which are formed in the "spec":
    */
    ErlDrvTermData *spec;
    int i      = 0;
    int length = 0; /* Number of result list entries */

    int spec_allocated = PACKET_ERL_DRV_TERM_DATA_LEN;
    spec = ALLOC(sizeof(* spec) * spec_allocated);

#   define RETURN_ERROR(Spec, Errno) \
    do {                    \
	FREE(Spec);        \
	return (Errno);     \
    } while(0)

    /* Spec is a name parmeter */
#   define PLACE_FOR(Spec, Index, N)                            \
    do {                                                        \
	int need;                                               \
	if ((Index) > spec_allocated) {                         \
	    erts_exit(ERTS_ERROR_EXIT,"Internal error in inet_drv, "           \
		     "miscalculated buffer size");              \
	}                                                       \
	need = (Index) + (N);                                   \
	if (need > INET_MAX_OPT_BUFFER/sizeof(ErlDrvTermData)) {\
	    RETURN_ERROR((Spec), -ENOMEM);                      \
	}                                                       \
	if (need > spec_allocated) {                            \
	    (Spec) = REALLOC((Spec),                            \
			     sizeof(* (Spec))                   \
			     * (spec_allocated = need + 20));   \
	}                                                       \
    } while (0)

    PLACE_FOR(spec, i, 2*LOAD_ATOM_CNT + LOAD_PORT_CNT);
    i = LOAD_ATOM (spec, i, am_inet_reply);
    i = LOAD_PORT (spec, i, desc->dport);
    i = LOAD_ATOM (spec, i, am_ok);

    while (buflen > 0) {
	int eopt = *buf;   /* "eopt" is 1-byte encoded */
	buf ++; buflen --;

	switch(eopt)
	{
	/* Local options allowed for SCTP. For TCP and UDP, the values of
	   these options are returned via "res" using integer encoding,
	   but here, we encode them as proper terms the same way as we do
	   it for all other SCTP options:
	*/
	case INET_LOPT_BUFFER:
	{
	    PLACE_FOR(spec, i, LOAD_ATOM_CNT + LOAD_INT_CNT + LOAD_TUPLE_CNT);
	    i = LOAD_ATOM (spec, i, am_buffer);
	    i = LOAD_INT  (spec, i, desc->bufsz);
	    i = LOAD_TUPLE(spec, i, 2);
	    break;
	}
	case INET_LOPT_MODE:
	{
	    PLACE_FOR(spec, i, 2*LOAD_ATOM_CNT + LOAD_TUPLE_CNT);
	    i = LOAD_ATOM (spec, i, am_mode);
	    switch (desc->mode)
	    {
	    	case INET_MODE_LIST  :
		{ i = LOAD_ATOM (spec, i, am_list);   break; }

		case INET_MODE_BINARY:
		{ i = LOAD_ATOM (spec, i, am_binary); break; }

		default: ASSERT (0);
	    }
	    i = LOAD_TUPLE (spec, i, 2);
	    break;
	}
	case INET_LOPT_ACTIVE:
	{
            if (desc->active == INET_MULTI)
                PLACE_FOR(spec, i, LOAD_ATOM_CNT + LOAD_INT_CNT + LOAD_TUPLE_CNT);
            else
                PLACE_FOR(spec, i, 2*LOAD_ATOM_CNT + LOAD_TUPLE_CNT);
	    i = LOAD_ATOM (spec, i, am_active);
	    switch (desc->active)
	    {
		case INET_ACTIVE :
		{ i = LOAD_ATOM (spec, i, am_true);  break; }

		case INET_PASSIVE:
		{ i = LOAD_ATOM (spec, i, am_false); break; }

		case INET_ONCE   :
		{ i = LOAD_ATOM (spec, i, am_once);  break; }

                case INET_MULTI  :
                { i = LOAD_INT(spec, i, desc->active_count); break; }

		default: ASSERT (0);
	    }
	    i = LOAD_TUPLE (spec, i, 2);
	    break;
	}

#ifdef HAVE_SETNS
	case INET_LOPT_NETNS:
	    if (desc->netns != NULL) {
		PLACE_FOR
		    (spec, i,
		     LOAD_ATOM_CNT + LOAD_BUF2BINARY_CNT + LOAD_TUPLE_CNT);
		i = LOAD_ATOM (spec, i, am_netns);
		i = LOAD_BUF2BINARY
		    (spec, i, desc->netns, strlen(desc->netns));
		i = LOAD_TUPLE (spec, i, 2);
		break;
	    }
	    else
		continue; /* Ignore */
#endif

	/* SCTP and generic INET options: */

	case SCTP_OPT_RTOINFO:
	{
	    struct       sctp_rtoinfo rti;
	    unsigned int sz  = sizeof(rti);

	    if (buflen < ASSOC_ID_LEN) RETURN_ERROR(spec, -EINVAL);
	    rti.srto_assoc_id = GET_ASSOC_ID(buf);
	    buf    += ASSOC_ID_LEN;
	    buflen -= ASSOC_ID_LEN;

	    if (sock_getopt(desc->s, IPPROTO_SCTP, SCTP_RTOINFO,
			    &rti, &sz) < 0) continue;
	    /* Fill in the response: */
	    PLACE_FOR(spec, i,
		      2*LOAD_ATOM_CNT + LOAD_ASSOC_ID_CNT +
		      3*LOAD_INT_CNT + 2*LOAD_TUPLE_CNT);
	    i = LOAD_ATOM	(spec, i, am_sctp_rtoinfo);
	    i = LOAD_ATOM	(spec, i, am_sctp_rtoinfo);
	    i = LOAD_ASSOC_ID	(spec, i, rti.srto_assoc_id);
	    i = LOAD_INT	(spec, i, rti.srto_initial);
	    i = LOAD_INT	(spec, i, rti.srto_max);
	    i = LOAD_INT	(spec, i, rti.srto_min);
	    i = LOAD_TUPLE	(spec, i, 5);
	    i = LOAD_TUPLE (spec, i, 2);
	    break;
	}
	case SCTP_OPT_ASSOCINFO:
	{
	    struct       sctp_assocparams ap;
	    unsigned int sz  = sizeof(ap);

	    if (buflen < ASSOC_ID_LEN) RETURN_ERROR(spec, -EINVAL);
	    ap.sasoc_assoc_id = GET_ASSOC_ID(buf);
	    buf    += ASSOC_ID_LEN;
	    buflen -= ASSOC_ID_LEN;

	    if (sock_getopt(desc->s, IPPROTO_SCTP, SCTP_ASSOCINFO,
			    &ap, &sz) < 0) continue;
	    /* Fill in the response: */
	    PLACE_FOR(spec, i,
		      2*LOAD_ATOM_CNT + LOAD_ASSOC_ID_CNT +
		      5*LOAD_INT_CNT + 2*LOAD_TUPLE_CNT);
	    i = LOAD_ATOM	(spec, i, am_sctp_associnfo);
	    i = LOAD_ATOM	(spec, i, am_sctp_assocparams);
	    i = LOAD_ASSOC_ID	(spec, i, ap.sasoc_assoc_id);
	    i = LOAD_INT	(spec, i, ap.sasoc_asocmaxrxt);
	    i = LOAD_INT	(spec, i, ap.sasoc_number_peer_destinations);
	    i = LOAD_INT	(spec, i, ap.sasoc_peer_rwnd);
	    i = LOAD_INT	(spec, i, ap.sasoc_local_rwnd);
	    i = LOAD_INT	(spec, i, ap.sasoc_cookie_life);
	    i = LOAD_TUPLE	(spec, i, 7);
	    i = LOAD_TUPLE	(spec, i, 2);
	    break;
	}
	case SCTP_OPT_INITMSG:
	{
	    struct       sctp_initmsg im;
	    unsigned int sz = sizeof(im);

	    if (sock_getopt(desc->s, IPPROTO_SCTP, SCTP_INITMSG,
			    &im, &sz) < 0) continue;
	    /* Fill in the response: */
	    PLACE_FOR(spec, i,
		      2*LOAD_ATOM_CNT +
		      4*LOAD_INT_CNT + 2*LOAD_TUPLE_CNT);
	    i = LOAD_ATOM	(spec, i, am_sctp_initmsg);
	    i = LOAD_ATOM	(spec, i, am_sctp_initmsg);
	    i = LOAD_INT	(spec, i, im.sinit_num_ostreams);
	    i = LOAD_INT	(spec, i, im.sinit_max_instreams);
	    i = LOAD_INT	(spec, i, im.sinit_max_attempts);
	    i = LOAD_INT	(spec, i, im.sinit_max_init_timeo);
	    i = LOAD_TUPLE	(spec, i, 5);
	    i = LOAD_TUPLE	(spec, i, 2);
	    break;
	}
	/* The following option returns a tuple {bool, int}:   */
	case INET_OPT_LINGER:
	{
	    struct linger lg;
	    unsigned int  sz = sizeof(lg);

	    if (sock_getopt(desc->s, SOL_SOCKET, SO_LINGER,
			    &lg, &sz) < 0) continue;
	    /* Fill in the response: */
	    PLACE_FOR(spec, i,
		      LOAD_ATOM_CNT + LOAD_BOOL_CNT +
		      LOAD_INT_CNT + 2*LOAD_TUPLE_CNT);
	    i = LOAD_ATOM	(spec, i, am_linger);
	    i = LOAD_BOOL	(spec, i, lg.l_onoff);
	    i = LOAD_INT	(spec, i, lg.l_linger);
	    i = LOAD_TUPLE	(spec, i, 2);
	    i = LOAD_TUPLE	(spec, i, 2);
	    break;
	}

#ifdef SO_BINDTODEVICE
	/* The following option returns a binary:   */
	case INET_OPT_BIND_TO_DEVICE: {
	    char ifname[IFNAMSIZ];
	    unsigned int  sz = sizeof(ifname);

	    if (sock_getopt(desc->s, SOL_SOCKET, SO_BINDTODEVICE,
			    &ifname, &sz) < 0) continue;
	    /* Fill in the response: */
	    PLACE_FOR(spec, i,
		      LOAD_ATOM_CNT + LOAD_BUF2BINARY_CNT + LOAD_TUPLE_CNT);
	    i = LOAD_ATOM (spec, i, am_bind_to_device);
	    i = LOAD_BUF2BINARY(spec, i, ifname, my_strnlen(ifname, sz));
	    i = LOAD_TUPLE (spec, i, 2);
	    break;
	}
#endif

	/* The following options just return an integer value: */
	case INET_OPT_RCVBUF   :
	case INET_OPT_SNDBUF   :
	case INET_OPT_REUSEADDR:
	case INET_OPT_DONTROUTE:
	case INET_OPT_PRIORITY :
	case INET_OPT_TOS      :
	case INET_OPT_TCLASS   :
	case INET_OPT_TTL      :
	case INET_OPT_IPV6_V6ONLY:
	case SCTP_OPT_AUTOCLOSE:
	case SCTP_OPT_MAXSEG   :
	/* The following options return true or false:	       */
	case SCTP_OPT_NODELAY  :
	case SCTP_OPT_DISABLE_FRAGMENTS:
	case SCTP_OPT_I_WANT_MAPPED_V4_ADDR:
	case INET_OPT_RECVTOS  :
	case INET_OPT_RECVTCLASS :
	case INET_OPT_RECVTTL :
	{
	    int res   = 0;
	    unsigned int sz = sizeof(res);
	    int proto = 0, type = 0, is_int = 0;
	    ErlDrvTermData tag = am_sctp_error;

	    switch(eopt)
	    {
	    case INET_OPT_RCVBUF   :
	    {
		proto  = SOL_SOCKET;
		type   = SO_RCVBUF;
		is_int = 1;
		tag    = am_recbuf;
		break;
	    }
	    case INET_OPT_SNDBUF   :
	    {
		proto  = SOL_SOCKET;
		type   = SO_SNDBUF;
		is_int = 1;
		tag    = am_sndbuf;
		break;
	    }
	    case INET_OPT_REUSEADDR:
	    {
		proto  = SOL_SOCKET;
		type   = SO_REUSEADDR;
		is_int = 0;
		tag    = am_reuseaddr;
		break;
	    }
	    case INET_OPT_DONTROUTE:
	    {
		proto  = SOL_SOCKET;
		type   = SO_DONTROUTE;
		is_int = 0;
		tag    = am_dontroute;
		break;
	    }
	    case INET_OPT_PRIORITY:
	    {
#	    if defined(SO_PRIORITY)
		proto  = SOL_SOCKET;
		type   = SO_PRIORITY;
		is_int = 1;
		tag    = am_priority;
		break;
#	    else
		/* Not supported -- ignore */
		continue;
#	    endif
	    }
	    case INET_OPT_TOS:
	    {
#	    if defined(IP_TOS) && defined(IPPROTO_IP)
		proto  = IPPROTO_IP;
		type   = IP_TOS;
		is_int = 1;
		tag    = am_tos;
		break;
#	    else
		/* Not supported -- ignore */
		continue;
#	    endif
	    }
	    case INET_OPT_TCLASS:
	    {
#           if defined(IPV6_TCLASS) && defined(IPPROTO_IPV6)
		proto  = IPPROTO_IPV6;
		type   = IPV6_TCLASS;
		is_int = 1;
		tag    = am_tclass;
		break;
#	    else
		/* Not supported -- ignore */
		continue;
#	    endif
	    }
	    case INET_OPT_TTL:
	    {
#           if defined(IP_TTL) && defined(IPPROTO_IP)
		proto  = IPPROTO_IP;
		type   = IP_TTL;
		is_int = 1;
		tag    = am_ttl;
		break;
#	    else
		/* Not supported -- ignore */
		continue;
#	    endif
	    }
	    case INET_OPT_RECVTOS:
	    {
#	    if defined(IP_RECVTOS) && defined(IPPROTO_IP)
		proto  = IPPROTO_IP;
		type   = IP_RECVTOS;
		is_int = 0;
		tag    = am_recvtos;
		break;
#	    else
		/* Not supported -- ignore */
		continue;
#	    endif
	    }
	    case INET_OPT_RECVTCLASS:
	    {
#           if defined(IPV6_RECVTCLASS) && defined(IPPROTO_IPV6)
		proto  = IPPROTO_IPV6;
		type   = IPV6_RECVTCLASS;
		is_int = 0;
		tag    = am_recvtclass;
		break;
#	    else
		/* Not supported -- ignore */
		continue;
#	    endif
	    }
	    case INET_OPT_RECVTTL:
	    {
#	    if defined(IP_RECVTTL) && defined(IPPROTO_IP)
		proto  = IPPROTO_IP;
		type   = IP_RECVTTL;
		is_int = 0;
		tag    = am_recvttl;
		break;
#	    else
		/* Not supported -- ignore */
		continue;
#	    endif
	    }
	    case INET_OPT_IPV6_V6ONLY:
#           if HAVE_DECL_IPV6_V6ONLY && defined(IPPROTO_IPV6)
	    {
		proto  = IPPROTO_IPV6;
		type   = IPV6_V6ONLY;
		tag    = am_ipv6_v6only;
		break;
	    }
#           elif defined(__WIN32__) && defined(HAVE_IN6) && defined(AF_INET6)
#               error Here is a fix for Win IPv6 SCTP needed
#           else
		/* Not supported -- ignore */
		continue;
#           endif
	    case SCTP_OPT_AUTOCLOSE:
	    {
		proto  = IPPROTO_SCTP;
		type   = SCTP_AUTOCLOSE;
		is_int = 1;
		tag    = am_sctp_autoclose;
		break;
	    }
	    case SCTP_OPT_MAXSEG   :
	   {
		proto  = IPPROTO_SCTP;
		type   = SCTP_MAXSEG;
		is_int = 1;
		tag    = am_sctp_maxseg;
		break;
	    }
	    case SCTP_OPT_NODELAY  :
	   {
		proto  = IPPROTO_SCTP;
		type   = SCTP_NODELAY;
		is_int = 0;
		tag    = am_sctp_nodelay;
		break;
	    }
	    case SCTP_OPT_DISABLE_FRAGMENTS:
	    {
		proto  = IPPROTO_SCTP;
		type   = SCTP_DISABLE_FRAGMENTS;
		is_int = 0;
		tag    = am_sctp_disable_fragments;
		break;
	    }
	    case SCTP_OPT_I_WANT_MAPPED_V4_ADDR:
	    {
		proto  = IPPROTO_SCTP;
		type   = SCTP_I_WANT_MAPPED_V4_ADDR;
		is_int = 0;
		tag    = am_sctp_i_want_mapped_v4_addr;
		break;
	    }
	    default:	 ASSERT(0);
	    }
	    if (sock_getopt (desc->s, proto, type, &res, &sz) < 0) continue;
	    /* Form the result: */
	    PLACE_FOR(spec, i, LOAD_ATOM_CNT +
		      (is_int ? LOAD_INT_CNT : LOAD_BOOL_CNT) +
		      LOAD_TUPLE_CNT);
	    i = LOAD_ATOM	(spec, i, tag);
	    if (is_int)
	    	i = LOAD_INT	(spec, i, res);
	    else
	    	i = LOAD_BOOL	(spec, i, res);
	    i = LOAD_TUPLE	(spec, i, 2);
	    break;
	}
	case SCTP_OPT_PRIMARY_ADDR:
	case SCTP_OPT_SET_PEER_PRIMARY_ADDR:
	{
	    /* These 2 options use completely isomorphic data structures: */
	    struct       sctp_setpeerprim sp;
	    unsigned int sz = sizeof(sp);

	    if (buflen < ASSOC_ID_LEN) RETURN_ERROR(spec, -EINVAL);
	    sp.sspp_assoc_id = GET_ASSOC_ID(buf);
	    buf    += ASSOC_ID_LEN;
	    buflen -= ASSOC_ID_LEN;

	    if (sock_getopt(desc->s, IPPROTO_SCTP,
			    (eopt == SCTP_OPT_PRIMARY_ADDR) ?
			    SCTP_PRIMARY_ADDR : SCTP_SET_PEER_PRIMARY_ADDR,
			    &sp, &sz) < 0) continue;
	    /* Fill in the response: */
	    PLACE_FOR(spec, i,
		      2*LOAD_ATOM_CNT + LOAD_ASSOC_ID_CNT +
		      LOAD_INET_GET_ADDRESS_CNT + 2*LOAD_TUPLE_CNT);
	    switch (eopt) {
	    case SCTP_OPT_PRIMARY_ADDR:
		i = LOAD_ATOM(spec, i, am_sctp_primary_addr);
		i = LOAD_ATOM(spec, i, am_sctp_prim);
		break;
	    case SCTP_OPT_SET_PEER_PRIMARY_ADDR:
		i = LOAD_ATOM(spec, i, am_sctp_set_peer_primary_addr);
		i = LOAD_ATOM(spec, i, am_sctp_setpeerprim);
		break;
	    default:
		ASSERT(0);
	    }
	    i = LOAD_ASSOC_ID	(spec, i, sp.sspp_assoc_id);
	    i = load_inet_get_address(spec, i, desc, &sp.sspp_addr);
	    i = LOAD_TUPLE	(spec, i, 3);
	    i = LOAD_TUPLE	(spec, i, 2);
	    break;
	}
	case SCTP_OPT_ADAPTATION_LAYER:
	{
	    struct       sctp_setadaptation ad;
	    unsigned int sz  = sizeof (ad);

	    if (sock_getopt(desc->s, IPPROTO_SCTP, SCTP_ADAPTATION_LAYER,
			    &ad, &sz) < 0) continue;
	    /* Fill in the response: */
	    PLACE_FOR(spec, i,
		      2*LOAD_ATOM_CNT + LOAD_INT_CNT + 2*LOAD_TUPLE_CNT);
	    i = LOAD_ATOM	(spec, i, am_sctp_adaptation_layer);
	    i = LOAD_ATOM	(spec, i, am_sctp_setadaptation);
	    i = LOAD_INT	(spec, i, sock_ntohl(ad.ssb_adaptation_ind));
	    i = LOAD_TUPLE	(spec, i, 2);
	    i = LOAD_TUPLE	(spec, i, 2);
	    break;
	}
	case SCTP_OPT_PEER_ADDR_PARAMS:
	{
	    struct sctp_paddrparams  ap;
	    unsigned int             sz = sizeof(ap);
	    int                      n;
	    char                    *before, *xerror;
	    ErlDrvSizeT              alen;

	    if (buflen < ASSOC_ID_LEN) RETURN_ERROR(spec, -EINVAL);
	    ap.spp_assoc_id = GET_ASSOC_ID(buf);
	    buf += ASSOC_ID_LEN;
	    buflen -= ASSOC_ID_LEN;
	    alen = buflen;
	    before = buf;
	    xerror =
	      inet_set_faddress
	      (desc->sfamily, (inet_address*) (&ap.spp_address),
	       &buf, &alen);
	    if (xerror != NULL) {
#ifdef EAFNOSUPPORT
	        if (xerror == str_eafnosupport) {
		    RETURN_ERROR(spec, -EAFNOSUPPORT);
		}
#else
		RETURN_ERROR(spec, -EINVAL);
#endif
	    }
	    buflen -= buf - before;

	    if (sock_getopt(desc->s, IPPROTO_SCTP, SCTP_PEER_ADDR_PARAMS,
			    &ap, &sz) < 0) continue;
	    /* Fill in the response: */
	    PLACE_FOR(spec, i,
		      2*LOAD_ATOM_CNT + LOAD_ASSOC_ID_CNT +
		      LOAD_INET_GET_ADDRESS_CNT + 4*LOAD_INT_CNT);
	    i = LOAD_ATOM	(spec, i, am_sctp_peer_addr_params);
	    i = LOAD_ATOM	(spec, i, am_sctp_paddrparams);
	    i = LOAD_ASSOC_ID	(spec, i, ap.spp_assoc_id);
	    i = load_inet_get_address(spec, i, desc, &ap.spp_address);
	    i = LOAD_INT	(spec, i, ap.spp_hbinterval);
	    i = LOAD_INT	(spec, i, ap.spp_pathmaxrxt);

	    /* The following fields are not suported in SOLARIS10,
	    ** so put 0s for "spp_pathmtu", "spp_sackdelay",
	    ** and empty list for "spp_flags":
	    */

#	    ifdef HAVE_STRUCT_SCTP_PADDRPARAMS_SPP_PATHMTU
	    i = LOAD_INT	(spec, i, ap.spp_pathmtu);
#           else
	    i = LOAD_INT	(spec, i, 0);
#           endif

#	    ifdef HAVE_STRUCT_SCTP_PADDRPARAMS_SPP_SACKDELAY
	    i = LOAD_INT	(spec, i, ap.spp_sackdelay);
#           else
	    i = LOAD_INT	(spec, i, 0);
#           endif

	    n = 0;
#	    ifdef HAVE_STRUCT_SCTP_PADDRPARAMS_SPP_FLAGS
	    PLACE_FOR(spec, i, 7*LOAD_ATOM_CNT);
	    /* Now Flags, as a list: */
	    if (ap.spp_flags & SPP_HB_ENABLE)
	    	{ i = LOAD_ATOM	(spec, i, am_hb_enable); 	     n++; }

	    if (ap.spp_flags & SPP_HB_DISABLE)
		{ i = LOAD_ATOM (spec, i, am_hb_disable); 	     n++; }

	    if (ap.spp_flags & SPP_HB_DEMAND)
		{ i = LOAD_ATOM (spec, i, am_hb_demand);	     n++; }

	    if (ap.spp_flags & SPP_PMTUD_ENABLE)
		{ i = LOAD_ATOM (spec, i, am_pmtud_enable);          n++; }

	    if (ap.spp_flags & SPP_PMTUD_DISABLE)
		{ i = LOAD_ATOM (spec, i, am_pmtud_disable);         n++; }
#	    ifdef HAVE_STRUCT_SCTP_PADDRPARAMS_SPP_SACKDELAY
	    /* SPP_SACKDELAY_* not in FreeBSD 7.1 */
	    if (ap.spp_flags & SPP_SACKDELAY_ENABLE)
		{ i = LOAD_ATOM (spec, i, am_sackdelay_enable);      n++; }

	    if (ap.spp_flags & SPP_SACKDELAY_DISABLE)
		{ i = LOAD_ATOM (spec, i, am_sackdelay_disable);     n++; }
#	    endif
#	    endif

	    PLACE_FOR(spec, i,
		      LOAD_NIL_CNT + LOAD_LIST_CNT + 2*LOAD_TUPLE_CNT);

	    /* Close up the Flags list: */
	    i = LOAD_NIL	(spec, i);
	    i = LOAD_LIST	(spec, i, n+1);

	    /* Close up the record: */
	    i = LOAD_TUPLE	(spec, i, 8);
	    /* Close up the result tuple: */
	    i = LOAD_TUPLE	(spec, i, 2);
	    break;
	}
	case SCTP_OPT_DEFAULT_SEND_PARAM:
	{
	    struct       sctp_sndrcvinfo sri;
	    unsigned int sz  = sizeof(sri);

	    if (buflen < ASSOC_ID_LEN) RETURN_ERROR(spec, -EINVAL);
	    sri.sinfo_assoc_id = GET_ASSOC_ID(buf);
	    buf += ASSOC_ID_LEN;
	    buflen -= ASSOC_ID_LEN;
	    if (sock_getopt(desc->s, IPPROTO_SCTP, SCTP_DEFAULT_SEND_PARAM,
			    &sri, &sz) < 0) continue;
	    /* Fill in the response: */
	    PLACE_FOR(spec, i, LOAD_ATOM_CNT +
		      SCTP_PARSE_SNDRCVINFO_CNT + LOAD_TUPLE_CNT);
	    i = LOAD_ATOM(spec, i, am_sctp_default_send_param);
	    i = sctp_parse_sndrcvinfo(spec, i, &sri);
	    i = LOAD_TUPLE(spec, i, 2);
	    break;
	}
	case SCTP_OPT_EVENTS:
	{
	    struct       sctp_event_subscribe evs;
	    unsigned int sz  = sizeof(evs);

	    if (sock_getopt(desc->s, IPPROTO_SCTP, SCTP_EVENTS,
			    &evs, &sz) < 0) continue;
	    /* Fill in the response: */
	    PLACE_FOR(spec, i,
		      2*LOAD_ATOM_CNT + 9*LOAD_BOOL_CNT + 2*LOAD_TUPLE_CNT);
	    i = LOAD_ATOM	(spec, i, am_sctp_events);
	    i = LOAD_ATOM	(spec, i, am_sctp_event_subscribe);
	    i = LOAD_BOOL	(spec, i, evs.sctp_data_io_event);
	    i = LOAD_BOOL	(spec, i, evs.sctp_association_event);
	    i = LOAD_BOOL	(spec, i, evs.sctp_address_event);
	    i = LOAD_BOOL	(spec, i, evs.sctp_send_failure_event);
	    i = LOAD_BOOL	(spec, i, evs.sctp_peer_error_event);
	    i = LOAD_BOOL	(spec, i, evs.sctp_shutdown_event);
	    i = LOAD_BOOL	(spec, i, evs.sctp_partial_delivery_event);
	    i = LOAD_BOOL	(spec, i, evs.sctp_adaptation_layer_event);
	    i = LOAD_BOOL	(spec, i, 0);/* NB: sctp_authentication_event
					      * is not yet supported in Linux
					      */
	    i = LOAD_TUPLE	(spec, i, 10);
	    i = LOAD_TUPLE	(spec, i, 2);
	    break;
	}
	/* The following option is not available in Solaris 10: */
#	if HAVE_DECL_SCTP_DELAYED_ACK_TIME
	case SCTP_OPT_DELAYED_ACK_TIME:
	{
	    struct       sctp_assoc_value av;
	    unsigned int sz  = sizeof(av);

	    if (buflen < ASSOC_ID_LEN) RETURN_ERROR(spec, -EINVAL);
	    av.assoc_id = GET_ASSOC_ID(buf);
	    buf    += ASSOC_ID_LEN;
	    buflen -= ASSOC_ID_LEN;

	    if (sock_getopt(desc->s, IPPROTO_SCTP, SCTP_DELAYED_ACK_TIME,
			    &av, &sz) < 0) continue;
	    /* Fill in the response: */
	    PLACE_FOR(spec, i, 2*LOAD_ATOM_CNT + LOAD_ASSOC_ID_CNT +
		      LOAD_INT_CNT + 2*LOAD_TUPLE_CNT);
	    i = LOAD_ATOM	(spec, i, am_sctp_delayed_ack_time);
	    i = LOAD_ATOM	(spec, i, am_sctp_assoc_value);
	    i = LOAD_ASSOC_ID	(spec, i, av.assoc_id);
	    i = LOAD_INT	(spec, i, av.assoc_value);
	    i = LOAD_TUPLE	(spec, i, 3);
	    i = LOAD_TUPLE	(spec, i, 2);
	    break;
	}
#	endif
	case SCTP_OPT_STATUS:
	{
	    struct       sctp_status  st;
	    unsigned int sz  = sizeof(st);

	    if (buflen < ASSOC_ID_LEN) RETURN_ERROR(spec, -EINVAL);
	    st.sstat_assoc_id = GET_ASSOC_ID(buf);
	    buf    += ASSOC_ID_LEN;
	    buflen -= ASSOC_ID_LEN;

	    if (sock_getopt(desc->s, IPPROTO_SCTP, SCTP_STATUS,
			    &st, &sz) < 0) continue;
	    /* Fill in the response: */
	    PLACE_FOR(spec, i, 3*LOAD_ATOM_CNT + LOAD_ASSOC_ID_CNT +
		      6*LOAD_INT_CNT + LOAD_PADDRINFO_CNT +
		      2*LOAD_TUPLE_CNT);
	    i = LOAD_ATOM	(spec, i, am_sctp_status);
	    i = LOAD_ATOM	(spec, i, am_sctp_status);
	    i = LOAD_ASSOC_ID   (spec, i, st.sstat_assoc_id);
	    switch(st.sstat_state)
	    {
            /*  SCTP_EMPTY is not supported on SOLARIS10: */
#	    if HAVE_DECL_SCTP_EMPTY
	    case SCTP_EMPTY:
		i = LOAD_ATOM	(spec, i, am_empty);
		break;
#	    endif
	    case SCTP_CLOSED:
		i = LOAD_ATOM   (spec, i, am_closed);
		break;
#           if HAVE_DECL_SCTP_BOUND
	    case SCTP_BOUND:
		i = LOAD_ATOM	(spec, i, am_bound);
		break;
#           endif
#           if HAVE_DECL_SCTP_LISTEN
	    case SCTP_LISTEN:
		i = LOAD_ATOM	(spec, i, am_listen);
		break;
#           endif
	    case SCTP_COOKIE_WAIT:
		i = LOAD_ATOM	(spec, i, am_cookie_wait);
		break;
	    case SCTP_COOKIE_ECHOED:
		i = LOAD_ATOM	(spec, i, am_cookie_echoed);
		break;
	    case SCTP_ESTABLISHED:
		i = LOAD_ATOM	(spec, i, am_established);
		break;
	    case SCTP_SHUTDOWN_PENDING:
		i = LOAD_ATOM	(spec, i, am_shutdown_pending);
		break;
	    case SCTP_SHUTDOWN_SENT:
		i = LOAD_ATOM	(spec, i, am_shutdown_sent);
		break;
	    case SCTP_SHUTDOWN_RECEIVED:
		i = LOAD_ATOM	(spec, i, am_shutdown_received);
		break;
	    case SCTP_SHUTDOWN_ACK_SENT:
		i = LOAD_ATOM	(spec, i, am_shutdown_ack_sent);
		break;
	    default:
		i = LOAD_ATOM	(spec, i, am_undefined);
		break;
	    }
	    i = LOAD_INT	(spec, i, st.sstat_rwnd);
	    i = LOAD_INT	(spec, i, st.sstat_unackdata);
	    i = LOAD_INT	(spec, i, st.sstat_penddata);
	    i = LOAD_INT	(spec, i, st.sstat_instrms);
	    i = LOAD_INT	(spec, i, st.sstat_outstrms);
	    i = LOAD_INT	(spec, i, st.sstat_fragmentation_point);
	    i = load_paddrinfo	(spec, i, desc, &st.sstat_primary);
	    /* Close up the record: */
	    i = LOAD_TUPLE	(spec, i, 10);
	    /* Close up the result tuple: */
	    i = LOAD_TUPLE	(spec, i, 2);
	    break;
	}
	case SCTP_OPT_GET_PEER_ADDR_INFO:
	{
	    struct sctp_paddrinfo  pai;
	    unsigned int           sz = sizeof(pai);
	    char                  *before, *xerror;
	    ErlDrvSizeT            alen;

	    if (buflen < ASSOC_ID_LEN) RETURN_ERROR(spec, -EINVAL);
	    pai.spinfo_assoc_id = GET_ASSOC_ID(buf);
	    buf    += ASSOC_ID_LEN;
	    buflen -= ASSOC_ID_LEN;
	    alen = buflen;
	    before = buf;
	    xerror =
	      inet_set_faddress
	      (desc->sfamily, (inet_address*) (&pai.spinfo_address),
	       &buf, &alen);
	    if (xerror != NULL) {
#ifdef EAFNOSUPPORT
	        if (xerror == str_eafnosupport) {
		    RETURN_ERROR(spec, -EAFNOSUPPORT);
		}
#else
		RETURN_ERROR(spec, -EINVAL);
#endif
	    }
	    buflen -= buf - before;

	    if (sock_getopt(desc->s, IPPROTO_SCTP, SCTP_GET_PEER_ADDR_INFO,
			    &pai, &sz) < 0) continue;
	    /* Fill in the response: */
	    PLACE_FOR(spec, i,
		      LOAD_ATOM_CNT + LOAD_PADDRINFO_CNT + LOAD_TUPLE_CNT);
	    i = LOAD_ATOM       (spec, i, am_sctp_get_peer_addr_info);
	    i = load_paddrinfo	(spec, i, desc, &pai);
	    i = LOAD_TUPLE	(spec, i, 2);
	    break;
	}
	default:
	    RETURN_ERROR(spec, -EINVAL); /* No more valid options */
	}
	/* If we get here one result has been successfully loaded */
	length ++;
    }
    if (buflen != 0) RETURN_ERROR(spec, -EINVAL); /* Optparam mismatch */

    PLACE_FOR(spec, i, LOAD_NIL_CNT + LOAD_LIST_CNT + 2*LOAD_TUPLE_CNT);

    /* If we get here, we have "length" options: */
    i = LOAD_NIL  (spec, i);
    i = LOAD_LIST (spec, i, length+1);

    /* Close up the {ok, List} response: */
    i = LOAD_TUPLE(spec, i, 2);
    /* Close up the {inet_reply, S, {ok, List}} response:    */
    i = LOAD_TUPLE(spec, i, 3);

    /* Now, convert "spec" into the returnable term: */
    erl_drv_send_term(desc->dport, driver_caller(desc->port), spec, i);
    FREE(spec);

    (*dest)[0] = INET_REP;
    return 1;   /* Response length */
#   undef PLACE_FOR
#   undef RETURN_ERROR
}
#endif

/* fill statistics reply, op codes from src and result in dest
** dst area must be a least 5*len + 1 bytes
*/
static ErlDrvSSizeT inet_fill_stat(inet_descriptor* desc,
				   char* src, ErlDrvSizeT len, char* dst)
{
    unsigned long val;
    int op;
    char* dst_start = dst;

    *dst++ = INET_REP_OK;     /* put reply code */
    while (len--) {
	op = *src++;
	*dst++ = op;  /* copy op code */
	switch(op) {
	case INET_STAT_RECV_CNT:
	    val = desc->recv_cnt;
	    break;
	case INET_STAT_RECV_MAX:
	    val = (unsigned long) desc->recv_max;
	    break;
	case INET_STAT_RECV_AVG:
	    val = (unsigned long) desc->recv_avg;
	    break;
	case INET_STAT_RECV_DVI:
	    val = (unsigned long) fabs(desc->recv_dvi);
	    break;
	case INET_STAT_SEND_CNT:
	    val = desc->send_cnt;
	    break;
	case INET_STAT_SEND_MAX:
	    val = desc->send_max;
	    break;
	case INET_STAT_SEND_AVG:
	    val = (unsigned long) desc->send_avg;
	    break;
	case INET_STAT_SEND_PND:
	    val = (unsigned long) driver_sizeq(desc->port);
	    break;
	case INET_STAT_RECV_OCT:
#ifdef ARCH_64
	    put_int64(desc->recv_oct, dst); /* write it all */
#else
	    put_int32(desc->recv_oct[1], dst);   /* write high 32bit */
	    put_int32(desc->recv_oct[0], dst+4); /* write low 32bit */
#endif
	    dst += 8;
	    continue;
	case INET_STAT_SEND_OCT:
#ifdef ARCH_64
	    put_int64(desc->send_oct, dst); /* write it all */
#else
	    put_int32(desc->send_oct[1], dst);   /* write high 32bit */
	    put_int32(desc->send_oct[0], dst+4); /* write low 32bit */
#endif
	    dst += 8;
	    continue;
	default: return -1; /* invalid argument */
	}
	put_int32(val, dst);  /* write 32bit value */
	dst += 4;
    }
    return dst - dst_start;  /* actual length */
}

static void
send_empty_out_q_msgs(inet_descriptor* desc)
{
  ErlDrvTermData msg[6];
  int msg_len = 0;

  if(NO_SUBSCRIBERS(&desc->empty_out_q_subs))
    return;

  msg_len = LOAD_ATOM(msg, msg_len, am_empty_out_q);
  msg_len = LOAD_PORT(msg, msg_len, desc->dport);
  msg_len = LOAD_TUPLE(msg, msg_len, 2);

  ASSERT(msg_len == sizeof(msg)/sizeof(*msg));

  send_to_subscribers(desc->dport,
		      &desc->empty_out_q_subs,
		      1,
		      msg,
		      msg_len);
}

/* subscribe and fill subscription reply, op codes from src and
** result in dest dst area must be a least 5*len + 1 bytes
*/
static ErlDrvSSizeT inet_subscribe(inet_descriptor* desc,
				   char* src, ErlDrvSizeT len, char* dst)
{
    unsigned long val;
    int op;
    char* dst_start = dst;

    *dst++ = INET_REP_OK;     /* put reply code */
    while (len--) {
	op = *src++;
	*dst++ = op;  /* copy op code */
	switch(op) {
	case INET_SUBS_EMPTY_OUT_Q:
	  val = driver_sizeq(desc->port);
	  if(val > 0)
	    if(!save_subscriber(&desc->empty_out_q_subs,
				driver_caller(desc->port)))
	      return 0;
	  break;
	default: return -1; /* invalid argument */
	}
	put_int32(val, dst);  /* write 32bit value */
	dst += 4;
    }
    return dst - dst_start;  /* actual length */
}

/* Terminate socket */
static void inet_stop(inet_descriptor* desc)
{
    erl_inet_close(desc);
#ifdef HAVE_SETNS
    if (desc->netns != NULL)
	FREE(desc->netns);
#endif
    FREE(desc);
}

static void inet_emergency_close(ErlDrvData data)
{
    /* valid for any (UDP, TCP or SCTP) descriptor */
    tcp_descriptor* tcp_desc = (tcp_descriptor*)data;
    inet_descriptor* desc = INETP(tcp_desc);
    DEBUGF(("inet_emergency_close(%ld) {s=%d\r\n",
	    (long)desc->port, desc->s));
    if (desc->s != INVALID_SOCKET) {
	sock_close(desc->s);
    }
}


static void set_default_msgq_limits(ErlDrvPort port)
{
    ErlDrvSizeT q_high = INET_HIGH_MSGQ_WATERMARK;
    ErlDrvSizeT q_low = INET_LOW_MSGQ_WATERMARK;
    if (q_low < ERL_DRV_BUSY_MSGQ_LIM_MIN)
	q_low = ERL_DRV_BUSY_MSGQ_LIM_MIN;
    else if (q_low > ERL_DRV_BUSY_MSGQ_LIM_MAX)
	q_low = ERL_DRV_BUSY_MSGQ_LIM_MAX;
    if (q_high < ERL_DRV_BUSY_MSGQ_LIM_MIN)
	q_high = ERL_DRV_BUSY_MSGQ_LIM_MIN;
    else if (q_high > ERL_DRV_BUSY_MSGQ_LIM_MAX)
	q_high = ERL_DRV_BUSY_MSGQ_LIM_MAX;
    erl_drv_busy_msgq_limits(port, &q_low, &q_high);
}

/* Allocate descriptor */
static ErlDrvData inet_start(ErlDrvPort port, int size, int protocol)
{
    inet_descriptor* desc;

    if ((desc = (inet_descriptor*) ALLOC(size)) == NULL)
	return NULL;

    desc->s = INVALID_SOCKET;
    desc->event = INVALID_EVENT;
    desc->event_mask = 0;
#ifdef __WIN32__
    desc->forced_events = 0;
    desc->send_would_block = 0;
#endif
    desc->port = port;
    desc->dport = driver_mk_port(port);
    desc->state = INET_STATE_CLOSED;
    desc->prebound = 0;
    desc->bufsz = INET_DEF_BUFFER;
    desc->hsz = 0;                     /* list header size */
    desc->htype = TCP_PB_RAW;          /* default packet type */
    desc->psize = 0;                   /* no size check */
    desc->stype = -1;                  /* bad stype */
    desc->sfamily = -1;
    desc->sprotocol = protocol;
    desc->mode    = INET_MODE_LIST;    /* list mode */
    desc->exitf   = 1;                 /* exit port when close on active
					  socket */
    desc->deliver = INET_DELIVER_TERM; /* standard term format */
    desc->active  = INET_PASSIVE;      /* start passive */
    desc->active_count = 0;
    desc->delimiter    = '\n';         /* line delimiting char */
    desc->oph = NULL;
    desc->opt = NULL;
    desc->op_ref = 0;

    desc->peer_ptr = NULL;
    desc->name_ptr = NULL;

#ifdef ARCH_64
    desc->recv_oct  = 0;
#else
    desc->recv_oct[0] = desc->recv_oct[1] = 0;
#endif
    desc->recv_cnt = 0;
    desc->recv_max = 0;
    desc->recv_avg = 0.0;
    desc->recv_dvi = 0.0;
#ifdef ARCH_64
    desc->send_oct = 0;
#else
    desc->send_oct[0] = desc->send_oct[1] = 0;
#endif
    desc->send_cnt = 0;
    desc->send_max = 0;
    desc->send_avg = 0.0;
    desc->empty_out_q_subs.subscriber = NO_PROCESS;
    desc->empty_out_q_subs.next = NULL;

    sys_memzero((char *)&desc->remote,sizeof(desc->remote));

    desc->flags = 0;

#ifdef HAVE_SETNS
    desc->netns = NULL;
#endif

    desc->recv_cmsgflags = 0;

    return (ErlDrvData)desc;
}

/* MAXHOSTNAMELEN could be 64 or 255 depending
on the platform. Instead, use INET_MAXHOSTNAMELEN
which is always 255 across all platforms */
#define INET_MAXHOSTNAMELEN 255

/*
** common TCP/UDP/SCTP control command
*/
static ErlDrvSSizeT inet_ctl(inet_descriptor* desc, int cmd, char* buf,
			     ErlDrvSizeT len, char** rbuf, ErlDrvSizeT rsize)
{
    switch (cmd) {

    case INET_REQ_GETSTAT: {
	  char* dst;
	  ErlDrvSizeT i;
	  int dstlen = 1;  /* Reply code */

	  for (i = 0; i < len; i++) {
	      switch(buf[i]) {
	      case INET_STAT_SEND_OCT: dstlen += 9; break;
	      case INET_STAT_RECV_OCT: dstlen += 9; break;
	      default: dstlen += 5; break;
	      }
	  }
	  DEBUGF(("inet_ctl(%ld): GETSTAT\r\n", (long) desc->port));
	  if (dstlen > INET_MAX_OPT_BUFFER) /* sanity check */
	      return 0;
	  if (dstlen > rsize) {
	      if ((dst = (char*) ALLOC(dstlen)) == NULL)
		  return 0;
	      *rbuf = dst;  /* call will free this buffer */
	  }
	  else
	      dst = *rbuf;  /* ok we fit in buffer given */
	  return inet_fill_stat(desc, buf, len, dst);
      }

    case INET_REQ_SUBSCRIBE: {
	  char* dst;
	  int dstlen = 1 /* Reply code */ + len*5;
	  DEBUGF(("inet_ctl(%ld): INET_REQ_SUBSCRIBE\r\n", (long) desc->port));
	  if (dstlen > INET_MAX_OPT_BUFFER) /* sanity check */
	      return 0;
	  if (dstlen > rsize) {
	      if ((dst = (char*) ALLOC(dstlen)) == NULL)
		  return 0;
	      *rbuf = dst;  /* call will free this buffer */
	  }
	  else
	      dst = *rbuf;  /* ok we fit in buffer given */
	  return inet_subscribe(desc, buf, len, dst);
      }

    case INET_REQ_GETOPTS: {    /* get options */
	ErlDrvSSizeT replen;
	DEBUGF(("inet_ctl(%ld): GETOPTS\r\n", (long)desc->port));
#ifdef HAVE_SCTP
        if (IS_SCTP(desc))
        {
            if ((replen = sctp_fill_opts(desc, buf, len, rbuf, rsize)) < 0)
                return ctl_error(-replen, rbuf, rsize);
        } else
#endif
	if ((replen = inet_fill_opts(desc, buf, len, rbuf, rsize)) < 0) {
	    return ctl_error(EINVAL, rbuf, rsize);
	}
	return replen;
    }

    case INET_REQ_GETIFLIST: {
	DEBUGF(("inet_ctl(%ld): GETIFLIST\r\n", (long)desc->port));
	if (!IS_OPEN(desc))
	    return ctl_xerror(EXBADPORT, rbuf, rsize);
	return inet_ctl_getiflist(desc, rbuf, rsize);
    }

    case INET_REQ_GETIFADDRS: {
	DEBUGF(("inet_ctl(%ld): GETIFADDRS\r\n", (long)desc->port));
	if (!IS_OPEN(desc))
	    return ctl_xerror(EXBADPORT, rbuf, rsize);
	return inet_ctl_getifaddrs(desc, rbuf, rsize);
    }

    case INET_REQ_IFGET: {
	DEBUGF(("inet_ctl(%ld): IFGET\r\n", (long)desc->port));
	if (!IS_OPEN(desc))
	    return ctl_xerror(EXBADPORT, rbuf, rsize);
	return inet_ctl_ifget(desc, buf, len, rbuf, rsize);
    }

    case INET_REQ_IFSET: {
	DEBUGF(("inet_ctl(%ld): IFSET\r\n", (long)desc->port));
	if (!IS_OPEN(desc))
	    return ctl_xerror(EXBADPORT, rbuf, rsize);
	return inet_ctl_ifset(desc, buf, len, rbuf, rsize);
    }

    case INET_REQ_SETOPTS:  {   /* set options */
	DEBUGF(("inet_ctl(%ld): SETOPTS\r\n", (long)desc->port));
	/* XXX fprintf(stderr,"inet_ctl(%ld): SETOPTS (len = %d)\r\n", (long)desc->port,(int) len); */
	switch(inet_set_opts(desc, buf, len)) {
	case -1:
	    return ctl_error(EINVAL, rbuf, rsize);
	case 0:
	    return ctl_reply(INET_REP_OK, NULL, 0, rbuf, rsize);
	case 1:
	    /*
	     * Let's hope that the descriptor really is a tcp_descriptor here.
	     */
	    /* fprintf(stderr,"Triggered tcp_deliver by setopt.\r\n"); */
	    tcp_deliver((tcp_descriptor *) desc, 0);
	    return ctl_reply(INET_REP_OK, NULL, 0, rbuf, rsize);
	default:
	    /* fprintf(stderr,"Triggered tcp_recv by setopt.\r\n"); */
	    /*
	     * Same as above, but active changed to once w/o header type
	     * change, so try a read instead of just deliver.
	     */
	    tcp_recv((tcp_descriptor *) desc, 0);
	    return ctl_reply(INET_REP_OK, NULL, 0, rbuf, rsize);
	}
    }

    case INET_REQ_GETSTATUS: {
	char tbuf[4];

	DEBUGF(("inet_ctl(%ld): GETSTATUS\r\n", (long)desc->port));
	put_int32(desc->state, tbuf);
	return ctl_reply(INET_REP_OK, tbuf, 4, rbuf, rsize);
    }

    case INET_REQ_GETTYPE: {
	char tbuf[8];

	DEBUGF(("inet_ctl(%ld): GETTYPE\r\n", (long)desc->port));
	if (desc->sfamily == AF_INET) {
	    put_int32(INET_AF_INET, &tbuf[0]);
	}
#if defined(HAVE_IN6) && defined(AF_INET6)
        else if (desc->sfamily == AF_INET6) {
	    put_int32(INET_AF_INET6, &tbuf[0]);
	}
#endif
#ifdef HAVE_SYS_UN_H
	else if (desc->sfamily == AF_UNIX) {
	    put_int32(INET_AF_LOCAL, &tbuf[0]);
	}
#endif
	else
	    return ctl_error(EINVAL, rbuf, rsize);

	if (desc->stype == SOCK_STREAM) {
	    put_int32(INET_TYPE_STREAM, &tbuf[4]);
	}
	else if (desc->stype == SOCK_DGRAM) {
	    put_int32(INET_TYPE_DGRAM, &tbuf[4]);
	}
#ifdef HAVE_SCTP
	else if (desc->stype == SOCK_SEQPACKET) {
	    put_int32(INET_TYPE_SEQPACKET, &tbuf[4]);
	}
#endif
	else
	    return ctl_error(EINVAL, rbuf, rsize);
	return ctl_reply(INET_REP_OK, tbuf, 8, rbuf, rsize);
    }


    case INET_REQ_GETFD: {
	char tbuf[4];

	DEBUGF(("inet_ctl(%ld): GETFD\r\n", (long)desc->port));
	if (!IS_OPEN(desc))
	    return ctl_error(EINVAL, rbuf, rsize);
	put_int32((long)desc->s, tbuf);
	return ctl_reply(INET_REP_OK, tbuf, 4, rbuf, rsize);
    }

    case INET_REQ_GETHOSTNAME: { /* get host name */
	char tbuf[INET_MAXHOSTNAMELEN + 1];

	DEBUGF(("inet_ctl(%ld): GETHOSTNAME\r\n", (long)desc->port));
	if (len != 0)
	    return ctl_error(EINVAL, rbuf, rsize);

        /* gethostname requires len to be max(hostname) + 1 */
	if (IS_SOCKET_ERROR(sock_hostname(tbuf, INET_MAXHOSTNAMELEN + 1)))
	    return ctl_error(sock_errno(), rbuf, rsize);
	return ctl_reply(INET_REP_OK, tbuf, strlen(tbuf), rbuf, rsize);
    }

    case INET_REQ_GETPADDRS: {
	DEBUGF(("inet_ctl(%ld): INET_GETPADDRS\r\n", (long)desc->port));

	if (len != 4) return ctl_error(EINVAL, rbuf, rsize);

	if (! IS_OPEN(desc)) return ctl_xerror(EXBADPORT, rbuf, rsize);

#ifdef HAVE_SCTP
	if (IS_SCTP(desc) && p_sctp_getpaddrs) {
	    struct sockaddr *sa;
	    Uint32 assoc_id;
	    int n;
	    ErlDrvSizeT rlen;

	    assoc_id = get_int32(buf);
	    n = p_sctp_getpaddrs(desc->s, assoc_id, &sa);
	    rlen = reply_inet_addrs(n, (inet_address *) sa, rbuf, rsize, 0);
	    if (n > 0) p_sctp_freepaddrs(sa);
	    return rlen;
	}
#endif
	{ /* Fallback to sock_peer */
	    inet_address addr;
	    SOCKLEN_T sz;
	    int i;

	    sz = sizeof(addr);
	    i = sock_peer(desc->s, (struct sockaddr *) &addr, &sz);
	    return reply_inet_addrs(i >= 0 ? 1 : i, &addr, rbuf, rsize, sz);
	}
    }

    case INET_REQ_PEER:  {      /* get peername */
	char tbuf[sizeof(inet_address)];
	inet_address peer;
	inet_address* ptr;
	unsigned int sz;

	DEBUGF(("inet_ctl(%ld): PEER\r\n", (long)desc->port));

	if (!(desc->state & INET_F_ACTIVE))
	    return ctl_error(ENOTCONN, rbuf, rsize);
	if ((ptr = desc->peer_ptr) != NULL) {
	    sz = desc->peer_addr_len;
	}
	else {
	    ptr = &peer;
            sz = sizeof(peer);
            sys_memzero((char *) &peer, sz);
	    if (IS_SOCKET_ERROR
		(sock_peer
		 (desc->s, (struct sockaddr*)ptr, &sz)))
		return ctl_error(sock_errno(), rbuf, rsize);
	}
	if (inet_get_address(tbuf, ptr, &sz) < 0)
	    return ctl_error(EINVAL, rbuf, rsize);
	return ctl_reply(INET_REP_OK, tbuf, sz, rbuf, rsize);
    }

    case INET_REQ_SETPEER: { /* set fake peername Port Address */
        char *xerror;
	if (len == 0) {
	    desc->peer_ptr = NULL;
	    return ctl_reply(INET_REP_OK, NULL, 0, rbuf, rsize);
	}
	else if (len < 2)
	    return ctl_error(EINVAL, rbuf, rsize);
	else if ((xerror = inet_set_faddress
		  (desc->sfamily, &desc->peer_addr, &buf, &len)) != NULL)
	    return ctl_xerror(xerror, rbuf, rsize);
	else {
	    desc->peer_ptr = &desc->peer_addr;
	    desc->peer_addr_len = (SOCKLEN_T) len;
	    return ctl_reply(INET_REP_OK, NULL, 0, rbuf, rsize);
	}
    }

    case INET_REQ_GETLADDRS: {
	DEBUGF(("inet_ctl(%ld): INET_GETLADDRS\r\n", (long)desc->port));

	if (len != 4) return ctl_error(EINVAL, rbuf, rsize);

	if (! IS_OPEN(desc)) return ctl_xerror(EXBADPORT, rbuf, rsize);

#ifdef HAVE_SCTP
	if (IS_SCTP(desc) && p_sctp_getladdrs) {
	    struct sockaddr *sa;
	    Uint32 assoc_id;
	    int n;
	    ErlDrvSizeT rlen;

	    assoc_id = get_int32(buf);
	    n = p_sctp_getladdrs(desc->s, assoc_id, &sa);
	    rlen = reply_inet_addrs(n, (inet_address *) sa, rbuf, rsize, 0);
	    if (n > 0) p_sctp_freeladdrs(sa);
	    return rlen;
	}
#endif
	{ /* Fallback to sock_name */
	    inet_address addr;
	    SOCKLEN_T sz;
	    int i;

	    sz = sizeof(addr);
	    sys_memzero((char *) &addr, sz);
	    i = sock_name(desc->s, (struct sockaddr *) &addr, &sz);
	    return reply_inet_addrs(i >= 0 ? 1 : i, &addr, rbuf, rsize, sz);
	}
    }

    case INET_REQ_NAME:  {      /* get sockname */
	char tbuf[sizeof(inet_address)];
	inet_address name;
	inet_address* ptr;
	unsigned int sz;

	DEBUGF(("inet_ctl(%ld): NAME\r\n", (long)desc->port));

	if ((ptr = desc->name_ptr) != NULL) {
	    sz = desc->name_addr_len;
	}
	else {
	    ptr = &name;
	    sz = sizeof(name);
	    sys_memzero((char *) &name, sz);
	    if (IS_SOCKET_ERROR
		(sock_name(desc->s, (struct sockaddr*)ptr, &sz)))
		return ctl_error(sock_errno(), rbuf, rsize);
	}
	if (inet_get_address(tbuf, ptr, &sz) < 0)
	    return ctl_error(EINVAL, rbuf, rsize);
	return ctl_reply(INET_REP_OK, tbuf, sz, rbuf, rsize);
    }

    case INET_REQ_SETNAME: { /* set fake sockname Port Address */
        char *xerror;
	if (len == 0) {
	    desc->name_ptr = NULL;
	    return ctl_reply(INET_REP_OK, NULL, 0, rbuf, rsize);
	}
	else if (len < 2)
	    return ctl_error(EINVAL, rbuf, rsize);
	else if ((xerror = inet_set_faddress
		  (desc->sfamily, &desc->name_addr, &buf, &len)) != NULL)
	    return ctl_xerror(xerror, rbuf, rsize);
	else {
	    desc->name_ptr = &desc->name_addr;
	    desc->name_addr_len = (SOCKLEN_T) len;
	    return ctl_reply(INET_REP_OK, NULL, 0, rbuf, rsize);
	}
    }

    case INET_REQ_BIND:  {      /* bind socket */
        char tbuf[2], *xerror;
	inet_address local;
	int port;

	DEBUGF(("inet_ctl(%ld): BIND\r\n", (long)desc->port));

	if (len < 2)
	    return ctl_error(EINVAL, rbuf, rsize);
	if (desc->state != INET_STATE_OPEN)
	    return ctl_xerror(EXBADPORT, rbuf, rsize);

	if ((xerror = inet_set_faddress
	     (desc->sfamily, &local, &buf, &len)) != NULL)
	    return ctl_xerror(xerror, rbuf, rsize);

	if (IS_SOCKET_ERROR(sock_bind(desc->s,(struct sockaddr*) &local, len)))
	    return ctl_error(sock_errno(), rbuf, rsize);

	desc->state = INET_STATE_OPEN;

	port = inet_address_port(&local);
	if (port == 0) {
	    SOCKLEN_T adrlen = sizeof(local);
	    sys_memzero((char *) &local, adrlen);
	    sock_name(desc->s, &local.sa, &adrlen);
	    port = inet_address_port(&local);
	}
        else if (port == -1) port = 0;
	put_int16(sock_ntohs((Uint16) port), tbuf);
	return ctl_reply(INET_REP_OK, tbuf, 2, rbuf, rsize);
    }

    case INET_REQ_IGNOREFD: {
      DEBUGF(("inet_ctl(%ld): IGNOREFD, IGNORED = %d\r\n",
	      (long)desc->port,(int)*buf));

      /*
       * FD can only be ignored for connected TCP connections for now,
       * possible to add UDP and SCTP support if needed.
       */
      if (!IS_CONNECTED(desc))
	  return ctl_error(ENOTCONN, rbuf, rsize);

      if (desc->stype != SOCK_STREAM)
	  return ctl_error(EINVAL, rbuf, rsize);

      if (*buf == 1 && !INET_IGNORED(desc)) {
	  sock_select(desc, (FD_READ|FD_WRITE|FD_CLOSE|ERL_DRV_USE_NO_CALLBACK), 0);
	  if (desc->active)
	    desc->flags |= INET_IGNORE_READ;
	  else
	    desc->flags |= INET_IGNORE_PASSIVE;
      } else if (*buf == 0 && INET_IGNORED(desc)) {
	  int flags = FD_CLOSE;
	  if (desc->flags & INET_IGNORE_READ)
	    flags |= FD_READ;
	  if (desc->flags & INET_IGNORE_WRITE)
	    flags |= FD_WRITE;
	  desc->flags = INET_IGNORE_CLEAR(desc);
	  if (flags != FD_CLOSE)
	    sock_select(desc, flags, 1);
      } else
	  return ctl_error(EINVAL, rbuf, rsize);

      return ctl_reply(INET_REP_OK, NULL, 0, rbuf, rsize);
    }

    case INET_REQ_GETSERVBYNAME: { /* L1 Name-String L2 Proto-String */
	char namebuf[256];
	char protobuf[256];
	char tbuf[2];
	struct servent* srv;
	short port;
	int n;

	if (len < 2)
	    return ctl_error(EINVAL, rbuf, rsize);
	n = get_int8(buf); buf++; len--;
	if (n >= len) /* the = sign makes the test inklude next length byte */
	    return ctl_error(EINVAL, rbuf, rsize);
	memcpy(namebuf, buf, n);
	namebuf[n] = '\0';
	len -= n; buf += n;
	n = get_int8(buf); buf++; len--;
	if (n > len)
	    return ctl_error(EINVAL, rbuf, rsize);
	memcpy(protobuf, buf, n);
	protobuf[n] = '\0';
	if ((srv = sock_getservbyname(namebuf, protobuf)) == NULL)
	    return ctl_error(EINVAL, rbuf, rsize);
	port = sock_ntohs(srv->s_port);
	put_int16(port, tbuf);
	return ctl_reply(INET_REP_OK, tbuf, 2, rbuf, rsize);
    }

    case INET_REQ_GETSERVBYPORT: { /* P1 P0 L1 Proto-String */
	char protobuf[256];
	unsigned short port;
	int n;
	struct servent* srv;

	if (len < 3)
	    return ctl_error(EINVAL, rbuf, rsize);
	port = get_int16(buf);
	port = sock_htons(port);
	buf += 2;
	n = get_int8(buf); buf++; len -= 3;
	if (n > len)
	    return ctl_error(EINVAL, rbuf, rsize);
	memcpy(protobuf, buf, n);
	protobuf[n] = '\0';
	if ((srv = sock_getservbyport(port, protobuf)) == NULL)
	    return ctl_error(EINVAL, rbuf, rsize);
	len = strlen(srv->s_name);
	return ctl_reply(INET_REP_OK, srv->s_name, len, rbuf, rsize);
    }

    default:
	return ctl_xerror(EXBADPORT, rbuf, rsize);
    }
}

/* update statistics on output packets */
static void inet_output_count(inet_descriptor* desc, ErlDrvSizeT len)
{
    unsigned long n = desc->send_cnt + 1;
#ifndef ARCH_64
    Uint32 t = desc->send_oct[0] + len;
    int c = (t < desc->send_oct[0]);
#endif
    double avg = desc->send_avg;

#ifdef ARCH_64
    desc->send_oct += len;
#else
    /* 64 bit octet count in 32 bit words */
    desc->send_oct[0] = t;
    desc->send_oct[1] += c;
#endif
    if (n == 0) /* WRAP, use old avg as input to a new sequence */
	n = 1;
    desc->send_avg += (len - avg) / n;
    if (len > desc->send_max)
	desc->send_max = len;
    desc->send_cnt = n;
}

/* update statistics on input packets */
static void inet_input_count(inet_descriptor* desc, ErlDrvSizeT len)
{
    unsigned long n = desc->recv_cnt + 1;
#ifndef ARCH_64
    Uint32 t = (desc->recv_oct[0] + len);
    int c = (t < desc->recv_oct[0]);
#endif
    double avg = desc->recv_avg;
    double dvi;

#ifdef ARCH_64
    desc->recv_oct += len;
#else
    /* 64 bit octet count in 32 bit words */
    desc->recv_oct[0] = t;
    desc->recv_oct[1] += c;
#endif

    if (n == 0) /* WRAP */
	n = 1;

    /* average packet length */
    avg = avg + (len - avg) / n;
    desc->recv_avg = avg;

    if (len > desc->recv_max)
	desc->recv_max = len;

    /* average deviation from average packet length */
    dvi = desc->recv_dvi;
    desc->recv_dvi = dvi + ((len - avg) - dvi) / n;
    desc->recv_cnt = n;
}

/*----------------------------------------------------------------------------

   TCP

-----------------------------------------------------------------------------*/

/*
** Set new size on buffer, used when packet size is determined
** and the buffer is to small.
** buffer must have a size of at least len bytes (counting from ptr_start!)
*/
static int tcp_expand_buffer(tcp_descriptor* desc, int len)
{
    ErlDrvBinary* bin;
    int offs1;
    int offs2;
    int used = desc->i_ptr_start - desc->i_buf->orig_bytes;
    int ulen = used + len;

    if (desc->i_bufsz >= ulen) /* packet will fit */
	return 0;
    else if (desc->i_buf->orig_size >= ulen) { /* buffer is large enough */
	desc->i_bufsz = ulen;  /* set "virtual" size */
	return 0;
    }

    DEBUGF(("tcp_expand_buffer(%ld): s=%d, from %ld to %d\r\n",
	    (long)desc->inet.port, desc->inet.s, desc->i_buf->orig_size, ulen));

    offs1 = desc->i_ptr_start - desc->i_buf->orig_bytes;
    offs2 = desc->i_ptr - desc->i_ptr_start;

    if ((bin = driver_realloc_binary(desc->i_buf, ulen)) == NULL)
	return -1;

    desc->i_buf = bin;
    desc->i_ptr_start = bin->orig_bytes + offs1;
    desc->i_ptr       = desc->i_ptr_start + offs2;
    desc->i_bufsz     = ulen;
    return 0;
}

/* push data into i_buf  */
static int tcp_push_buffer(tcp_descriptor* desc, char* buf, int len)
{
    ErlDrvBinary* bin;

    if (desc->i_buf == NULL) {
	bin = alloc_buffer(len);
	sys_memcpy(bin->orig_bytes, buf, len);
	desc->i_buf = bin;
	desc->i_bufsz = len;
	desc->i_ptr_start = desc->i_buf->orig_bytes;
	desc->i_ptr = desc->i_ptr_start + len;
    }
    else {
	char* start =  desc->i_buf->orig_bytes;
	int sz_before = desc->i_ptr_start - start;
	int sz_filled = desc->i_ptr - desc->i_ptr_start;

	if (len <= sz_before) {
	    sys_memcpy(desc->i_ptr_start - len, buf, len);
	    desc->i_ptr_start -= len;
	}
	else {
	    bin = alloc_buffer(desc->i_bufsz+len);
	    sys_memcpy(bin->orig_bytes, buf, len);
	    sys_memcpy(bin->orig_bytes+len, desc->i_ptr_start, sz_filled);
	    free_buffer(desc->i_buf);
	    desc->i_bufsz += len;
	    desc->i_buf = bin;
	    desc->i_ptr_start = bin->orig_bytes;
	    desc->i_ptr = desc->i_ptr_start + sz_filled + len;
	}
    }
    desc->i_remain = 0;
    return 0;
}

/* clear CURRENT input buffer */
static void tcp_clear_input(tcp_descriptor* desc)
{
    if (desc->i_buf != NULL)
	free_buffer(desc->i_buf);
    desc->i_buf = NULL;
    desc->i_remain    = 0;
    desc->i_ptr       = NULL;
    desc->i_ptr_start = NULL;
    desc->i_bufsz     = 0;
}

/* clear QUEUED output */
static void tcp_clear_output(tcp_descriptor* desc)
{
    ErlDrvPort ix  = desc->inet.port;
    ErlDrvSizeT qsz = driver_sizeq(ix);

    driver_deq(ix, qsz);
    send_empty_out_q_msgs(INETP(desc));
}


/* Move data so that ptr_start point at buf->orig_bytes */
static void tcp_restart_input(tcp_descriptor* desc)
{
    if (desc->i_ptr_start != desc->i_buf->orig_bytes) {
	int n = desc->i_ptr - desc->i_ptr_start;

	DEBUGF(("tcp_restart_input: move %d bytes\r\n", n));
	sys_memmove(desc->i_buf->orig_bytes, desc->i_ptr_start, n);
	desc->i_ptr_start = desc->i_buf->orig_bytes;
	desc->i_ptr = desc->i_ptr_start + n;
    }
}


static int tcp_inet_init(void)
{
    DEBUGF(("tcp_inet_init() {}\r\n"));
    return 0;
}

/* initialize the TCP descriptor */

static ErlDrvData prep_tcp_inet_start(ErlDrvPort port, char* args)
{
    tcp_descriptor* desc;
    DEBUGF(("tcp_inet_start(%ld) {\r\n", (long)port));

    desc = (tcp_descriptor*)
	inet_start(port, sizeof(tcp_descriptor), IPPROTO_TCP);
    if (desc == NULL)
	return ERL_DRV_ERROR_ERRNO;
    desc->high = INET_HIGH_WATERMARK;
    desc->low  = INET_LOW_WATERMARK;
    desc->send_timeout = INET_INFINITY;
    desc->send_timeout_close = 0;
    desc->busy_on_send = 0;
    desc->i_buf = NULL;
    desc->i_ptr = NULL;
    desc->i_ptr_start = NULL;
    desc->i_remain = 0;
    desc->i_bufsz = 0;
    desc->tcp_add_flags = 0;
    desc->http_state = 0;
    desc->mtd = NULL;
    desc->mtd_cache = NULL;
    desc->multi_first = desc->multi_last = NULL;
    DEBUGF(("tcp_inet_start(%ld) }\r\n", (long)port));
    return (ErlDrvData) desc;
}

static ErlDrvData tcp_inet_start(ErlDrvPort port, char* args)
{
    ErlDrvData data = prep_tcp_inet_start(port, args);
    set_default_msgq_limits(port);
    return data;
}
/* Copy a descriptor, by creating a new port with same settings
 * as the descriptor desc.
 * return NULL on error (SYSTEM_LIMIT no ports avail)
 */
static tcp_descriptor* tcp_inet_copy(tcp_descriptor* desc,SOCKET s,
				     ErlDrvTermData owner, int* err)
{
    ErlDrvSizeT q_low, q_high;
    ErlDrvPort port = desc->inet.port;
    tcp_descriptor* copy_desc;

    copy_desc = (tcp_descriptor*) prep_tcp_inet_start(port, NULL);

    /* Setup event if needed */
    if ((copy_desc->inet.s = s) != INVALID_SOCKET) {
	if ((copy_desc->inet.event = sock_create_event(INETP(copy_desc))) ==
	    INVALID_EVENT) {
	    *err = sock_errno();
	    FREE(copy_desc);
	    return NULL;
	}
    }

    /* Some flags must be inherited at this point */
    copy_desc->inet.mode     = desc->inet.mode;
    copy_desc->inet.exitf    = desc->inet.exitf;
    copy_desc->inet.deliver  = desc->inet.deliver;
    copy_desc->inet.htype    = desc->inet.htype;
    copy_desc->inet.psize    = desc->inet.psize;
    copy_desc->inet.stype    = desc->inet.stype;
    copy_desc->inet.sfamily  = desc->inet.sfamily;
    copy_desc->inet.hsz      = desc->inet.hsz;
    copy_desc->inet.bufsz    = desc->inet.bufsz;
    copy_desc->high          = desc->high;
    copy_desc->low           = desc->low;
    copy_desc->send_timeout  = desc->send_timeout;
    copy_desc->send_timeout_close = desc->send_timeout_close;

    copy_desc->tcp_add_flags = desc->tcp_add_flags
        & (TCP_ADDF_SHOW_ECONNRESET | TCP_ADDF_LINGER_ZERO);

    /* The new port will be linked and connected to the original caller */
    port = driver_create_port(port, owner, "tcp_inet", (ErlDrvData) copy_desc);
    if ((long)port == -1) {
	*err = INET_ERRNO_SYSTEM_LIMIT;
	FREE(copy_desc);
	return NULL;
    }

    /* Read busy msgq limits of parent */
    q_low = q_high = ERL_DRV_BUSY_MSGQ_READ_ONLY;
    erl_drv_busy_msgq_limits(desc->inet.port, &q_low, &q_high);
    /* Write same busy msgq limits to child */
    erl_drv_busy_msgq_limits(port, &q_low, &q_high);

    copy_desc->inet.port = port;
    copy_desc->inet.dport = driver_mk_port(port);

    *err = 0;
    return copy_desc;
}

/*
** Check Special cases:
** 1. we are a listener doing nb accept -> report error on accept !
** 2. we are doing accept -> restore listener state
*/
static void tcp_close_check(tcp_descriptor* desc)
{
    /* XXX:PaN - multiple clients to handle! */
    if (desc->inet.state == INET_STATE_ACCEPTING) {
	inet_async_op *this_op = desc->inet.opt;
	sock_select(INETP(desc), FD_ACCEPT, 0);
	desc->inet.state = INET_STATE_LISTENING;
	if (this_op != NULL) {
	    driver_demonitor_process(desc->inet.port, &(this_op->monitor));
	}
	async_error_am(INETP(desc), am_closed);
    }
    else if (desc->inet.state == INET_STATE_MULTI_ACCEPTING) {
	int id,req;
	ErlDrvTermData caller;
	ErlDrvMonitor monitor;

	sock_select(INETP(desc), FD_ACCEPT, 0);
	desc->inet.state = INET_STATE_LISTENING;
	while (deq_multi_op(desc,&id,&req,&caller,NULL,&monitor) == 0) {
	    driver_demonitor_process(desc->inet.port, &monitor);
	    send_async_error(desc->inet.dport, id, caller, am_closed);
	}
    }
    else if (desc->inet.state == INET_STATE_CONNECTING) {
	async_error_am(INETP(desc), am_closed);
    }
    else if (desc->inet.state == INET_STATE_CONNECTED) {
	async_error_am_all(INETP(desc), am_closed);
    }
    clean_multi_timers(desc, desc->inet.port);
}

/*
** Cleanup & Free
*/
static void tcp_inet_stop(ErlDrvData e)
{
    tcp_descriptor* desc = (tcp_descriptor*)e;
    DEBUGF(("tcp_inet_stop(%ld) {s=%d\r\n",
	    (long)desc->inet.port, desc->inet.s));

    tcp_close_check(desc);
    tcp_clear_input(desc);

#ifdef HAVE_SENDFILE
    if(desc->tcp_add_flags & TCP_ADDF_SENDFILE) {
        desc->tcp_add_flags &= ~TCP_ADDF_SENDFILE;
        close(desc->sendfile.dup_file_fd);
        DEBUGF(("tcp_inet_stop(%p): SENDFILE dup closed %d\r\n",
                desc->inet.port, desc->sendfile.dup_file_fd));
    }
#endif

    DEBUGF(("tcp_inet_stop(%ld) }\r\n", (long)desc->inet.port));
    inet_stop(INETP(desc));
}

/* Closes a tcp descriptor without leaving things hanging; the VM keeps trying
 * to flush IO queues as long as it contains anything even after the port has
 * been closed from the erlang side, which is desired behavior (Think escripts
 * writing to files) but pretty hopeless if the underlying fd has been set to
 * INVALID_SOCKET through desc_close.
 *
 * This function should be used in place of desc_close/erl_inet_close in all
 * TCP-related operations. Note that this only closes the desc cleanly; it
 * will be freed through tcp_inet_stop later on. */
static void tcp_desc_close(tcp_descriptor* desc)
{

    tcp_clear_input(desc);
    tcp_clear_output(desc);

    erl_inet_close(INETP(desc));
}

static void tcp_inet_recv_timeout(ErlDrvData e, ErlDrvTermData dummy)
{
    tcp_descriptor* desc = (tcp_descriptor*)e;
    ASSERT(!desc->inet.active);
    sock_select(INETP(desc),(FD_READ|FD_CLOSE),0);
    desc->i_remain = 0;
    async_error_am(INETP(desc), am_timeout);
}

/* TCP requests from Erlang */
static ErlDrvSSizeT tcp_inet_ctl(ErlDrvData e, unsigned int cmd,
				 char* buf, ErlDrvSizeT len,
				 char** rbuf, ErlDrvSizeT rsize)
{
    tcp_descriptor* desc = (tcp_descriptor*)e;

    cmd -= ERTS_INET_DRV_CONTROL_MAGIC_NUMBER;
    switch(cmd) {
    case INET_REQ_OPEN: { /* open socket and return internal index */
	int domain;
	DEBUGF(("tcp_inet_ctl(%ld): OPEN\r\n", (long)desc->inet.port));
	if (len != 2) return ctl_error(EINVAL, rbuf, rsize);
	switch(buf[0]) {
	case INET_AF_INET:
	    domain = AF_INET;
	    break;
#if defined(HAVE_IN6) && defined(AF_INET6)
	case INET_AF_INET6:
	    domain = AF_INET6;
	    break;
#endif
#ifdef HAVE_SYS_UN_H
	case INET_AF_LOCAL:
	    domain = AF_UNIX;
	    break;
#endif
	default:
	    return ctl_xerror(str_eafnosupport, rbuf, rsize);
	}
	if (buf[1] != INET_TYPE_STREAM) return ctl_error(EINVAL, rbuf, rsize);
	return inet_ctl_open(INETP(desc), domain, SOCK_STREAM, rbuf, rsize);
	break;
    }

    case INET_REQ_FDOPEN: {  /* pass in an open (and optionally bound) socket */
	int domain;
        int bound;
	DEBUGF(("tcp_inet_ctl(%ld): FDOPEN\r\n", (long)desc->inet.port));
	if (len != 6 && len != 10) return ctl_error(EINVAL, rbuf, rsize);
	switch(buf[0]) {
	case INET_AF_INET:
	    domain = AF_INET;
	    break;
#if defined(HAVE_IN6) && defined(AF_INET6)
	case INET_AF_INET6:
	    domain = AF_INET6;
	    break;
#endif
#ifdef HAVE_SYS_UN_H
	case INET_AF_LOCAL:
	    domain = AF_UNIX;
	    break;
#endif
	default:
	    return ctl_xerror(str_eafnosupport, rbuf, rsize);
	}
	if (buf[1] != INET_TYPE_STREAM) return ctl_error(EINVAL, rbuf, rsize);

        if (len == 6) bound = 1;
        else bound = get_int32(buf+2+4);

	return inet_ctl_fdopen(INETP(desc), domain, SOCK_STREAM,
                               (SOCKET) get_int32(buf+2),
                               bound, rbuf, rsize);
	break;
    }

    case INET_REQ_LISTEN: { /* argument backlog */

	int backlog;
	DEBUGF(("tcp_inet_ctl(%ld): LISTEN\r\n", (long)desc->inet.port));
	if (desc->inet.state == INET_STATE_CLOSED)
	    return ctl_xerror(EXBADPORT, rbuf, rsize);
	if (!IS_OPEN(INETP(desc)))
	    return ctl_xerror(EXBADPORT, rbuf, rsize);
	if (len != 2)
	    return ctl_error(EINVAL, rbuf, rsize);
	backlog = get_int16(buf);
	if (IS_SOCKET_ERROR(sock_listen(desc->inet.s, backlog)))
	    return ctl_error(sock_errno(), rbuf, rsize);
	desc->inet.state = INET_STATE_LISTENING;
	return ctl_reply(INET_REP_OK, NULL, 0, rbuf, rsize);
    }


    case INET_REQ_CONNECT: {   /* do async connect */
	int code;
	char tbuf[2], *xerror;
	unsigned timeout;

	DEBUGF(("tcp_inet_ctl(%ld): CONNECT\r\n", (long)desc->inet.port));
	/* INPUT: Timeout(4), Port(2), Address(N) */

	if (!IS_OPEN(INETP(desc)))
	    return ctl_xerror(EXBADPORT, rbuf, rsize);
	if (IS_CONNECTED(INETP(desc)))
	    return ctl_error(EISCONN, rbuf, rsize);
	if (IS_CONNECTING(INETP(desc)))
	    return ctl_error(EINVAL, rbuf, rsize);
	if (len < 6)
	    return ctl_error(EINVAL, rbuf, rsize);
	timeout = get_int32(buf);
	buf += 4;
	len -= 4;
	if ((xerror = inet_set_faddress
	     (desc->inet.sfamily, &desc->inet.remote, &buf, &len)) != NULL)
	    return ctl_xerror(xerror, rbuf, rsize);

	code = sock_connect(desc->inet.s,
			    (struct sockaddr*) &desc->inet.remote, len);
	if (IS_SOCKET_ERROR(code) &&
		((sock_errno() == ERRNO_BLOCK) ||  /* Winsock2 */
		 (sock_errno() == EINPROGRESS))) {	/* Unix & OSE!! */
          sock_select(INETP(desc), FD_CONNECT, 1);
	    desc->inet.state = INET_STATE_CONNECTING;
	    if (timeout != INET_INFINITY)
		driver_set_timer(desc->inet.port, timeout);
	    enq_async(INETP(desc), tbuf, INET_REQ_CONNECT);
	}
	else if (code == 0) { /* ok we are connected */
	    desc->inet.state = INET_STATE_CONNECTED;
	    if (desc->inet.active)
		sock_select(INETP(desc), (FD_READ|FD_CLOSE), 1);
	    enq_async(INETP(desc), tbuf, INET_REQ_CONNECT);
	    async_ok(INETP(desc));
	}
	else {
	    return ctl_error(sock_errno(), rbuf, rsize);
	}
	return ctl_reply(INET_REP_OK, tbuf, 2, rbuf, rsize);
    }

    case INET_REQ_ACCEPT: {  /* do async accept */
	char tbuf[2];
	unsigned timeout;
	inet_address remote;
	unsigned int n;
	SOCKET s;

	DEBUGF(("tcp_inet_ctl(%ld): ACCEPT\r\n", (long)desc->inet.port));
	/* INPUT: Timeout(4) */

	if ((desc->inet.state != INET_STATE_LISTENING && desc->inet.state != INET_STATE_ACCEPTING &&
	     desc->inet.state != INET_STATE_MULTI_ACCEPTING) || len != 4) {
	    return ctl_error(EINVAL, rbuf, rsize);
	}

	timeout = get_int32(buf);

	if (desc->inet.state == INET_STATE_ACCEPTING) {
	    unsigned long time_left = 0;
	    int oid = 0;
	    ErlDrvTermData ocaller = ERL_DRV_NIL;
	    int oreq = 0;
	    unsigned otimeout = 0;
	    ErlDrvTermData caller = driver_caller(desc->inet.port);
	    MultiTimerData *mtd = NULL,*omtd = NULL;
	    ErlDrvMonitor monitor, omonitor;


	    if (driver_monitor_process(desc->inet.port, caller ,&monitor) != 0) {
		return ctl_xerror("noproc", rbuf, rsize);
	    }
	    deq_async_w_tmo(INETP(desc),&oid,&ocaller,&oreq,&otimeout,&omonitor);
	    if (otimeout != INET_INFINITY) {
		driver_read_timer(desc->inet.port, &time_left);
		driver_cancel_timer(desc->inet.port);
		if (time_left <= 0) {
		    time_left = 1;
		}
		omtd = add_multi_timer(desc, desc->inet.port, ocaller,
				       time_left, &tcp_inet_multi_timeout);
	    }
	    enq_old_multi_op(desc, oid, oreq, ocaller, omtd, &omonitor);
	    if (timeout != INET_INFINITY) {
		mtd = add_multi_timer(desc, desc->inet.port, caller,
				      timeout, &tcp_inet_multi_timeout);
	    }
	    enq_multi_op(desc, tbuf, INET_REQ_ACCEPT, caller, mtd, &monitor);
	    desc->inet.state = INET_STATE_MULTI_ACCEPTING;
	    return ctl_reply(INET_REP_OK, tbuf, 2, rbuf, rsize);
	} else if (desc->inet.state == INET_STATE_MULTI_ACCEPTING) {
	    ErlDrvTermData caller = driver_caller(desc->inet.port);
	    MultiTimerData *mtd = NULL;
	    ErlDrvMonitor monitor;

	    if (driver_monitor_process(desc->inet.port, caller ,&monitor) != 0) {
		return ctl_xerror("noproc", rbuf, rsize);
	    }
	    if (timeout != INET_INFINITY) {
		mtd = add_multi_timer(desc, desc->inet.port, caller,
				      timeout, &tcp_inet_multi_timeout);
	    }
	    enq_multi_op(desc, tbuf, INET_REQ_ACCEPT, caller, mtd, &monitor);
	    return ctl_reply(INET_REP_OK, tbuf, 2, rbuf, rsize);
 	} else {
	    n = sizeof(desc->inet.remote);
	    sys_memzero((char *) &remote, n);
	    s = sock_accept(desc->inet.s, (struct sockaddr*) &remote, &n);
	    if (s == INVALID_SOCKET) {
		if (sock_errno() == ERRNO_BLOCK) {
		    ErlDrvMonitor monitor;
		    if (driver_monitor_process(desc->inet.port, driver_caller(desc->inet.port),
					       &monitor) != 0) {
			return ctl_xerror("noproc", rbuf, rsize);
		    }
		    enq_async_w_tmo(INETP(desc), tbuf, INET_REQ_ACCEPT, timeout, &monitor);
		    desc->inet.state = INET_STATE_ACCEPTING;
		    sock_select(INETP(desc),FD_ACCEPT,1);
		    if (timeout != INET_INFINITY) {
			driver_set_timer(desc->inet.port, timeout);
		    }
		} else {
		    return ctl_error(sock_errno(), rbuf, rsize);
		}
	    } else {
		ErlDrvTermData caller = driver_caller(desc->inet.port);
		tcp_descriptor* accept_desc;
		int err;

		if ((accept_desc = tcp_inet_copy(desc,s,caller,&err)) == NULL) {
		    sock_close(s);
		    return ctl_error(err, rbuf, rsize);
		}
		/* FIXME: may MUST lock access_port
		 * 1 - Port is accessible via the erlang:ports()
		 * 2 - Port is accessible via callers process_info(links)
		 */
		accept_desc->inet.remote = remote;
		SET_NONBLOCKING(accept_desc->inet.s);
#ifdef __WIN32__
		driver_select(accept_desc->inet.port, accept_desc->inet.event,
			      ERL_DRV_READ, 1);
#endif
		accept_desc->inet.state = INET_STATE_CONNECTED;
		enq_async(INETP(desc), tbuf, INET_REQ_ACCEPT);
		async_ok_port(INETP(desc), accept_desc->inet.dport);
	    }
	    return ctl_reply(INET_REP_OK, tbuf, 2, rbuf, rsize);
	}
    }
    case INET_REQ_CLOSE:
	DEBUGF(("tcp_inet_ctl(%ld): CLOSE\r\n", (long)desc->inet.port));
	tcp_close_check(desc);
	tcp_desc_close(desc);
	return ctl_reply(INET_REP_OK, NULL, 0, rbuf, rsize);


    case TCP_REQ_RECV: {
	unsigned timeout;
	char tbuf[2];
	int n;

	DEBUGF(("tcp_inet_ctl(%ld): RECV (s=%d)\r\n",
		(long)desc->inet.port, desc->inet.s));
	/* INPUT: Timeout(4),  Length(4) */
	if (!IS_CONNECTED(INETP(desc))) {
	    if (desc->tcp_add_flags & TCP_ADDF_DELAYED_CLOSE_RECV) {
		desc->tcp_add_flags &= ~(TCP_ADDF_DELAYED_CLOSE_RECV|
					 TCP_ADDF_DELAYED_CLOSE_SEND);
		if (desc->tcp_add_flags & TCP_ADDF_DELAYED_ECONNRESET) {
		    desc->tcp_add_flags &= ~TCP_ADDF_DELAYED_ECONNRESET;
		    return ctl_reply(INET_REP_ERROR, "econnreset", 10, rbuf, rsize);
		} else
		    return ctl_reply(INET_REP_ERROR, "closed", 6, rbuf, rsize);
	    }
	    return ctl_error(ENOTCONN, rbuf, rsize);
	}
	if (desc->inet.active || (len != 8))
	    return ctl_error(EINVAL, rbuf, rsize);
	timeout = get_int32(buf);
	buf += 4;
	n = get_int32(buf);
	DEBUGF(("tcp_inet_ctl(%ld) timeout = %d, n = %d\r\n",
		(long)desc->inet.port,timeout,n));
	if ((desc->inet.htype != TCP_PB_RAW) && (n != 0))
	    return ctl_error(EINVAL, rbuf, rsize);
	if (n > TCP_MAX_PACKET_SIZE)
	    return ctl_error(ENOMEM, rbuf, rsize);
	if (enq_async(INETP(desc), tbuf, TCP_REQ_RECV) < 0)
	    return ctl_error(EALREADY, rbuf, rsize);

	if (INET_IGNORED(INETP(desc)) || tcp_recv(desc, n) == 0) {
	    if (timeout == 0)
		async_error_am(INETP(desc), am_timeout);
	    else {
		if (timeout != INET_INFINITY)
                    add_multi_timer(desc, INETP(desc)->port, 0,
                                    timeout, &tcp_inet_recv_timeout);
		if (!INET_IGNORED(INETP(desc)))
		    sock_select(INETP(desc),(FD_READ|FD_CLOSE),1);
		else
		  INETP(desc)->flags |= INET_IGNORE_READ;
	    }
	}
	return ctl_reply(INET_REP_OK, tbuf, 2, rbuf, rsize);
    }

    case TCP_REQ_UNRECV: {
	DEBUGF(("tcp_inet_ctl(%ld): UNRECV\r\n", (long)desc->inet.port));
	if (!IS_CONNECTED(INETP(desc)))
   	    return ctl_error(ENOTCONN, rbuf, rsize);
	tcp_push_buffer(desc, buf, len);
	if (desc->inet.active)
	    tcp_deliver(desc, 0);
	return ctl_reply(INET_REP_OK, NULL, 0, rbuf, rsize);
    }
    case TCP_REQ_SHUTDOWN: {
	int how;
	DEBUGF(("tcp_inet_ctl(%ld): FDOPEN\r\n", (long)desc->inet.port));
	if (!IS_CONNECTED(INETP(desc))) {
	    return ctl_error(ENOTCONN, rbuf, rsize);
	}
	if (len != 1) {
	    return ctl_error(EINVAL, rbuf, rsize);
	}
	how = buf[0];
	if (how != TCP_SHUT_RD && driver_sizeq(desc->inet.port) > 0) {
	    if (how == TCP_SHUT_WR) {
		desc->tcp_add_flags |= TCP_ADDF_PENDING_SHUT_WR;
	    } else if (how == TCP_SHUT_RDWR) {
		desc->tcp_add_flags |= TCP_ADDF_PENDING_SHUT_RDWR;
	    }
	    return ctl_reply(INET_REP_OK, NULL, 0, rbuf, rsize);
	}
	if (IS_SOCKET_ERROR(sock_shutdown(INETP(desc)->s, how))) {
	    if (how != TCP_SHUT_RD)
		desc->tcp_add_flags |= TCP_ADDF_SHUTDOWN_WR_DONE;
	    return ctl_error(sock_errno(), rbuf, rsize);
	} else {
	    return ctl_reply(INET_REP_OK, NULL, 0, rbuf, rsize);
	}
    }

    case TCP_REQ_SENDFILE: {
#ifdef HAVE_SENDFILE
        const ErlDrvSizeT required_len =
            sizeof(desc->sendfile.dup_file_fd) +
            sizeof(Uint64) * 2;

        int raw_file_fd;

        DEBUGF(("tcp_inet_ctl(%ld): SENDFILE\r\n", (long)desc->inet.port));

        if (len != required_len) {
            return ctl_error(EINVAL, rbuf, rsize);
        } else if (!IS_CONNECTED(INETP(desc))) {
            return ctl_error(ENOTCONN, rbuf, rsize);
        } else if (desc->tcp_add_flags & TCP_ADDF_SENDFILE) {
            /* This should not happen as prim_inet.erl makes
               sure that only the controlling process can
               use the sendfile operation. But we add this
               check here anyways just in case that prim_inet
               is changed... */
            return ctl_error(EINVAL, rbuf, rsize);
        }

        sys_memcpy(&raw_file_fd, buf, sizeof(raw_file_fd));
        buf += sizeof(raw_file_fd);

        desc->sendfile.dup_file_fd = dup(raw_file_fd);

        DEBUGF(("tcp_inet_ctl(%p): SENDFILE dup %d\r\n",
                desc->inet.port, desc->sendfile.dup_file_fd));

        if(desc->sendfile.dup_file_fd == -1) {
            return ctl_error(errno, rbuf, rsize);
        }

        desc->sendfile.offset = get_int64(buf);
        buf += sizeof(Uint64);

        desc->sendfile.length = get_int64(buf);
        buf += sizeof(Uint64);

        ASSERT(desc->sendfile.offset >= 0);
        ASSERT(desc->sendfile.length >= 0);

        desc->sendfile.ioq_skip = driver_sizeq(desc->inet.port);
        desc->sendfile.bytes_sent = 0;

        desc->inet.caller = driver_caller(desc->inet.port);
        desc->tcp_add_flags |= TCP_ADDF_SENDFILE;

        /* See if we can finish sending without selecting & rescheduling. */
        if (tcp_inet_sendfile(desc) == 0) {
            if(desc->sendfile.length > 0) {
                sock_select(INETP(desc), FD_WRITE, 1);
            }
        }
        return ctl_reply(INET_REP_OK, NULL, 0, rbuf, rsize);
#else
        return ctl_error(ENOTSUP, rbuf, rsize);
#endif
    }

    default:
	DEBUGF(("tcp_inet_ctl(%ld): %u\r\n", (long)desc->inet.port, cmd));
	return inet_ctl(INETP(desc), cmd, buf, len, rbuf, rsize);
    }

}

static void tcp_inet_send_timeout(ErlDrvData e, ErlDrvTermData dummy)
{
    tcp_descriptor* desc = (tcp_descriptor*)e;
    ASSERT(IS_BUSY(INETP(desc)));
    ASSERT(desc->busy_on_send);
    desc->inet.caller = desc->inet.busy_caller;
    desc->inet.state &= ~INET_F_BUSY;
    desc->busy_on_send = 0;
    set_busy_port(desc->inet.port, 0);
    inet_reply_error_am(INETP(desc), am_timeout);
    if (desc->send_timeout_close) {
        tcp_desc_close(desc);
    }
}

/*
** tcp_inet_timeout:
** called when timer expire:
** TCP socket may be:
**
** a)  receiving   -- send timeout
** b)  connecting  -- close socket
** c)  accepting   -- reset listener
**
*/

static void tcp_inet_timeout(ErlDrvData e)
{
    tcp_descriptor* desc = (tcp_descriptor*)e;
    int state = desc->inet.state;

    DEBUGF(("tcp_inet_timeout(%ld) {s=%d\r\n",
	    (long)desc->inet.port, desc->inet.s));
    if ((state & INET_F_MULTI_CLIENT)) { /* Multi-client always means multi-timers */
	fire_multi_timers(desc, desc->inet.port, e);
    } else if ((state & INET_STATE_CONNECTED) == INET_STATE_CONNECTED) {
        fire_multi_timers(desc, desc->inet.port, e);
    }
    else if ((state & INET_STATE_CONNECTING) == INET_STATE_CONNECTING) {
	/* assume connect timeout */
	/* close the socket since it's not usable (see man pages) */
	tcp_desc_close(desc);
	async_error_am(INETP(desc), am_timeout);
    }
    else if ((state & INET_STATE_ACCEPTING) == INET_STATE_ACCEPTING) {
	inet_async_op *this_op = desc->inet.opt;
	/* timer is set on accept */
	sock_select(INETP(desc), FD_ACCEPT, 0);
	if (this_op != NULL) {
	    driver_demonitor_process(desc->inet.port, &(this_op->monitor));
	}
	desc->inet.state = INET_STATE_LISTENING;
	async_error_am(INETP(desc), am_timeout);
    }
    DEBUGF(("tcp_inet_timeout(%ld) }\r\n", (long)desc->inet.port));
}

static void tcp_inet_multi_timeout(ErlDrvData e, ErlDrvTermData caller)
{
    tcp_descriptor* desc = (tcp_descriptor*)e;
    int id,req;
    ErlDrvMonitor monitor;

    if (remove_multi_op(desc, &id, &req, caller, NULL, &monitor) != 0) {
	return;
    }
    driver_demonitor_process(desc->inet.port, &monitor);
    if (desc->multi_first == NULL) {
	sock_select(INETP(desc),FD_ACCEPT,0);
	desc->inet.state = INET_STATE_LISTENING; /* restore state */
    }
    send_async_error(desc->inet.dport, id, caller, am_timeout);
}



/*
** command:
**   output on a socket only !
**   a reply code will be sent to connected (caller later)
**   {inet_reply, S, Status}
** NOTE! normal sockets use the the tcp_inet_commandv
** but distribution still uses the tcp_inet_command!!
*/

static void tcp_inet_command(ErlDrvData e, char *buf, ErlDrvSizeT len)
{
    tcp_descriptor* desc = (tcp_descriptor*)e;
    desc->inet.caller = driver_caller(desc->inet.port);

    DEBUGF(("tcp_inet_command(%ld) {s=%d\r\n",
	    (long)desc->inet.port, desc->inet.s));
    if (!IS_CONNECTED(INETP(desc)))
	inet_reply_error(INETP(desc), ENOTCONN);
    else if (tcp_send(desc, buf, len) == 0)
	inet_reply_ok(INETP(desc));
    DEBUGF(("tcp_inet_command(%ld) }\r\n", (long)desc->inet.port));
}

static void tcp_inet_commandv(ErlDrvData e, ErlIOVec* ev)
{
    tcp_descriptor* desc = (tcp_descriptor*)e;
    desc->inet.caller = driver_caller(desc->inet.port);

    DEBUGF(("tcp_inet_commanv(%ld) {s=%d\r\n",
	    (long)desc->inet.port, desc->inet.s));
    if (!IS_CONNECTED(INETP(desc))) {
	if (desc->tcp_add_flags & TCP_ADDF_DELAYED_CLOSE_SEND) {
	    desc->tcp_add_flags &= ~TCP_ADDF_DELAYED_CLOSE_SEND;
	    if (desc->tcp_add_flags & TCP_ADDF_DELAYED_ECONNRESET) {
		/* Don't clear flag. Leave it enabled for the next receive
		 * operation.
		 */
		inet_reply_error(INETP(desc), ECONNRESET);
	    } else
		inet_reply_error_am(INETP(desc), am_closed);
	}
	else
	    inet_reply_error(INETP(desc), ENOTCONN);
    }
    else if (desc->tcp_add_flags & TCP_ADDF_PENDING_SHUTDOWN)
	tcp_shutdown_error(desc, EPIPE);
    else if (tcp_sendv(desc, ev) == 0)
	inet_reply_ok(INETP(desc));
    DEBUGF(("tcp_inet_commandv(%ld) }\r\n", (long)desc->inet.port));
}

static void tcp_inet_flush(ErlDrvData e)
{
    tcp_descriptor* desc = (tcp_descriptor*)e;
    int discard_output;

    /* Discard send queue to avoid hanging port (OTP-7615) */
    discard_output = !(desc->inet.event_mask & FD_WRITE);

    discard_output |= desc->tcp_add_flags & TCP_ADDF_LINGER_ZERO;

#ifdef HAVE_SENDFILE
    /* The old file driver aborted when it was stopped during sendfile, so
     * we'll clear the flag and discard all output. It is the job of
     * tcp_inet_stop to close the extra sendfile fd. */
    if(desc->tcp_add_flags & TCP_ADDF_SENDFILE) {
        discard_output = 1;
    }
#endif

    if (discard_output) {
        tcp_clear_output(desc);
    }
}

static void tcp_inet_process_exit(ErlDrvData e, ErlDrvMonitor *monitorp)
{
    tcp_descriptor* desc = (tcp_descriptor*)e;
    ErlDrvTermData who = driver_get_monitored_process(desc->inet.port,monitorp);
    int state = desc->inet.state;

    if ((state & INET_STATE_MULTI_ACCEPTING) == INET_STATE_MULTI_ACCEPTING) {
	int id,req;
	MultiTimerData *timeout;
	if (remove_multi_op(desc, &id, &req, who, &timeout, NULL) != 0) {
	    return;
	}
	if (timeout != NULL) {
	    remove_multi_timer(desc, desc->inet.port, timeout);
	}
	if (desc->multi_first == NULL) {
	    sock_select(INETP(desc),FD_ACCEPT,0);
	    desc->inet.state = INET_STATE_LISTENING; /* restore state */
	}
    } else if ((state & INET_STATE_ACCEPTING) == INET_STATE_ACCEPTING) {
	int did,drid;
	ErlDrvTermData dcaller;
	deq_async(INETP(desc), &did, &dcaller, &drid);
	driver_cancel_timer(desc->inet.port);
	sock_select(INETP(desc),FD_ACCEPT,0);
	desc->inet.state = INET_STATE_LISTENING; /* restore state */
    }
}

static void inet_stop_select(ErlDrvEvent event, void* _)
{
#ifdef __WIN32__
    WSACloseEvent((HANDLE)event);
#else
    sock_close((SOCKET)(long)event);
#endif
}

/* The peer socket has closed, cleanup and send event */
static int tcp_recv_closed(tcp_descriptor* desc)
{
#ifdef DEBUG
    long port = (long) desc->inet.port; /* Used after driver_exit() */
#endif
    int blocking_send = 0;
    DEBUGF(("tcp_recv_closed(%ld): s=%d, in %s, line %d\r\n",
	    port, desc->inet.s, __FILE__, __LINE__));
    if (IS_BUSY(INETP(desc))) {
	/* A send is blocked */
	desc->inet.caller = desc->inet.busy_caller;
	tcp_clear_output(desc);
	if (desc->busy_on_send) {
            cancel_multi_timer(desc, INETP(desc)->port, &tcp_inet_send_timeout);
	    desc->busy_on_send = 0;
	    DEBUGF(("tcp_recv_closed(%ld): busy on send\r\n", port));
	}
	desc->inet.state &= ~INET_F_BUSY;
	set_busy_port(desc->inet.port, 0);
	inet_reply_error_am(INETP(desc), am_closed);
	DEBUGF(("tcp_recv_closed(%ld): busy reply 'closed'\r\n", port));
        blocking_send = 1;
    }
#ifdef HAVE_SENDFILE
    if (desc->tcp_add_flags & TCP_ADDF_SENDFILE) {
        tcp_sendfile_aborted(desc, ENOTCONN);
        blocking_send = 1;
    }
#endif
    if (!blocking_send) {
        /* No blocking send op to reply to right now.
         * If next op is a send, make sure it returns {error,closed}
         * rather than {error,enotconn}.
         */
        desc->tcp_add_flags |= TCP_ADDF_DELAYED_CLOSE_SEND;
    }

    if (!desc->inet.active) {
        /* We must cancel any timer here ! */
        clean_multi_timers(desc, INETP(desc)->port);
	/* passive mode do not terminate port ! */
	tcp_clear_input(desc);
	if (desc->inet.exitf) {
	    tcp_desc_close(desc);
	} else {
	    desc_close_read(INETP(desc));
	}
	async_error_am_all(INETP(desc), am_closed);
	/* next time EXBADSEQ will be delivered  */
	DEBUGF(("tcp_recv_closed(%ld): passive reply all 'closed'\r\n", port));
    } else {
	tcp_clear_input(desc);
	tcp_closed_message(desc);
	if (desc->inet.exitf) {
	    driver_exit(desc->inet.port, 0);
	} else {
	    desc_close_read(INETP(desc));
	}
	DEBUGF(("tcp_recv_closed(%ld): active close\r\n", port));
    }
    DEBUGF(("tcp_recv_closed(%ld): done\r\n", port));
    return -1;
}


/* We have a read error determine the action */
static int tcp_recv_error(tcp_descriptor* desc, int err)
{
    if (err != ERRNO_BLOCK) {
	if (IS_BUSY(INETP(desc))) {
	    /* A send is blocked */
	    desc->inet.caller = desc->inet.busy_caller;
	    tcp_clear_output(desc);
	    if (desc->busy_on_send) {
                cancel_multi_timer(desc, INETP(desc)->port, &tcp_inet_send_timeout);
		desc->busy_on_send = 0;
	    }
	    desc->inet.state &= ~INET_F_BUSY;
	    set_busy_port(desc->inet.port, 0);
	    inet_reply_error_am(INETP(desc), am_closed);
	}
#ifdef HAVE_SENDFILE
        if (desc->tcp_add_flags & TCP_ADDF_SENDFILE) {
            tcp_sendfile_aborted(desc, err);
        }
#endif
	if (!desc->inet.active) {
	    /* We must cancel any timer here ! */
            clean_multi_timers(desc, INETP(desc)->port);
	    tcp_clear_input(desc);
	    if (desc->inet.exitf) {
		tcp_desc_close(desc);
	    } else {
		desc_close_read(INETP(desc));
	    }
	    async_error_am_all(INETP(desc), error_atom(err));
	} else {
	    tcp_clear_input(desc);
	    tcp_error_message(desc, err); /* first error */
	    tcp_closed_message(desc);     /* then closed */
	    if (desc->inet.exitf)
		driver_exit(desc->inet.port, err);
	    else
		desc_close_read(INETP(desc));
	}
	return -1;
    }
    return 0;
}



/*
** Calculate number of bytes that remain to read before deliver
** Assume buf, ptr_start, ptr has been setup
**
** return  > 0 if more to read
**         = 0 if holding complete packet
**         < 0 on error
**
** if return value == 0 then *len will hold the length of the first packet
**    return value > 0 then if *len == 0 then value means upperbound
**                             *len > 0  then value means exact
**
*/
static int tcp_remain(tcp_descriptor* desc, int* len)
{
    char* ptr = desc->i_ptr_start;
    int nfill = (desc->i_ptr - desc->i_buf->orig_bytes); /* filled */
    int nsz   = desc->i_bufsz - nfill;                   /* remain */
    int n = desc->i_ptr - ptr;  /* number of bytes read */
    int tlen;

    tlen = packet_get_length(desc->inet.htype, ptr, n,
                             desc->inet.psize, desc->i_bufsz,
                             desc->inet.delimiter, &desc->http_state);

    DEBUGF(("tcp_remain(%ld): s=%d, n=%d, nfill=%d nsz=%d, tlen %d\r\n",
	    (long)desc->inet.port, desc->inet.s, n, nfill, nsz, tlen));

    if (tlen > 0) {
        if (tlen <= n) { /* got a packet */
            *len = tlen;
            DEBUGF((" => nothing remain packet=%d\r\n", tlen));
            return 0;
        }
        else { /* need known more */
            if (tcp_expand_buffer(desc, tlen) < 0)
                return -1;
            *len = tlen - n;
            DEBUGF((" => remain=%d\r\n", *len));
            return *len;
        }
    }
    else if (tlen == 0) { /* need unknown more */
        *len = 0;
        if (nsz == 0) {
            if (nfill == n) {
                if (desc->inet.psize != 0 && desc->inet.psize > nfill) {
                    if (tcp_expand_buffer(desc, desc->inet.psize) < 0)
                        return -1;
                    return desc->inet.psize;
                }
                else
                    goto error;
            }
            DEBUGF((" => restart more=%d\r\n", nfill - n));
            return nfill - n;
        }
        else {
            DEBUGF((" => more=%d \r\n", nsz));
            return nsz;
        }
    }

error:
    DEBUGF((" => packet error\r\n"));
    return -1;
}

/*
** Deliver all packets ready
** if len == 0 then check start with a check for ready packet
*/
static int tcp_deliver(tcp_descriptor* desc, int len)
{
    int count = 0;
    int n;

    /* Poll for ready packet */
    if (len == 0) {
	/* empty buffer or waiting for more input */
	if ((desc->i_buf == NULL) || (desc->i_remain > 0))
	    return 0;
	if ((n = tcp_remain(desc, &len)) != 0) {
	    if (n < 0) /* packet error */
		return n;
	    if (len > 0)  /* more data pending */
		desc->i_remain = len;
	    return 0;
	}
    }

    while (len > 0) {
	int code;

	inet_input_count(INETP(desc), len);

	/* deliver binary? */
	if (len*4 >= desc->i_buf->orig_size*3) { /* >=75% */
	    code = tcp_reply_binary_data(desc, desc->i_buf,
					 (desc->i_ptr_start -
					  desc->i_buf->orig_bytes),
					 len);
	    if (code < 0)
		return code;

	    /* something after? */
	    if (desc->i_ptr_start + len == desc->i_ptr) { /* no */
		tcp_clear_input(desc);
	    }
	    else { /* move trail to beginning of a new buffer */
		ErlDrvBinary* bin = alloc_buffer(desc->i_bufsz);
		char* ptr_end = desc->i_ptr_start + len;
		int sz = desc->i_ptr - ptr_end;

		memcpy(bin->orig_bytes, ptr_end, sz);
		free_buffer(desc->i_buf);
		desc->i_buf = bin;
		desc->i_ptr_start = desc->i_buf->orig_bytes;
		desc->i_ptr = desc->i_ptr_start + sz;
		desc->i_remain = 0;
	    }
	}
	else {
	    code = tcp_reply_data(desc, desc->i_ptr_start, len);
	    /* XXX The buffer gets thrown away on error  (code < 0)    */
	    /* Windows needs workaround for this in tcp_inet_event...  */
	    if (code < 0)
		return code;
	    desc->i_ptr_start += len;
	    if (desc->i_ptr_start == desc->i_ptr)
		tcp_clear_input(desc);
	    else
		desc->i_remain = 0;
	}

	count++;
	len = 0;

	if (!desc->inet.active) {
            cancel_multi_timer(desc, INETP(desc)->port, &tcp_inet_recv_timeout);
	    sock_select(INETP(desc),(FD_READ|FD_CLOSE),0);
	    if (desc->i_buf != NULL)
		tcp_restart_input(desc);
	}
	else if (desc->i_buf != NULL) {
	    if ((n = tcp_remain(desc, &len)) != 0) {
		if (n < 0) /* packet error */
		    return n;
		tcp_restart_input(desc);
		if (len > 0)
		    desc->i_remain = len;
		len = 0;
	    }
	}
    }
    return count;
}


static int tcp_recv(tcp_descriptor* desc, int request_len)
{
    int n;
    int len;
    int nread;

    if (desc->i_buf == NULL) {  /* allocate a read buffer */
	int sz = (request_len > 0) ? request_len : desc->inet.bufsz;

	if ((desc->i_buf = alloc_buffer(sz)) == NULL)
	    return -1;
	/* XXX: changing bufsz during recv SHOULD/MAY? affect
	 * ongoing operation but is not now
	 */
	desc->i_bufsz = sz; /* use i_bufsz not i_buf->orig_size ! */
	desc->i_ptr_start = desc->i_buf->orig_bytes;
	desc->i_ptr = desc->i_ptr_start;
	nread = sz;
	if (request_len > 0)
	    desc->i_remain = request_len;
	else
	    desc->i_remain = 0;
    }
    else if (request_len > 0) { /* we have a data in buffer and a request */
	n = desc->i_ptr - desc->i_ptr_start;
	if (n >= request_len)
	    return tcp_deliver(desc, request_len);
	else if (tcp_expand_buffer(desc, request_len) < 0)
	    return tcp_recv_error(desc, ENOMEM);
	else
	    desc->i_remain = nread = request_len - n;
    }
    else if (desc->i_remain == 0) {  /* poll remain from buffer data */
	if ((nread = tcp_remain(desc, &len)) < 0)
	    return tcp_recv_error(desc, EMSGSIZE);
	else if (nread == 0)
	    return tcp_deliver(desc, len);
	else if (len > 0)
	    desc->i_remain = len;  /* set remain */
    }
    else  /* remain already set use it */
	nread = desc->i_remain;

    DEBUGF(("tcp_recv(%ld): s=%d about to read %d bytes...\r\n",
	    (long)desc->inet.port, desc->inet.s, nread));

    n = sock_recv(desc->inet.s, desc->i_ptr, nread, 0);

    if (IS_SOCKET_ERROR(n)) {
	int err = sock_errno();
	if (err == ECONNRESET) {
	    DEBUGF((" => detected close (connreset)\r\n"));
	    if (desc->tcp_add_flags & TCP_ADDF_SHOW_ECONNRESET)
		return tcp_recv_error(desc, err);
	    else
		return tcp_recv_closed(desc);
	}
	if (err == ERRNO_BLOCK) {
	    DEBUGF((" => would block\r\n"));
	    return 0;
	}
	else {
	    DEBUGF((" => error: %d\r\n", err));
	    return tcp_recv_error(desc, err);
	}
    }
    else if (n == 0) {
	DEBUGF(("  => detected close\r\n"));
	return tcp_recv_closed(desc);
    }

    DEBUGF((" => got %d bytes\r\n", n));
    desc->i_ptr += n;
    if (desc->i_remain > 0) {
	desc->i_remain -= n;
	if (desc->i_remain == 0)
	    return tcp_deliver(desc, desc->i_ptr - desc->i_ptr_start);
    }
    else {
        nread = tcp_remain(desc, &len);
	if (nread < 0)
	    return tcp_recv_error(desc, EMSGSIZE);
	else if (nread == 0)
            return tcp_deliver(desc, len);
	else if (len > 0)
	    desc->i_remain = len;  /* set remain */
    }
    return 0;
}


#ifdef __WIN32__


static int winsock_event_select(inet_descriptor *desc, int flags, int on)
{
    int save_event_mask = desc->event_mask;

    desc->forced_events = 0;
    if (on)
	desc->event_mask |= flags;
    else
	desc->event_mask &= (~flags);
    DEBUGF(("port %d: winsock_event_select: "
	    "flags=%02X, on=%d, event_mask=%02X\n",
	    desc->port, flags, on, desc->event_mask));
    /* The RIGHT WAY (TM) to do this is to make sure:
       A) The cancelling of all network events is done with
          NULL as the event parameter (bug in NT's winsock),
       B) The actual event handle is reset so that it is only
          raised if one of the requested network events is active,
       C) Avoid race conditions by making sure that the event cannot be set
          while we are preparing to set the correct network event mask.
       The simplest way to do it is to turn off all events, reset the
       event handle and then, if event_mask != 0, turn on the appropriate
       events again. */
    if (WSAEventSelect(desc->s, NULL, 0) != 0) {
	DEBUGF(("port %d: winsock_event_select: "
		"WSAEventSelect returned error, code %d.\n",
		sock_errno()));
	desc->event_mask = save_event_mask;
	return -1;
    }
    if (!ResetEvent(desc->event)) {
	DEBUGF(("port %d: winsock_event_select: "
		"ResetEvent returned error, code %d.\n",
		GetLastError()));
	desc->event_mask = 0;
	return -1;
    }
    if (desc->event_mask != 0) {
	if (WSAEventSelect(desc->s,
			     desc->event,
			     desc->event_mask) != 0) {
	    DEBUGF(("port %d: winsock_event_select: "
		    "WSAEventSelect returned error, code %d.\n",
		    sock_errno()));
	    desc->event_mask = 0;
	    return -1;
	}

	/* Now, WSAEventSelect() is trigged only when the queue goes from
	   full to empty or from empty to full; therefore we need an extra test
	   to see whether it is writeable, readable or closed... */
	if ((desc->event_mask & FD_WRITE)) {
	    int do_force = 1;
	    if (desc->send_would_block) {
		TIMEVAL tmo = {0,0};
		FD_SET fds;
		int ret;

		FD_ZERO(&fds);
		FD_SET(desc->s,&fds);
		do_force = (select(desc->s+1,0,&fds,0,&tmo) > 0);
	    }
	    if (do_force) {
		SetEvent(desc->event);
		desc->forced_events |= FD_WRITE;
	    }
	}
	if ((desc->event_mask & (FD_READ|FD_CLOSE))) {
	    int readable = 0;
	    int closed = 0;
	    TIMEVAL tmo = {0,0};
	    FD_SET fds;
	    int ret;
	    unsigned long arg;

	    FD_ZERO(&fds);
	    FD_SET(desc->s,&fds);
	    ret = select(desc->s+1,&fds,0,0,&tmo);
	    if (ret > 0) {
		++readable;
		if (ioctlsocket(desc->s,FIONREAD,&arg) != 0) {
		    ++closed;	/* Which gives a FD_CLOSE event */
		} else {
		    closed = (arg == 0);
		}
	    }
	    if ((desc->event_mask & FD_READ) && readable && !closed) {
		SetEvent(desc->event);
		desc->forced_events |= FD_READ;
	    }
	    if ((desc->event_mask & FD_CLOSE) && closed) {
		SetEvent(desc->event);
		desc->forced_events |= FD_CLOSE;
	    }
	}
    }
    return 0;
}

static void tcp_inet_event(ErlDrvData e, ErlDrvEvent event)
{
    tcp_descriptor* desc = (tcp_descriptor*)e;
    WSANETWORKEVENTS netEv;
    int err;

    DEBUGF(("tcp_inet_event(%ld) {s=%d\r\n",
	    (long)desc->inet.port, desc->inet.s));
    if (WSAEnumNetworkEvents(desc->inet.s, desc->inet.event,
					&netEv) != 0) {
	DEBUGF((" => EnumNetworkEvents = %d\r\n", sock_errno() ));
	goto error;
    }

    DEBUGF((" => event=%02X, mask=%02X\r\n",
	    netEv.lNetworkEvents, desc->inet.event_mask));

    /* Add the forced events. */

    netEv.lNetworkEvents |= desc->inet.forced_events;

    /*
     * Calling WSAEventSelect() with a mask of 0 doesn't always turn off
     * all events.  To avoid acting on events we don't want, we mask
     * the events with mask for the events we really want.
     */

#ifdef DEBUG
    if ((netEv.lNetworkEvents & ~(desc->inet.event_mask)) != 0) {
	DEBUGF(("port %d:  ... unexpected event: %d\r\n",
		desc->inet.port, netEv.lNetworkEvents & ~(desc->inet.event_mask)));
    }
#endif
    netEv.lNetworkEvents &= desc->inet.event_mask;

    if (netEv.lNetworkEvents & FD_READ) {
	if (tcp_inet_input(desc, event) < 0) {
	    goto error;
	}
	if (netEv.lNetworkEvents & FD_CLOSE) {
	    /*
	     * We must loop to read out the remaining packets (if any).
	     */
	    for (;;) {
		DEBUGF(("Retrying read due to closed port\r\n"));
		/* XXX The buffer will be thrown away on error (empty que).
		   Possible SMP FIXME. */
		if (!desc->inet.active && (desc->inet.opt) == NULL) {
		    goto error;
		}
		if (tcp_inet_input(desc, event) < 0) {
		    goto error;
		}
	    }
	}
    }
    if (netEv.lNetworkEvents & FD_WRITE) {
	desc->inet.send_would_block = 0;
	if (tcp_inet_output(desc, event) < 0)
	    goto error;
    }
    if (netEv.lNetworkEvents & FD_CONNECT) {
	if ((err = netEv.iErrorCode[FD_CONNECT_BIT]) != 0) {
	    async_error(INETP(desc), err);
	} else {
	    tcp_inet_output(desc, event);
	}
    } else if (netEv.lNetworkEvents & FD_ACCEPT) {
	if ((err = netEv.iErrorCode[FD_ACCEPT_BIT]) != 0)
	    async_error(INETP(desc), err);
	else
	    tcp_inet_input(desc, event);
    }
    if (netEv.lNetworkEvents & FD_CLOSE) {
	/* error in err = netEv.iErrorCode[FD_CLOSE_BIT] */
	DEBUGF(("Detected close in %s, line %d\r\n", __FILE__, __LINE__));
	if (desc->tcp_add_flags & TCP_ADDF_SHOW_ECONNRESET) {
	    err = netEv.iErrorCode[FD_CLOSE_BIT];
	    if (err == ECONNRESET)
		tcp_recv_error(desc, err);
	    else if (err == ECONNABORTED && IS_CONNECTED(INETP(desc))) {
		/* translate this error to ECONNRESET */
		tcp_recv_error(desc, ECONNRESET);
	    }
	    else
		tcp_recv_closed(desc);
	}
	else
	    tcp_recv_closed(desc);
    }
    DEBUGF(("tcp_inet_event(%ld) }\r\n", (long)desc->inet.port));
    return;

 error:
    DEBUGF(("tcp_inet_event(%ld) error}\r\n", (long)desc->inet.port));
    return;
}

#endif /* __WIN32__ */


/* socket has input:
** 1. INET_STATE_ACCEPTING  => non block accept ?
** 2. INET_STATE_CONNECTED => read input
*/
static int tcp_inet_input(tcp_descriptor* desc, HANDLE event)
{
    int ret = 0;
#ifdef DEBUG
    long port = (long) desc->inet.port;  /* Used after driver_exit() */
#endif
    ASSERT(!INET_IGNORED(INETP(desc)));
    DEBUGF(("tcp_inet_input(%ld) {s=%d\r\n", port, desc->inet.s));
    /* XXX fprintf(stderr,"tcp_inet_input(%ld) {s=%d}\r\n",(long) desc->inet.port, desc->inet.s); */
    if (desc->inet.state == INET_STATE_ACCEPTING) {
	SOCKET s;
	unsigned int len;
	inet_address remote;
	inet_async_op *this_op = desc->inet.opt;

	len = sizeof(desc->inet.remote);
	sys_memzero((char *) &remote, len);
	s = sock_accept(desc->inet.s, (struct sockaddr*) &remote, &len);
	if (s == INVALID_SOCKET && sock_errno() == ERRNO_BLOCK) {
	    /* Just try again, no real error, just a ghost trigger from poll,
	       keep the default return code and everything else as is */
	    goto done;
	}

	sock_select(INETP(desc),FD_ACCEPT,0);
	desc->inet.state = INET_STATE_LISTENING; /* restore state */

	if (this_op != NULL) {
	    driver_demonitor_process(desc->inet.port, &(this_op->monitor));
	}


	driver_cancel_timer(desc->inet.port); /* posssibly cancel a timer */

	if (s == INVALID_SOCKET) {
	    ret = async_error(INETP(desc), sock_errno());
	    goto done;
	}
	else {
	    ErlDrvTermData caller;
	    tcp_descriptor* accept_desc;
	    int err;

	    if (desc->inet.opt == NULL) {
		/* No caller setup */
		sock_close(s);
		ret = async_error(INETP(desc), EINVAL);
		goto done;
	    }
	    caller = desc->inet.opt->caller;
	    if ((accept_desc = tcp_inet_copy(desc,s,caller,&err)) == NULL) {
		sock_close(s);
		ret = async_error(INETP(desc), err);
		goto done;
	    }
	    /* FIXME: may MUST lock port
	     * 1 - Port is accessible via the erlang:ports()
	     * 2 - Port is accessible via callers process_info(links)
	     */
	    accept_desc->inet.remote = remote;
	    SET_NONBLOCKING(accept_desc->inet.s);
#ifdef __WIN32__
	    driver_select(accept_desc->inet.port, accept_desc->inet.event,
			  ERL_DRV_READ, 1);
#endif
	    accept_desc->inet.state = INET_STATE_CONNECTED;
	    ret =  async_ok_port(INETP(desc), accept_desc->inet.dport);
	    goto done;
	}
    } else if (desc->inet.state == INET_STATE_MULTI_ACCEPTING) {
	SOCKET s;
	unsigned int len;
	inet_address remote;
	int id,req;
	ErlDrvTermData caller;
	MultiTimerData *timeout;
	ErlDrvMonitor monitor;
#ifdef HARDDEBUG
	int times = 0;
#endif

	while (desc->inet.state == INET_STATE_MULTI_ACCEPTING) {
	    len = sizeof(desc->inet.remote);
	    sys_memzero((char *) &remote, len);
	    s = sock_accept(desc->inet.s, (struct sockaddr*) &remote, &len);
	    if (s == INVALID_SOCKET && sock_errno() == ERRNO_BLOCK) {
		/* Just try again, no real error, keep the last return code */
		goto done;
	    }
#ifdef HARDDEBUG
	    if (++times > 1) {
		erts_fprintf(stderr,"Accepts in one suite: %d :-)\r\n",times);
	    }
#endif
	    if (deq_multi_op(desc,&id,&req,&caller,&timeout,&monitor) != 0) {
		ret = -1;
		goto done;
	    }

	    if (desc->multi_first == NULL) {
		sock_select(INETP(desc),FD_ACCEPT,0);
		desc->inet.state = INET_STATE_LISTENING; /* restore state */
	    }

	    if (timeout != NULL) {
		remove_multi_timer(desc, desc->inet.port, timeout);
	    }

	    driver_demonitor_process(desc->inet.port, &monitor);


	    if (s == INVALID_SOCKET) { /* Not ERRNO_BLOCK, that's handled right away */
		ret = send_async_error(desc->inet.dport,
				       id, caller, error_atom(sock_errno()));
		goto done;
	    }
	    else {
		tcp_descriptor* accept_desc;
		int err;

		if ((accept_desc = tcp_inet_copy(desc,s,caller,&err)) == NULL) {
		    sock_close(s);
		    ret = send_async_error(desc->inet.dport,
					   id, caller, error_atom(err));
		    goto done;
		}
		accept_desc->inet.remote = remote;
		SET_NONBLOCKING(accept_desc->inet.s);
#ifdef __WIN32__
		driver_select(accept_desc->inet.port, accept_desc->inet.event,
			      ERL_DRV_READ, 1);
#endif
		accept_desc->inet.state = INET_STATE_CONNECTED;
		ret =  send_async_ok_port(desc->inet.dport,
					  id, caller, accept_desc->inet.dport);
	    }
	}
    }
    else if (IS_CONNECTED(INETP(desc))) {
	ret = tcp_recv(desc, 0);
	goto done;
    }
    else {
	/* maybe a close op from connection attempt?? */
	sock_select(INETP(desc),FD_ACCEPT,0);
	DEBUGF(("tcp_inet_input(%ld): s=%d bad state: %04x\r\n",
		port, desc->inet.s, desc->inet.state));
    }
 done:
    DEBUGF(("tcp_inet_input(%ld) }\r\n", port));
    return ret;
}

static int tcp_send_or_shutdown_error(tcp_descriptor* desc, int err)
{
    int show_econnreset = (err == ECONNRESET
			   && desc->tcp_add_flags & TCP_ADDF_SHOW_ECONNRESET);

    /*
     * If the port is busy, we must do some clean-up before proceeding.
     */
    if (IS_BUSY(INETP(desc))) {
	desc->inet.caller = desc->inet.busy_caller;
	if (desc->busy_on_send) {
            cancel_multi_timer(desc, INETP(desc)->port, &tcp_inet_send_timeout);
	    desc->busy_on_send = 0;
	}
	desc->inet.state &= ~INET_F_BUSY;
	set_busy_port(desc->inet.port, 0);
    }

    /*
     * We used to handle "expected errors" differently from unexpected ones.
     * Now we handle all errors in the same way (unless the show_econnreset
     * socket option is enabled). We just have to distinguish between passive
     * and active sockets.
     */
    DEBUGF(("driver_failure_eof(%ld) in %s, line %d\r\n",
	    (long)desc->inet.port, __FILE__, __LINE__));
    if (desc->inet.active) {
        ErlDrvTermData err_atom;
	if (show_econnreset) {
	    tcp_error_message(desc, err);
            err_atom = error_atom(err);
	} else {
            err_atom = am_closed;
	}
        tcp_closed_message(desc);
        if (!(desc->tcp_add_flags & TCP_ADDF_SENDFILE))
            inet_reply_error_am(INETP(desc), err_atom);

	if (desc->inet.exitf)
	    driver_exit(desc->inet.port, 0);
	else
	    tcp_desc_close(desc);
    } else {
	tcp_close_check(desc);

	if (desc->inet.caller) {
            if (!(desc->tcp_add_flags & TCP_ADDF_SENDFILE)) {
                if (show_econnreset)
                    inet_reply_error(INETP(desc), err);
                else
                    inet_reply_error_am(INETP(desc), am_closed);
            }
	}
	else {
	    /* No blocking send op to reply to right now.
	     * If next op is a send, make sure it returns {error,closed}
	     * rather than {error,enotconn}.
	     */
	    desc->tcp_add_flags |= TCP_ADDF_DELAYED_CLOSE_SEND;
	}
        tcp_desc_close(desc);

	/*
	 * Make sure that the next receive operation gets an {error,closed}
	 * result rather than {error,enotconn}. That means that the caller
	 * can safely ignore errors in the send operations and handle them
	 * in the receive operation.
	 */
	desc->tcp_add_flags |= TCP_ADDF_DELAYED_CLOSE_RECV;

	if (show_econnreset) {
	    /* Return {error, econnreset} instead of {error, closed}
	     * on send or receive operations.
	     */
	    desc->tcp_add_flags |= TCP_ADDF_DELAYED_ECONNRESET;
	}
    }
    return -1;
}

static int tcp_send_error(tcp_descriptor* desc, int err)
{
    /* EPIPE errors usually occur in one of three ways:
     * 1. We write to a socket when we've already shutdown() the write side. On
     *    Windows the error returned for this is ESHUTDOWN rather than EPIPE.
     * 2. The TCP peer sends us an RST through no fault of our own (perhaps
     *    by aborting the connection using SO_LINGER) and we then attempt
     *    to write to the socket. On Linux and Windows we would actually
     *    receive an ECONNRESET error for this, but on the BSDs, Darwin,
     *    Illumos and presumably Solaris, it's an EPIPE.
     * 3. We cause the TCP peer to send us an RST by writing to a socket
     *    after we receive a FIN from them. Our first write will be
     *    successful, but if the they have closed the connection (rather
     *    than just shutting down the write side of it) this will cause their
     *    OS to send us an RST. Then, when we attempt to write to the socket
     *    a second time, we will get an EPIPE error. On Windows we get an
     *    ECONNABORTED.
     *
     * What we are going to do here is to treat all EPIPE messages that aren't
     * of type 1 as ECONNRESET errors. This will allow users who have the
     * show_econnreset socket option enabled to receive {error, econnreset} on
     * both send and recv operations to indicate that an RST has been received.
     */
#ifdef __WIN_32__
    if (err == ECONNABORTED)
	err = ECONNRESET;
#endif
    if (err == EPIPE && !(desc->tcp_add_flags & TCP_ADDF_SHUTDOWN_WR_DONE))
	err = ECONNRESET;
    return tcp_send_or_shutdown_error(desc, err);
}

static int tcp_shutdown_error(tcp_descriptor* desc, int err)
{
    return tcp_send_or_shutdown_error(desc, err);
}

static void tcp_inet_delay_send(ErlDrvData data, ErlDrvTermData dummy)
{
    tcp_descriptor *desc = (tcp_descriptor*)data;
    (void)tcp_inet_output(desc, (HANDLE) INETP(desc)->s);
}

/*
** Send non-blocking vector data
*/
static int tcp_sendv(tcp_descriptor* desc, ErlIOVec* ev)
{
    ErlDrvSizeT sz;
    char buf[4];
    ErlDrvSizeT h_len;
    ssize_t n;
    ErlDrvPort ix = desc->inet.port;
    ErlDrvSizeT len = ev->size;

     switch(desc->inet.htype) {
     case TCP_PB_1:
         put_int8(len, buf);
         h_len = 1;
         break;
     case TCP_PB_2:
         put_int16(len, buf);
         h_len = 2;
         break;
     case TCP_PB_4:
         put_int32(len, buf);
         h_len = 4;
         break;
     default:
         if (len == 0)
             return 0;
         h_len = 0;
         break;
     }

    inet_output_count(INETP(desc), len+h_len);

    if (h_len > 0) {
	ev->iov[0].iov_base = buf;
	ev->iov[0].iov_len = h_len;
	ev->size += h_len;
    }

    sz = driver_sizeq(ix);

    if ((desc->tcp_add_flags & TCP_ADDF_SENDFILE) || sz > 0) {
	driver_enqv(ix, ev, 0);
	if (sz+ev->size >= desc->high) {
	    DEBUGF(("tcp_sendv(%ld): s=%d, sender forced busy\r\n",
		    (long)desc->inet.port, desc->inet.s));
	    desc->inet.state |= INET_F_BUSY;  /* mark for low-watermark */
	    desc->inet.busy_caller = desc->inet.caller;
	    set_busy_port(desc->inet.port, 1);
	    if (desc->send_timeout != INET_INFINITY) {
		desc->busy_on_send = 1;
                add_multi_timer(desc, INETP(desc)->port,
                                0 /* arg */, desc->send_timeout /* timeout */,
                                &tcp_inet_send_timeout);
	    }
	    return 1;
	}
    }
    else {
	int vsize = (ev->vsize > MAX_VSIZE) ? MAX_VSIZE : ev->vsize;

	DEBUGF(("tcp_sendv(%ld): s=%d, about to send "LLU","LLU" bytes\r\n",
		(long)desc->inet.port, desc->inet.s, (llu_t)h_len, (llu_t)len));

	if (INET_IGNORED(INETP(desc))) {
	    INETP(desc)->flags |= INET_IGNORE_WRITE;
	    n = 0;
	} else if (desc->tcp_add_flags & TCP_ADDF_DELAY_SEND) {
            driver_enqv(ix, ev, 0);
            add_multi_timer(desc, INETP(desc)->port, 0,
                            0, &tcp_inet_delay_send);
	    return 0;
	} else if (IS_SOCKET_ERROR(sock_sendv(desc->inet.s, ev->iov,
					      vsize, &n, 0))) {
	    if ((sock_errno() != ERRNO_BLOCK) && (sock_errno() != EINTR)) {
		int err = sock_errno();
		DEBUGF(("tcp_sendv(%ld): s=%d, "
			"sock_sendv(size=2) errno = %d\r\n",
			(long)desc->inet.port, desc->inet.s, err));
		return tcp_send_error(desc, err);
	    }
#ifdef __WIN32__
	    desc->inet.send_would_block = 1;
#endif
	    n = 0;
	}
	else if (n == ev->size) {
	    ASSERT(NO_SUBSCRIBERS(&INETP(desc)->empty_out_q_subs));
	    return 0;
	}
	else {
	    DEBUGF(("tcp_sendv(%ld): s=%d, only sent "
		    LLU"/%d of "LLU"/%d bytes/items\r\n",
		    (long)desc->inet.port, desc->inet.s,
		    (llu_t)n, vsize, (llu_t)ev->size, ev->vsize));
	}

	DEBUGF(("tcp_sendv(%ld): s=%d, Send failed, queuing\r\n",
		(long)desc->inet.port, desc->inet.s));
	driver_enqv(ix, ev, n);
	if (!INET_IGNORED(INETP(desc)))
	    sock_select(INETP(desc),(FD_WRITE|FD_CLOSE), 1);
    }
    return 0;
}

/*
** Send non blocking data
*/
static int tcp_send(tcp_descriptor* desc, char* ptr, ErlDrvSizeT len)
{
    int sz;
    char buf[4];
    int h_len;
    int n;
    ErlDrvPort ix = desc->inet.port;
    SysIOVec iov[2];

    switch(desc->inet.htype) {
    case TCP_PB_1:
	put_int8(len, buf);
	h_len = 1;
	break;
    case TCP_PB_2:
	put_int16(len, buf);
	h_len = 2;
	break;
    case TCP_PB_4:
	put_int32(len, buf);
	h_len = 4;
	break;
    default:
	if (len == 0)
	    return 0;
	h_len = 0;
	break;
    }

    inet_output_count(INETP(desc), len+h_len);

    sz = driver_sizeq(ix);

    if ((desc->tcp_add_flags & TCP_ADDF_SENDFILE) || sz > 0) {
	if (h_len > 0)
	    driver_enq(ix, buf, h_len);
	driver_enq(ix, ptr, len);
	if (sz+h_len+len >= desc->high) {
	    DEBUGF(("tcp_send(%ld): s=%d, sender forced busy\r\n",
		    (long)desc->inet.port, desc->inet.s));
	    desc->inet.state |= INET_F_BUSY;  /* mark for low-watermark */
	    desc->inet.busy_caller = desc->inet.caller;
	    set_busy_port(desc->inet.port, 1);
	    if (desc->send_timeout != INET_INFINITY) {
		desc->busy_on_send = 1;
                add_multi_timer(desc, INETP(desc)->port,
                                0 /* arg */, desc->send_timeout /* timeout */,
                                &tcp_inet_send_timeout);
	    }
	    return 1;
	}
    }
    else {
	iov[0].iov_base = buf;
	iov[0].iov_len = h_len;
	iov[1].iov_base = ptr;
	iov[1].iov_len = len;

	DEBUGF(("tcp_send(%ld): s=%d, about to send "LLU","LLU" bytes\r\n",
		(long)desc->inet.port, desc->inet.s, (llu_t)h_len, (llu_t)len));
	if (INET_IGNORED(INETP(desc))) {
	    INETP(desc)->flags |= INET_IGNORE_WRITE;
	    n = 0;
	} else if (desc->tcp_add_flags & TCP_ADDF_DELAY_SEND) {
	    sock_send(desc->inet.s, buf, 0, 0);
	    n = 0;
	} else 	if (IS_SOCKET_ERROR(sock_sendv(desc->inet.s,iov,2,&n,0))) {
	    if ((sock_errno() != ERRNO_BLOCK) && (sock_errno() != EINTR)) {
		int err = sock_errno();
		DEBUGF(("tcp_send(%ld): s=%d,sock_sendv(size=2) errno = %d\r\n",
			(long)desc->inet.port, desc->inet.s, err));
		return tcp_send_error(desc, err);
	    }
#ifdef __WIN32__
	    desc->inet.send_would_block = 1;
#endif
	    n = 0;
	}
	else if (n == len+h_len) {
	    ASSERT(NO_SUBSCRIBERS(&INETP(desc)->empty_out_q_subs));
	    return 0;
	}

	DEBUGF(("tcp_send(%ld): s=%d, Send failed, queuing",
		(long)desc->inet.port, desc->inet.s));

	if (n < h_len) {
	    driver_enq(ix, buf+n, h_len-n);
	    driver_enq(ix, ptr, len);
	}
	else {
	    n -= h_len;
	    driver_enq(ix, ptr+n, len-n);
	}
	if (!INET_IGNORED(INETP(desc)))
	    sock_select(INETP(desc),(FD_WRITE|FD_CLOSE), 1);
    }
    return 0;
}

/* shutdown on the socket:
** Assume caller has confirmed TCP_ADDF_PENDING_SHUTDOWN is set.
*/
static void tcp_shutdown_async(tcp_descriptor* desc)
{
    int how;

    how = (desc->tcp_add_flags & TCP_ADDF_PENDING_SHUT_WR) ?
		TCP_SHUT_WR : TCP_SHUT_RDWR;
    if (IS_SOCKET_ERROR(sock_shutdown(INETP(desc)->s, how)))
	tcp_shutdown_error(desc, sock_errno());
    else
	desc->tcp_add_flags |= TCP_ADDF_SHUTDOWN_WR_DONE;
}

static void tcp_inet_drv_output(ErlDrvData data, ErlDrvEvent event)
{
    (void)tcp_inet_output((tcp_descriptor*)data, (HANDLE)event);
}

static void tcp_inet_drv_input(ErlDrvData data, ErlDrvEvent event)
{
    (void)tcp_inet_input((tcp_descriptor*)data, (HANDLE)event);
}

#ifdef HAVE_SENDFILE
static int tcp_sendfile_completed(tcp_descriptor* desc) {
    ErlDrvTermData spec[LOAD_PORT_CNT + LOAD_TUPLE_CNT * 2 +
        LOAD_ATOM_CNT * 2 + LOAD_UINT_CNT * 2];
    Uint32 sent_low, sent_high;
    int i;

    desc->tcp_add_flags &= ~TCP_ADDF_SENDFILE;
    close(desc->sendfile.dup_file_fd);

    DEBUGF(("tcp_sendfile_completed(%p): SENDFILE dup closed %d\r\n",
            desc->inet.port, desc->sendfile.dup_file_fd));

    /* While we flushed the output queue prior to sending the file, we've
     * deferred clearing busy status until now as there's no point in doing so
     * while we still have a file to send.
     *
     * The watermark is checked since more data may have been added while we
     * were sending the file. */

    if (driver_sizeq(desc->inet.port) <= desc->low) {
        if (IS_BUSY(INETP(desc))) {
            desc->inet.caller = desc->inet.busy_caller;
            desc->inet.state &= ~INET_F_BUSY;

            set_busy_port(desc->inet.port, 0);

            /* if we have a timer then cancel and send ok to client */
            if (desc->busy_on_send) {
                cancel_multi_timer(desc, INETP(desc)->port,
                                   &tcp_inet_send_timeout);
                desc->busy_on_send = 0;
            }

            inet_reply_ok(INETP(desc));
        }
    }

    if (driver_sizeq(desc->inet.port) == 0) {
        sock_select(INETP(desc), FD_WRITE, 0);
        send_empty_out_q_msgs(INETP(desc));

        if (desc->tcp_add_flags & TCP_ADDF_PENDING_SHUTDOWN) {
            tcp_shutdown_async(desc);
        }
    }

    sent_low = ((Uint64)desc->sendfile.bytes_sent >> 0) & 0xFFFFFFFF;
    sent_high = ((Uint64)desc->sendfile.bytes_sent >> 32) & 0xFFFFFFFF;

    i = LOAD_ATOM(spec, 0, am_sendfile);
    i = LOAD_PORT(spec, i, desc->inet.dport);
    i = LOAD_ATOM(spec, i, am_ok);
    i = LOAD_UINT(spec, i, sent_low);
    i = LOAD_UINT(spec, i, sent_high);
    i = LOAD_TUPLE(spec, i, 3);
    i = LOAD_TUPLE(spec, i, 3);

    ASSERT(i == sizeof(spec)/sizeof(*spec));

    return erl_drv_output_term(desc->inet.dport, spec, i);
}

static int tcp_sendfile_aborted(tcp_descriptor* desc, int socket_error) {
    ErlDrvTermData spec[LOAD_PORT_CNT + LOAD_TUPLE_CNT * 2 + LOAD_ATOM_CNT * 3];
    int i;

    /* We don't clean up sendfile state here, as that's done in tcp_desc_close
     * following normal error handling. All we do here is report the failure. */

    i = LOAD_ATOM(spec, 0, am_sendfile);
    i = LOAD_PORT(spec, i, desc->inet.dport);
    i = LOAD_ATOM(spec, i, am_error);

    switch (socket_error) {
    case ECONNRESET:
    case ENOTCONN:
    case EPIPE:
        i = LOAD_ATOM(spec, i, am_closed);
        break;
    default:
        i = LOAD_ATOM(spec, i, error_atom(socket_error));
    }

    i = LOAD_TUPLE(spec, i, 2);
    i = LOAD_TUPLE(spec, i, 3);

    ASSERT(i == sizeof(spec)/sizeof(*spec));

    return erl_drv_output_term(desc->inet.dport, spec, i);
}

static int tcp_inet_sendfile(tcp_descriptor* desc) {
    ErlDrvPort ix = desc->inet.port;
    int result = 0;
    ssize_t n;

    DEBUGF(("tcp_inet_sendfile(%ld) {s=%d\r\n", (long)ix, desc->inet.s));

    /* If there was any data in the queue by the time sendfile was issued,
     * we'll need to skip it first. Note that we don't clear busy status until
     * we're finished sending the file. */
    while (desc->sendfile.ioq_skip > 0) {
        ssize_t bytes_to_send;
        SysIOVec* iov;
        int vsize;

        ASSERT(driver_sizeq(ix) >= desc->sendfile.ioq_skip);

        if ((iov = driver_peekq(ix, &vsize)) == NULL) {
            ERTS_INTERNAL_ERROR("ioq empty when sendfile.ioq_skip > 0");
        }

        bytes_to_send = MIN(desc->sendfile.ioq_skip, iov[0].iov_len);
        n = sock_send(desc->inet.s, iov[0].iov_base, bytes_to_send, 0);

        if (!IS_SOCKET_ERROR(n)) {
            desc->sendfile.ioq_skip -= n;
            driver_deq(ix, n);
        } else if (sock_errno() == ERRNO_BLOCK) {
#ifdef __WIN32__
            desc->inet.send_would_block = 1;
#endif
            goto done;
        } else if (sock_errno() != EINTR) {
            goto socket_error;
        }
    }

    while (desc->sendfile.length > 0) {
        /* For some reason the maximum ssize_t cannot be used as the max size.
         * 1GB seems to work on all platforms */
        const Sint64 SENDFILE_CHUNK_SIZE = ((1UL << 30) - 1);

        ssize_t bytes_to_send = MIN(SENDFILE_CHUNK_SIZE, desc->sendfile.length);
        off_t offset = desc->sendfile.offset;

#if defined(__linux__)
        n = sendfile(desc->inet.s, desc->sendfile.dup_file_fd, &offset,
            bytes_to_send);
#elif defined(__FreeBSD__) || defined(__DragonFly__) || defined(__DARWIN__)
        {
            off_t bytes_sent;
            int error;

    #if defined(__DARWIN__)
            bytes_sent = bytes_to_send;

            error = sendfile(desc->sendfile.dup_file_fd, desc->inet.s, offset,
                &bytes_sent, NULL, 0);
            n = bytes_sent;
    #else
            error = sendfile(desc->sendfile.dup_file_fd, desc->inet.s, offset,
                bytes_to_send, NULL, &bytes_sent, 0);
            n = bytes_sent;
    #endif

            if(error < 0) {
                /* EAGAIN/EINTR report partial success by setting bytes_sent,
                 * so we have to skip error handling if nonzero, and skip EOF
                 * handling if zero, as it's possible that we didn't manage to
                 * send anything at all before being interrupted by a
                 * signal. */
                if((errno != EAGAIN && errno != EINTR) || bytes_sent == 0) {
                    n = -1;
                }
            }
        }
#elif defined(__sun) && defined(__SVR4) && defined(HAVE_SENDFILEV)
        {
            sendfilevec_t sfvec[1];
            size_t bytes_sent;
            ssize_t error;

            sfvec[0].sfv_fd = desc->sendfile.dup_file_fd;
            sfvec[0].sfv_len = bytes_to_send;
            sfvec[0].sfv_off = offset;
            sfvec[0].sfv_flag = 0;

            error = sendfilev(desc->inet.s, sfvec, 1, &bytes_sent);
            n = bytes_sent;

            if(error < 0) {
                if(errno == EINVAL) {
                    /* On some solaris versions (I've seen it on SunOS 5.10),
                     * using a sfv_len larger than the filesize will result in
                     * a (-1 && errno == EINVAL). We translate this to a
                     * successful send of the data.*/
                } else {
                    /* EAGAIN/EINTR behavior is identical to *BSD. */
                    if((errno != EAGAIN && errno != EINTR) || bytes_sent == 0) {
                        n = -1;
                    }
                }
            }
        }
#else
        #error "Unsupported sendfile syscall; update configure test."
#endif

        if (n > 0) {
            desc->sendfile.bytes_sent += n;
            desc->sendfile.offset += n;
            desc->sendfile.length -= n;
        } else if (n == 0) {
            /* EOF. */
            desc->sendfile.length = 0;
            break;
        } else if (IS_SOCKET_ERROR(n) && sock_errno() != EINTR) {
            if (sock_errno() != ERRNO_BLOCK) {
                goto socket_error;
            }

#ifdef __WIN32__
            desc->inet.send_would_block = 1;
#endif
            break;
        }
    }

    if (desc->sendfile.length == 0) {
        tcp_sendfile_completed(desc);
    }

    goto done;

socket_error: {
        int socket_errno = sock_errno();

        DEBUGF(("tcp_inet_sendfile(%ld): send errno = %d (errno %d)\r\n",
            (long)desc->inet.port, socket_errno, errno));

        tcp_sendfile_aborted(desc, socket_errno);
        result = tcp_send_error(desc, socket_errno);

        goto done;
    }

done:
    DEBUGF(("tcp_inet_sendfile(%ld) }\r\n", (long)desc->inet.port));
    return result;
}
#endif /* HAVE_SENDFILE */

/* socket ready for ouput:
** 1. INET_STATE_CONNECTING => non block connect ?
** 2. INET_STATE_CONNECTED  => write output
*/
static int tcp_inet_output(tcp_descriptor* desc, HANDLE event)
{
    int ret = 0;
    ErlDrvPort ix = desc->inet.port;

    ASSERT(!INET_IGNORED(INETP(desc)));
    DEBUGF(("tcp_inet_output(%ld) {s=%d\r\n",
	    (long)desc->inet.port, desc->inet.s));
    if (desc->inet.state == INET_STATE_CONNECTING) {
	sock_select(INETP(desc),FD_CONNECT,0);

	driver_cancel_timer(ix);  /* posssibly cancel a timer */
#ifndef __WIN32__
	/*
	 * XXX This is strange.  This *should* work on Windows NT too,
	 * but doesn't.  An bug in Winsock 2.0 for Windows NT?
	 *
	 * See "Unix Netwok Programming", W.R.Stevens, p 412 for a
	 * discussion about Unix portability and non blocking connect.
	 */

#ifndef SO_ERROR
	{
	    int sz, code;
            sz = sizeof(desc->inet.remote);
            sys_memzero((char *) &desc->inet.remote, sz);
	    code = sock_peer(desc->inet.s,
                             (struct sockaddr*) &desc->inet.remote, &sz);
	    if (IS_SOCKET_ERROR(code)) {
		desc->inet.state = INET_STATE_OPEN;  /* restore state */
		ret =  async_error(INETP(desc), sock_errno());
		goto done;
	    }
	}
#else
	{
	    int error = 0;	/* Has to be initiated, we check it */
	    unsigned int sz = sizeof(error); /* even if we get -1 */
	    int code = sock_getopt(desc->inet.s, SOL_SOCKET, SO_ERROR,
				   (void *)&error, &sz);

	    if ((code < 0) || error) {
		desc->inet.state = INET_STATE_OPEN;  /* restore state */
		ret = async_error(INETP(desc), error);
		goto done;
	    }
	}
#endif /* SO_ERROR */
#endif /* !__WIN32__ */

	desc->inet.state = INET_STATE_CONNECTED;
	if (desc->inet.active)
	    sock_select(INETP(desc),(FD_READ|FD_CLOSE),1);
	async_ok(INETP(desc));
    }
    else if (IS_CONNECTED(INETP(desc))) {

#ifdef HAVE_SENDFILE
        if(desc->tcp_add_flags & TCP_ADDF_SENDFILE) {
            return tcp_inet_sendfile(desc);
        }
#endif

        for (;;) {
	    int vsize;
	    ssize_t n;
	    SysIOVec* iov;

	    if ((iov = driver_peekq(ix, &vsize)) == NULL) {
		sock_select(INETP(desc), FD_WRITE, 0);
		send_empty_out_q_msgs(INETP(desc));
		if (desc->tcp_add_flags & TCP_ADDF_PENDING_SHUTDOWN)
		    tcp_shutdown_async(desc);
		goto done;
	    }
	    vsize = vsize > MAX_VSIZE ? MAX_VSIZE : vsize;
	    DEBUGF(("tcp_inet_output(%ld): s=%d, About to send %d items\r\n",
		    (long)desc->inet.port, desc->inet.s, vsize));
	    if (IS_SOCKET_ERROR(sock_sendv(desc->inet.s, iov, vsize, &n, 0))) {
	    write_error:
		if ((sock_errno() != ERRNO_BLOCK) && (sock_errno() != EINTR)) {
		    DEBUGF(("tcp_inet_output(%ld): sock_sendv(%d) errno = %d (errno %d)\r\n",
			    (long)desc->inet.port, vsize, sock_errno(), errno));
		    ret =  tcp_send_error(desc, sock_errno());
		    goto done;
		}
#ifdef __WIN32__
		desc->inet.send_would_block = 1;
#endif
                /* If DELAY_SEND is set ready_output may have
                   been called without doing select so we do
                   a select in order to get into the correct
                   state */
                if (desc->tcp_add_flags & TCP_ADDF_DELAY_SEND)
                    sock_select(INETP(desc), FD_WRITE, 1);
		goto done;
	    } else if (n == 0) { /* Workaround for redhat/CentOS 6.3 returning
				    0 when sending packets with
				    sizes > (max 32 bit signed int) */
	      size_t howmuch = 0x7FFFFFFF; /* max signed 32 bit */
	      int x;
	      for(x = 0; x < vsize && iov[x].iov_len == 0; ++x)
		;
	      if (x < vsize) {
		if (howmuch > iov[x].iov_len) {
		  howmuch = iov[x].iov_len;
		}
		n = sock_send(desc->inet.s, iov[x].iov_base,howmuch,0);
		if (IS_SOCKET_ERROR(n)) {
		  goto write_error;
		}
	      }
	    }
	    if (driver_deq(ix, n) <= desc->low) {
		if (IS_BUSY(INETP(desc))) {
		    desc->inet.caller = desc->inet.busy_caller;
		    desc->inet.state &= ~INET_F_BUSY;
		    set_busy_port(desc->inet.port, 0);
		    /* if we have a timer then cancel and send ok to client */
		    if (desc->busy_on_send) {
                        cancel_multi_timer(desc, INETP(desc)->port, &tcp_inet_send_timeout);
			desc->busy_on_send = 0;
		    }
		    inet_reply_ok(INETP(desc));
		}
	    }
	}
    }
    else {
	sock_select(INETP(desc),FD_CONNECT,0);
	DEBUGF(("tcp_inet_output(%ld): bad state: %04x\r\n",
		(long)desc->inet.port, desc->inet.state));
    }
 done:
    DEBUGF(("tcp_inet_output(%ld) }\r\n", (long)desc->inet.port));
    return ret;
}

/*-----------------------------------------------------------------------------

   UDP & SCTP (the latter in a 1<->M Mode)

-----------------------------------------------------------------------------*/

#if defined(HAVE_SO_BSDCOMPAT)
#if defined(__linux__)
#include <sys/utsname.h>
static int should_use_so_bsdcompat(void)
{
    /* SMP: FIXME this is probably not SMP safe but may be ok anyway? */
    static int init_done;
    static int so_bsdcompat_is_obsolete;

    if (!init_done) {
	struct utsname utsname;
	unsigned int version, patchlevel;

	init_done = 1;
	if (uname(&utsname) < 0) {
	    fprintf(stderr, "uname: %s\r\n", strerror(sock_errno()));
	    return 1;
	}
	/* Format is <version>.<patchlevel>.<sublevel><extraversion>
	   where the first three are unsigned integers and the last
	   is an arbitrary string. We only care about the first two. */
	if (sscanf(utsname.release, "%u.%u", &version, &patchlevel) != 2) {
	    fprintf(stderr, "uname: unexpected release '%s'\r\n",
		    utsname.release);
	    return 1;
	}
	/* SO_BSDCOMPAT is deprecated and triggers warnings in 2.5
	   kernels. It is a no-op in 2.4 but not in 2.2 kernels. */
	if (version > 2 || (version == 2 && patchlevel >= 5))
	    so_bsdcompat_is_obsolete = 1;
    }
    return !so_bsdcompat_is_obsolete;
}
#else	/* __linux__ */
#define should_use_so_bsdcompat() 1
#endif	/* __linux__ */
#endif	/* HAVE_SO_BSDCOMPAT */



#ifdef HAVE_SCTP
/* Copy a descriptor, by creating a new port with same settings
 * as the descriptor desc.
 * return NULL on error (ENFILE no ports avail)
 */
static ErlDrvData packet_inet_start(ErlDrvPort port, char* args, int protocol);

static udp_descriptor* sctp_inet_copy(udp_descriptor* desc, SOCKET s, int* err)
{
    ErlDrvSizeT q_low, q_high;
    ErlDrvPort port = desc->inet.port;
    udp_descriptor* copy_desc;

    copy_desc = (udp_descriptor*) packet_inet_start(port, NULL, IPPROTO_SCTP);

    /* Setup event if needed */
    if ((copy_desc->inet.s = s) != INVALID_SOCKET) {
	if ((copy_desc->inet.event = sock_create_event(INETP(copy_desc))) ==
	    INVALID_EVENT) {
	    *err = sock_errno();
	    FREE(copy_desc);
	    return NULL;
	}
    }

    /* Some flags must be inherited at this point */
    copy_desc->inet.mode     = desc->inet.mode;
    copy_desc->inet.exitf    = desc->inet.exitf;
    copy_desc->inet.deliver  = desc->inet.deliver;
    copy_desc->inet.htype    = desc->inet.htype;
    copy_desc->inet.psize    = desc->inet.psize;
    copy_desc->inet.stype    = desc->inet.stype;
    copy_desc->inet.sfamily  = desc->inet.sfamily;
    copy_desc->inet.hsz      = desc->inet.hsz;
    copy_desc->inet.bufsz    = desc->inet.bufsz;

    /* The new port will be linked and connected to the caller */
    port = driver_create_port(port, desc->inet.caller, "sctp_inet",
			      (ErlDrvData) copy_desc);
    if ((long)port == -1) {
	*err = ENFILE;
	FREE(copy_desc);
	return NULL;
    }

    /* Read busy msgq limits of parent */
    q_low = q_high = ERL_DRV_BUSY_MSGQ_READ_ONLY;
    erl_drv_busy_msgq_limits(desc->inet.port, &q_low, &q_high);
    /* Write same busy msgq limits to child */
    erl_drv_busy_msgq_limits(port, &q_low, &q_high);

    copy_desc->inet.port = port;
    copy_desc->inet.dport = driver_mk_port(port);
    *err = 0;

    return copy_desc;
}
#endif



#ifdef HAVE_UDP
static int packet_inet_init()
{
    return 0;
}

static ErlDrvData packet_inet_start(ErlDrvPort port, char* args, int protocol)
{
    /* "inet_start" returns "ErlDrvData", but in fact it is "inet_descriptor*",
       so we can preserve it as "ErlDrvData":
    */
    ErlDrvData	    drvd = inet_start(port, sizeof(udp_descriptor),
				      protocol);
    udp_descriptor* desc = (udp_descriptor*) drvd;

    if (desc == NULL)
	return ERL_DRV_ERROR_ERRNO;

    desc->read_packets = INET_PACKET_POLL;
    desc->i_bufsz = 0;
    desc->i_buf = NULL;
    desc->i_ptr = NULL;
    return drvd;
}

static ErlDrvData udp_inet_start(ErlDrvPort port, char *args)
{
    ErlDrvData data = packet_inet_start(port, args, IPPROTO_UDP);
    set_default_msgq_limits(port);
    return data;
}
#endif

#ifdef HAVE_SCTP
static ErlDrvData sctp_inet_start(ErlDrvPort port, char *args)
{
    ErlDrvData data = packet_inet_start(port, args, IPPROTO_SCTP);
    set_default_msgq_limits(port);
    return data;
}
#endif

#ifdef HAVE_UDP
static void packet_inet_stop(ErlDrvData e)
{
    /* There should *never* be any "empty out q" subscribers on
       an UDP or SCTP socket!
       NB: as in "inet_start", we  can always cast "ErlDRvData"
       into "udp_descriptor*" or "inet_descriptor*":
    */
    udp_descriptor * udesc = (udp_descriptor*) e;
    inet_descriptor* descr = INETP(udesc);
    if (udesc->i_buf != NULL) {
	release_buffer(udesc->i_buf);
	udesc->i_buf = NULL;
    }

    ASSERT(NO_SUBSCRIBERS(&(descr->empty_out_q_subs)));
    inet_stop(descr);
}

static int packet_error(udp_descriptor* udesc, int err)
{
    inet_descriptor * desc = INETP(udesc);
    if (!desc->active)
	async_error(desc, err);
    driver_failure_posix(desc->port, err);
    return -1;
}

/*
** Various functions accessible via "port_control" on the Erlang side:
*/
static ErlDrvSSizeT packet_inet_ctl(ErlDrvData e, unsigned int cmd, char* buf,
				    ErlDrvSizeT len, char** rbuf, ErlDrvSizeT rsize)
{
    ErlDrvSSizeT replen;
    udp_descriptor * udesc = (udp_descriptor *) e;
    inet_descriptor* desc  = INETP(udesc);
    int type = SOCK_DGRAM;
    int af = AF_INET;

    cmd -= ERTS_INET_DRV_CONTROL_MAGIC_NUMBER;
    switch(cmd) {
    case INET_REQ_OPEN:   /* open socket and return internal index */
	DEBUGF(("packet_inet_ctl(%ld): OPEN\r\n", (long)desc->port));
	if (len != 2) {
	    return ctl_error(EINVAL, rbuf, rsize);
	}
	switch (buf[0]) {
	case INET_AF_INET:  af = AF_INET; break;
#if defined(HAVE_IN6) && defined(AF_INET6)
	case INET_AF_INET6: af = AF_INET6; break;
#endif
#ifdef HAVE_SYS_UN_H
	case INET_AF_LOCAL: af = AF_UNIX; break;
#endif
	default:
	    return ctl_xerror(str_eafnosupport, rbuf, rsize);
	}
	switch (buf[1]) {
	case INET_TYPE_STREAM: type = SOCK_STREAM; break;
	case INET_TYPE_DGRAM: type = SOCK_DGRAM; break;
#ifdef HAVE_SCTP
	case INET_TYPE_SEQPACKET: type = SOCK_SEQPACKET; break;
#endif
	default:
	    return ctl_error(EINVAL, rbuf, rsize);
	}
	replen = inet_ctl_open(desc, af, type, rbuf, rsize);

	if ((*rbuf)[0] != INET_REP_ERROR) {
	    if (desc->active)
		sock_select(desc,FD_READ,1);
#ifdef HAVE_SO_BSDCOMPAT
	    /*
	     * Make sure that sending UDP packets to a non existing port on an
	     * existing machine doesn't close the socket. (Linux behaves this
	     * way)
	     */
	    if (should_use_so_bsdcompat()) {
		int one = 1;
		/* Ignore errors */
		sock_setopt(desc->s, SOL_SOCKET, SO_BSDCOMPAT, &one,
			    sizeof(one));
	    }
#endif
	}
	return replen;


    case INET_REQ_FDOPEN: {  /* pass in an open (and optionally bound) socket */
	SOCKET s;
        int bound;
	DEBUGF(("packet inet_ctl(%ld): FDOPEN\r\n", (long)desc->port));
	if (len != 6 && len != 10) {
	    return ctl_error(EINVAL, rbuf, rsize);
	}
	switch (buf[0]) {
	case INET_AF_INET:  af = AF_INET; break;
#if defined(HAVE_IN6) && defined(AF_INET6)
	case INET_AF_INET6: af = AF_INET6; break;
#endif
#ifdef HAVE_SYS_UN_H
	case INET_AF_LOCAL: af = AF_UNIX; break;
#endif
	default:
	    return ctl_xerror(str_eafnosupport, rbuf, rsize);
	}
	switch (buf[1]) {
	case INET_TYPE_STREAM: type = SOCK_STREAM; break;
	case INET_TYPE_DGRAM: type = SOCK_DGRAM; break;
#ifdef HAVE_SCTP
	case INET_TYPE_SEQPACKET: type = SOCK_SEQPACKET; break;
#endif
	default:
	    return ctl_error(EINVAL, rbuf, rsize);
	}
	s = (SOCKET)get_int32(buf+2);

        if (len == 6) bound = 1;
        else bound = get_int32(buf+2+4);

	replen = inet_ctl_fdopen(desc, af, type, s, bound, rbuf, rsize);

	if ((*rbuf)[0] != INET_REP_ERROR) {
	    if (desc->active)
		sock_select(desc,FD_READ,1);
#ifdef HAVE_SO_BSDCOMPAT
	    /*
	     * Make sure that sending UDP packets to a non existing port on an
	     * existing machine doesn't close the socket. (Linux behaves this
	     * way)
	     */
	    if (should_use_so_bsdcompat()) {
		int one = 1;
		/* Ignore errors */
		sock_setopt(desc->s, SOL_SOCKET, SO_BSDCOMPAT, &one,
			    sizeof(one));
	    }
#endif
	}
	return replen;
    }


    case INET_REQ_CLOSE:
	DEBUGF(("packet_inet_ctl(%ld): CLOSE\r\n", (long)desc->port));
	erl_inet_close(desc);
	return ctl_reply(INET_REP_OK, NULL, 0, rbuf, rsize);


    case INET_REQ_CONNECT:  {
	/* UDP and SCTP connect operations are completely different. UDP
	   connect means only setting the default peer addr locally,  so
	   it is always synchronous. SCTP connect means actual establish-
	   ing of an SCTP association with a remote peer, so it is async-
	   ronous, and similar to TCP connect. However, unlike TCP, SCTP
	   allows the socket to have multiple simultaneous associations:
	*/
	int code;
	char tbuf[2];
#ifdef HAVE_SCTP
	unsigned timeout;
#endif
	DEBUGF(("packet_inet_ctl(%ld): CONNECT\r\n", (long)desc->port));

	/* INPUT: [ Timeout(4), Port(2), Address(N) ] */

	if (!IS_OPEN(desc))
	    return ctl_xerror(EXBADPORT, rbuf, rsize);

#ifdef HAVE_SCTP
	if (IS_SCTP(desc)) {
	    inet_address remote;
	    char *xerror;

	    if (IS_CONNECTING(desc))
		return ctl_error(EINVAL, rbuf, rsize);
	    if (len < 6)
		return ctl_error(EINVAL, rbuf, rsize);
	    timeout = get_int32(buf);
	    buf += 4;
	    len -= 4;

	    /* For SCTP, we do not set the peer's addr in desc->remote, as
	       multiple peers are possible: */
	    if ((xerror = inet_set_faddress
		 (desc->sfamily, &remote, &buf, &len)) != NULL)
	        return ctl_xerror(xerror, rbuf, rsize);

	    code = sock_connect(desc->s, &remote.sa, len);

	    if (IS_SOCKET_ERROR(code) && (sock_errno() == EINPROGRESS)) {
		/* XXX: Unix only -- WinSock would have a different cond! */
		if (timeout != INET_INFINITY)
		    driver_set_timer(desc->port, timeout);
		enq_async(desc, tbuf, INET_REQ_CONNECT);
		async_ok(desc);
	    }
	    else if (code == 0) { /* OK we are connected */
		enq_async(desc, tbuf, INET_REQ_CONNECT);
		async_ok(desc);
	    }
	    else {
		return ctl_error(sock_errno(), rbuf, rsize);
	    }
	    return ctl_reply(INET_REP_OK, tbuf, 2, rbuf, rsize);
	}
#endif
	/* UDP */
	if (len == 0) {
	    /* What does it mean???  NULL sockaddr??? */
	    sock_connect(desc->s, (struct sockaddr*) NULL, 0);
	    desc->state &= ~INET_F_ACTIVE;
	    enq_async(desc, tbuf, INET_REQ_CONNECT);
	    async_ok (desc);
	}
	else if (len < 6)
	    return ctl_error(EINVAL, rbuf, rsize);
	else {
	    char *xerror;
	    /* Ignore timeout */
	    buf += 4;
	    len -= 4;
	    if ((xerror = inet_set_faddress
		 (desc->sfamily, &desc->remote, &buf, &len)) != NULL)
	        return ctl_xerror(xerror, rbuf, rsize);

	    code = sock_connect(desc->s,
				(struct sockaddr*) &desc->remote, len);
	    if (IS_SOCKET_ERROR(code)) {
		sock_connect(desc->s, (struct sockaddr*) NULL, 0);
		desc->state &= ~INET_F_ACTIVE;
		return ctl_error(sock_errno(), rbuf, rsize);
	    }
	    else /* ok we are connected */ {
		enq_async(desc, tbuf, INET_REQ_CONNECT);
		desc->state |= INET_F_ACTIVE;
		async_ok (desc);
	    }
	}
	return ctl_reply(INET_REP_OK, tbuf, 2, rbuf, rsize);
    }

#ifdef HAVE_SCTP
    case INET_REQ_LISTEN:
	{	/* LISTEN is only for SCTP sockets, not UDP. This code is borrowed
		   from the TCP section. Returns: {ok,[]} on success.
		*/
	    int backlog;

	    DEBUGF(("packet_inet_ctl(%ld): LISTEN\r\n", (long)desc->port));
	    if (!IS_SCTP(desc))
		return ctl_xerror(EXBADPORT, rbuf, rsize);
	    if (!IS_OPEN(desc))
		return ctl_xerror(EXBADPORT, rbuf, rsize);

	    if (len != 2)
		return ctl_error(EINVAL, rbuf, rsize);
	    backlog = get_int16(buf);

	    if (IS_SOCKET_ERROR(sock_listen(desc->s, backlog)))
		return ctl_error(sock_errno(), rbuf, rsize);

	    desc->state = INET_STATE_LISTENING;   /* XXX: not used? */
	    return ctl_reply(INET_REP_OK, NULL, 0, rbuf, rsize);
	}

    case SCTP_REQ_BINDX:
	{   /* Multi-homing bind for SCTP: */
	    /* Add additional addresses by calling sctp_bindx with one address
	       at a time, since this is what some OSes promise will work.
	       Buff structure: Flags(1), ListItem,...:
	    */
	    inet_address addr;
	    char* curr;
	    int   add_flag, rflag;

	    if (!IS_SCTP(desc))
		return ctl_xerror(EXBADPORT, rbuf, rsize);

	    curr = buf;
	    add_flag = get_int8(curr);
	    curr++;

	    /* Make the real flags: */
	    rflag = add_flag ? SCTP_BINDX_ADD_ADDR : SCTP_BINDX_REM_ADDR;

	    while (curr < buf+len)
		{
		    char *xerror;
		    /* List item format: see "inet_set_faddress": */
		    ErlDrvSizeT alen  = buf + len - curr;
		    xerror = inet_set_faddress
		      (desc->sfamily, &addr, &curr, &alen);
		    if (xerror != NULL)
		        return ctl_xerror(xerror, rbuf, rsize);

		    /* Invoke the call: */
		    if (p_sctp_bindx(desc->s, (struct sockaddr *)&addr, 1,
				     rflag) < 0)
			return ctl_error(sock_errno(), rbuf, rsize);
		}

	    desc->state = INET_STATE_OPEN;

	    return ctl_reply(INET_REP_OK, NULL, 0, rbuf, rsize);
	}

    case SCTP_REQ_PEELOFF:
	{
	    Uint32 assoc_id;
	    udp_descriptor* new_udesc;
	    int err;
	    SOCKET new_socket;

	    DEBUGF(("packet_inet_ctl(%ld): PEELOFF\r\n", (long)desc->port));
	    if (!IS_SCTP(desc))
		return ctl_xerror(EXBADPORT, rbuf, rsize);
	    if (!IS_OPEN(desc))
		return ctl_xerror(EXBADPORT, rbuf, rsize);
	    if (! p_sctp_peeloff)
		return ctl_error(ENOTSUP, rbuf, rsize);

	    if (len != 4)
		return ctl_error(EINVAL, rbuf, rsize);
	    assoc_id = get_int32(buf);

	    new_socket = p_sctp_peeloff(desc->s, assoc_id);
	    if (IS_SOCKET_ERROR(new_socket)) {
		return ctl_error(sock_errno(), rbuf, rsize);
	    }

	    desc->caller = driver_caller(desc->port);
	    if ((new_udesc = sctp_inet_copy(udesc, new_socket, &err)) == NULL) {
		sock_close(new_socket);
		desc->caller = 0;
		return ctl_error(err, rbuf, rsize);
	    }
	    new_udesc->inet.state = INET_STATE_CONNECTED;
	    new_udesc->inet.stype = SOCK_STREAM;
	    SET_NONBLOCKING(new_udesc->inet.s);

	    inet_reply_ok_port(desc, new_udesc->inet.dport);
	    (*rbuf)[0] = INET_REP;
	    return 1;
	}
#endif  /* HAVE_SCTP */

    case PACKET_REQ_RECV:
	{	/* THIS IS A FRONT-END for "recv*" requests. It only enqueues the
		   request  and possibly returns the data  immediately available.
		   The actual data returning function is the back-end ("*input"):
		*/
	    unsigned timeout;
	    char tbuf[2];

	    DEBUGF(("packet_inet_ctl(%ld): RECV\r\n", (long)desc->port));
	    /* INPUT: Timeout(4), Length(4) */
	    if (!IS_OPEN(desc))
		return ctl_xerror(EXBADPORT, rbuf, rsize);
	    if (desc->active || (len != 8))
		return ctl_error(EINVAL, rbuf, rsize);
	    timeout = get_int32(buf);
	    /* The 2nd arg, Length(4), is ignored for both UDP and SCTP protocols,
	       since they are msg-oriented. */

	    if (enq_async(desc, tbuf, PACKET_REQ_RECV) < 0)
		return ctl_error(EALREADY, rbuf, rsize);

	    if (packet_inet_input(udesc, desc->event) == 0) {
		if (timeout == 0)
		    async_error_am(desc, am_timeout);
		else {
		    if (timeout != INET_INFINITY)
			driver_set_timer(desc->port, timeout);
		}
	    }
	    return ctl_reply(INET_REP_OK, tbuf, 2, rbuf, rsize);
	}

    default:
	/* Delegate the request to the INET layer. In particular,
	   INET_REQ_BIND goes here. If the req is not recognised
	   there either, an error is returned:
	*/
	return inet_ctl(desc, cmd, buf, len, rbuf, rsize);
    }
}

static void packet_inet_timeout(ErlDrvData e)
{
    udp_descriptor  * udesc = (udp_descriptor*) e;
    inet_descriptor * desc  = INETP(udesc);
    if (!(desc->active)) {
	sock_select(desc, FD_READ, 0);
        async_error_am (desc, am_timeout);
    } else {
        (void)packet_inet_input(udesc, (HANDLE) desc->s);
    }
}


/* THIS IS A "send*" REQUEST; on the Erlang side: "port_command".
** input should be: Family Address buffer .
** For UDP,  buffer (after Address) is just data to be sent.
** For SCTP, buffer contains a list representing 2 items:
**   (1) 6 parms for sctp_sndrcvinfo, as in sctp_get_sendparams();
**   (2) 0+ real data bytes.
** There is no destination address -- SCTP send is performed over
** an existing association, using "sctp_sndrcvinfo" specified.
*/
static void packet_inet_command(ErlDrvData e, char* buf, ErlDrvSizeT len)
{
    udp_descriptor * udesc= (udp_descriptor*) e;
    inet_descriptor* desc = INETP(udesc);
    char* ptr		  = buf;
    char* qtr;
    char* xerror;
    ErlDrvSizeT sz;
    int code;
    inet_address other;

    desc->caller = driver_caller(desc->port);

    if (!IS_OPEN(desc)) {
	inet_reply_error(desc, EINVAL);
	return;
    }

#ifdef HAVE_SCTP
    if (IS_SCTP(desc))
    {
	ErlDrvSizeT   data_len;
	struct iovec  iov[1];		 /* For real data              */
	struct msghdr mhdr;		 /* Message wrapper            */
	struct sctp_sndrcvinfo *sri;     /* The actual ancillary data  */
	union {                          /* For ancillary data         */
	    struct cmsghdr hdr;
	    char ancd[CMSG_SPACE(sizeof(*sri))];
	} cmsg;

	if (len < SCTP_GET_SENDPARAMS_LEN) goto return_einval;

	/* The ancillary data */
	sri = (struct sctp_sndrcvinfo *) (CMSG_DATA(&cmsg.hdr));
	/* Get the "sndrcvinfo" from the buffer, advancing the "ptr": */
	ptr  = sctp_get_sendparams(sri, ptr);

	/* The ancillary data wrapper */
	cmsg.hdr.cmsg_level = IPPROTO_SCTP;
	cmsg.hdr.cmsg_type  = SCTP_SNDRCV;
	cmsg.hdr.cmsg_len   = CMSG_LEN(sizeof(*sri));

	data_len = (buf + len) - ptr;
	/* The whole msg.
	 * Solaris (XPG 4.2) requires iovlen >= 1 even for data_len == 0.
	 */
	mhdr.msg_name           = NULL;	        /* Already connected  */
	mhdr.msg_namelen        = 0;
	iov[0].iov_len          = data_len;
	iov[0].iov_base         = ptr;          /* The real data */
	mhdr.msg_iov            = iov;
	mhdr.msg_iovlen         = 1;
	mhdr.msg_control        = cmsg.ancd;    /* For ancillary data  */
	mhdr.msg_controllen     = cmsg.hdr.cmsg_len;
	VALGRIND_MAKE_MEM_DEFINED(mhdr.msg_control, mhdr.msg_controllen); /*suppress "uninitialised bytes"*/
	mhdr.msg_flags          = 0;            /* Not used with "sendmsg"   */

	inet_output_count(desc, data_len);
	/* Now do the actual sending. NB: "flags" in "sendmsg" itself are NOT
	   used: */
	code = sock_sendmsg(desc->s, &mhdr, 0);
	goto check_result_code;
    }
#endif
    {
        ErlDrvSizeT anc_len;

        /* UDP socket. Even if it is connected, there is an address prefix
           here -- ignored for connected sockets: */
        sz = len;
        qtr = ptr;
        xerror = inet_set_faddress(desc->sfamily, &other, &qtr, &sz);
        if (xerror != NULL) {
            inet_reply_error_am(desc, driver_mk_atom(xerror));
            return;
        }
        len -= (qtr - ptr);
        ptr = qtr;

        /* Here comes ancillary data */
        if (len < 4) goto return_einval;
        anc_len = get_int32(ptr);
        len -= 4; ptr += 4;
        if (len < anc_len) goto return_einval;

        if (anc_len == 0 && !!0/*XXX-short-circuit-for-testing*/) {
            /* Empty ancillary data */
            /* Now "ptr" is the user data ptr, "len" is data length: */
            inet_output_count(desc, len);
            if (desc->state & INET_F_ACTIVE) {
                /* connected (ignore address) */
                code = sock_send(desc->s, ptr, len, 0);
            }
            else {
                code = sock_sendto(desc->s, ptr, len, 0, &other.sa, sz);
            }
        }
        else {
#ifdef __WIN32__
            goto return_einval; /* Can not send ancillary data on Windows */
#else
            struct iovec iov[1];
            struct msghdr mhdr;
            union { /* For ancillary data         */
                struct cmsghdr hdr;
                char ancd[SIZEOF_ANCILLARY_DATA];
            } cmsg;
            sys_memset(&iov, '\0', sizeof(iov));
            sys_memset(&mhdr, '\0', sizeof(mhdr));
            sys_memset(&cmsg, '\0', sizeof(cmsg));
            if (desc->state & INET_F_ACTIVE) {
                /* connected (ignore address) */
                mhdr.msg_name = NULL;
                mhdr.msg_namelen = 0;
            }
            else {
                mhdr.msg_name = &other;
                mhdr.msg_namelen = sz;
            }
            mhdr.msg_control = cmsg.ancd;
            mhdr.msg_controllen = sizeof(cmsg.ancd);
            if (compile_ancillary_data(&mhdr, ptr, anc_len) != 0) {
                goto return_einval;
            }
            if (mhdr.msg_controllen == 0) {
                /* XXX Testing - only possible for anc_len == 0 */
                mhdr.msg_control = NULL;
            }
            len -= anc_len;
            ptr += anc_len;
            /* Now "ptr" is the user data ptr, "len" is data length: */
            iov[0].iov_len = len;
            iov[0].iov_base = ptr;
            mhdr.msg_iov = iov;
            mhdr.msg_iovlen = 1;
            mhdr.msg_flags = 0;
            inet_output_count(desc, len);
            code = sock_sendmsg(desc->s, &mhdr, 0);
#endif
        }
    }

#ifdef HAVE_SCTP
 check_result_code:
    /* "code" analysis is the same for both SCTP and UDP cases above: */
#endif
    if (IS_SOCKET_ERROR(code)) {
	int err = sock_errno();
	inet_reply_error(desc, err);
    }
    else
	inet_reply_ok(desc);
    return;

 return_einval:
    inet_reply_error(desc, EINVAL);
    return;
}

#endif /* HAVE_UDP */


#ifdef __WIN32__
static void packet_inet_event(ErlDrvData e, ErlDrvEvent event)
{
    udp_descriptor * udesc = (udp_descriptor*)e;
    inet_descriptor* desc  = INETP(udesc);
    WSANETWORKEVENTS netEv;

    if ((WSAEnumNetworkEvents)(desc->s, desc->event, &netEv) != 0) {
	DEBUGF(( "port %d: EnumNetwrokEvents = %d\r\n",
		desc->port, sock_errno() ));
	return; /* -1; */
    }
    netEv.lNetworkEvents |= desc->forced_events;
    if (netEv.lNetworkEvents & FD_READ) {
	packet_inet_input(udesc, (HANDLE)event);
    }
}

#endif

#ifdef HAVE_UDP
static void packet_inet_drv_input(ErlDrvData e, ErlDrvEvent event)
{
    (void)  packet_inet_input((udp_descriptor*)e, (HANDLE)event);
}

/*
** THIS IS A BACK-END FOR "recv*" REQUEST, which actually receives the
**	data requested, and delivers them to the caller:
*/
static int packet_inet_input(udp_descriptor* udesc, HANDLE event)
{
    inet_descriptor* desc = INETP(udesc);
    int n;
    inet_address other;
    char abuf[sizeof(inet_address)];  /* buffer address; enough??? */
    int packet_count = udesc->read_packets;
    int count = 0;     /* number of packets delivered to owner */
#ifndef __WIN32__
    struct msghdr mhdr;	  	     /* Top-level msg structure    */
    struct iovec  iov[1]; 	     /* Data or Notification Event */
    char   ancd[ANC_BUFF_SIZE];      /* Ancillary Data		   */
#endif
#ifdef HAVE_SCTP
    int short_recv = 0;
#endif

    while(packet_count--) {
	unsigned int len = sizeof(other);

	sys_memzero((char *) &other, sizeof(other));

	/* udesc->i_buf is only kept between SCTP fragments */
#ifdef HAVE_SCTP
	if (udesc->i_buf != NULL) {
            ErlDrvBinary* tmp;
	    int bufsz;
            ASSERT(IS_SCTP(desc));
	    bufsz = desc->bufsz + (udesc->i_ptr - udesc->i_buf->orig_bytes);
	    if ((tmp = realloc_buffer(udesc->i_buf, bufsz)) == NULL) {
		release_buffer(udesc->i_buf);
		udesc->i_buf = NULL;
		return packet_error(udesc, ENOMEM);
	    } else {
		udesc->i_ptr =
		    tmp->orig_bytes + (udesc->i_ptr - udesc->i_buf->orig_bytes);
		udesc->i_buf = tmp;
		udesc->i_bufsz = bufsz;
	    }
	} else
#endif
        {
            ASSERT(udesc->i_buf == NULL);
	    udesc->i_bufsz = desc->bufsz + len;
	    if ((udesc->i_buf = alloc_buffer(udesc->i_bufsz)) == NULL)
		return packet_error(udesc, ENOMEM);
	    /* pointer to message start */
	    udesc->i_ptr = udesc->i_buf->orig_bytes + len;
	}

	/* Note: On Windows NT, recvfrom() fails if the socket is connected. */
#ifdef HAVE_SCTP
	/* For SCTP we must use recvmsg() */
	if (IS_SCTP(desc)) {
	    iov->iov_base = udesc->i_ptr; /* Data will come here    */
	    iov->iov_len = desc->bufsz; /* Remaining buffer space */

	    mhdr.msg_name	= &other; /* Peer addr comes into "other" */
	    mhdr.msg_namelen	= len;
	    mhdr.msg_iov	= iov;
	    mhdr.msg_iovlen	= 1;
	    mhdr.msg_control	= ancd;
	    mhdr.msg_controllen	= ANC_BUFF_SIZE;
	    mhdr.msg_flags	= 0;	   /* To be filled by "recvmsg"    */

	    /* Do the actual SCTP receive: */
	    n = sock_recvmsg(desc->s, &mhdr, 0);
	    len = mhdr.msg_namelen;
	    goto check_result;
	}
#endif
	/* Use recv() instead on connected sockets. */
	if ((desc->state & INET_F_ACTIVE)) {
	    n = sock_recv(desc->s, udesc->i_ptr, desc->bufsz, 0);
	    other = desc->remote;
	    goto check_result;
	}
#ifndef __WIN32__
        /* recvmsg() does not exist in the Winsock API */
        if (desc->recv_cmsgflags) {
            /* Use recvmsg() */
            iov->iov_base = udesc->i_ptr;
            iov->iov_len = desc->bufsz;
            mhdr.msg_name = &other;
            mhdr.msg_namelen = len;
            mhdr.msg_iov = iov;
            mhdr.msg_iovlen = 1;
	    mhdr.msg_control = ancd;
            mhdr.msg_controllen = ANC_BUFF_SIZE;
            mhdr.msg_flags = 0;
            n = sock_recvmsg(desc->s, &mhdr, 0);
            len = mhdr.msg_namelen;
            goto check_result;
        }
#endif
	n = sock_recvfrom(desc->s, udesc->i_ptr, desc->bufsz,
			  0, &other.sa, &len);
    check_result:
	/* Analyse the result: */
	if (IS_SOCKET_ERROR(n)) {
	    int err = sock_errno();
	    if (err != ERRNO_BLOCK) {
		/* real error */
		release_buffer(udesc->i_buf);
		udesc->i_buf = NULL;
		if (!desc->active) {
		    async_error(desc, err);
                    driver_cancel_timer(desc->port);
		    sock_select(desc,FD_READ,0);
		}
		else {
		    /* This is for an active desc only: */
		    packet_error_message(udesc, err);
		}
		return count;
	    }
	    /* would block error - try again */
	    if (!desc->active
#ifdef HAVE_SCTP
		|| short_recv
#endif
		) {
		sock_select(desc,FD_READ,1);
	    }
#ifdef HAVE_SCTP
            if (!short_recv) {
#endif
                release_buffer(udesc->i_buf);
                udesc->i_buf = NULL;
#ifdef HAVE_SCTP
            }
#endif
	    return count;		/* strange, not ready */
	}

#ifdef HAVE_SCTP
	if (IS_SCTP(desc) && (short_recv = !(mhdr.msg_flags & MSG_EOR))) {
	    /* SCTP non-final message fragment */
	    inet_input_count(desc, n);
	    udesc->i_ptr += n;
	    continue; /* wait for more fragments */
	}
#endif

	{
	    /* message received */
	    int code;
            void *mp;
	    char * ptr;
	    int nsz;

	    inet_input_count(desc, n);
	    udesc->i_ptr += n;
	    inet_get_address(abuf, &other, &len);
	    /* Copy formatted address to the buffer allocated; "len" is the
	       actual length which must be <= than the original reserved.
	       This means that the addr + data in the buffer are contiguous,
	       but they may start not at the "orig_bytes", instead at "ptr":
	    */
	    ASSERT (len <= sizeof(other));
	    ptr = udesc->i_buf->orig_bytes + sizeof(other) - len;
	    sys_memcpy(ptr, abuf, len);

	    nsz = udesc->i_ptr - ptr;

	    /* Check if we need to reallocate binary */
	    if ((desc->mode == INET_MODE_BINARY)
		&& (desc->hsz < (nsz - len))
		&& (nsz + BIN_REALLOC_MARGIN(desc->bufsz) < udesc->i_bufsz)) {
		ErlDrvBinary* tmp;
		int bufsz;
		bufsz = udesc->i_ptr - udesc->i_buf->orig_bytes;
		if ((tmp = realloc_buffer(udesc->i_buf, bufsz)) != NULL) {
		    udesc->i_buf = tmp;
		    udesc->i_bufsz = bufsz;
		    udesc->i_ptr = NULL;  /* not used from here */
		}
	    }
            mp = NULL;
#ifdef HAVE_SCTP
	    if (IS_SCTP(desc)) mp = &mhdr;
#endif
#ifndef __WIN32__
            if (desc->recv_cmsgflags) mp = &mhdr;
#endif
	    /* Actual parsing and return of the data received, occur here: */
	    code = packet_reply_binary_data(desc, len, udesc->i_buf,
					    (sizeof(other) - len),
					    nsz,
					    mp);
	    free_buffer(udesc->i_buf);
	    udesc->i_buf = NULL;
	    if (code < 0)
		return count;
	    count++;
	    if (!desc->active) {
                driver_cancel_timer(desc->port);
		sock_select(desc,FD_READ,0);
		return count;  /* passive mode (read one packet only) */
	    }
	}
    } /*  while(packet_count--) { */

    /* we ran out of tries (packet_count) either on an active socket
     * that got that many messages or an SCTP socket that got that
     * many message fragments but still not the final
     */
#ifdef HAVE_SCTP
    if (short_recv) {
	sock_select(desc, FD_READ, 1);
    }
#endif

    /* We set a timer on the port to trigger now.
       This emulates a "yield" operation as that is
       what we want to do here. We do *NOT* do a deselect
       as that is expensive, instead we check if the
       socket it still active when the timeout triggers
       and if it is not, then we just ignore the timeout */
    driver_set_timer(desc->port, 0);

    return count;
}

#endif

/*---------------------------------------------------------------------------*/

#ifdef __WIN32__

/*
 * Although we no longer need to lookup all of winsock2 dynamically,
 * there are still some function(s) we need to look up.
 */
static void find_dynamic_functions(void)
{
    char kernel_dll_name[] = "kernel32";
    HMODULE module;
    module = GetModuleHandle(kernel_dll_name);
    fpSetHandleInformation = (module != NULL) ?
	(BOOL (WINAPI *)(HANDLE,DWORD,DWORD))
	    GetProcAddress(module,"SetHandleInformation") :
	NULL;
}



/*
 * We must make sure that the socket handles are not inherited
 * by port programs (if there are inherited, the sockets will not
 * get closed when the emulator terminates, and epmd and other Erlang
 * nodes will not notice that we have exited).
 *
 * XXX It is not clear whether this works/is necessary in Windows 95.
 * There could also be problems with Winsock implementations from other
 * suppliers than Microsoft.
 */

static SOCKET
make_noninheritable_handle(SOCKET s)
{
    if (s != INVALID_SOCKET) {
	if (fpSetHandleInformation != NULL) {
	    (*fpSetHandleInformation)((HANDLE) s, HANDLE_FLAG_INHERIT, 0);
	} else {
	    HANDLE non_inherited;
	    HANDLE this_process = GetCurrentProcess();
	    if (DuplicateHandle(this_process, (HANDLE) s,
				this_process, &non_inherited, 0,
				FALSE, DUPLICATE_SAME_ACCESS)) {
		sock_close(s);
		s = (SOCKET) non_inherited;
	    }
	}
    }
    return s;
}

#endif  /* UDP for __WIN32__ */

/*
 * Multi-timers
 */

static void fire_multi_timers(tcp_descriptor *desc, ErlDrvPort port,
			      ErlDrvData data)
{
    ErlDrvTime next_timeout = 0;
    if (!desc->mtd) {
        ASSERT(0);
        return;
    }
#ifdef DEBUG
    {
	ErlDrvTime chk = erl_drv_monotonic_time(ERL_DRV_MSEC);
	ASSERT(chk >= desc->mtd->when);
    }
#endif
    do {
	MultiTimerData save = *desc->mtd;

        /* We first remove the timer so that the timeout_functions has
           can call clean_multi_timers without breaking anything */
        if (desc->mtd_cache == NULL) {
            desc->mtd_cache = desc->mtd;
        } else {
            FREE(desc->mtd);
        }

        desc->mtd = save.next;
        if (desc->mtd != NULL)
            desc->mtd->prev = NULL;

	(*(save.timeout_function))(data,save.caller);

        if (desc->mtd == NULL)
	    return;

	next_timeout = desc->mtd->when - erl_drv_monotonic_time(ERL_DRV_MSEC);
    } while (next_timeout <= 0);

    driver_set_timer(port, (unsigned long) next_timeout);
}

static void clean_multi_timers(tcp_descriptor *desc, ErlDrvPort port)
{
    if (desc->mtd) {
	driver_cancel_timer(port);
    }
    while (desc->mtd) {
	MultiTimerData *p = desc->mtd;
	desc->mtd = p->next;
        FREE(p);
    }
    desc->mtd = NULL;
    if (desc->mtd_cache) {
        FREE(desc->mtd_cache);
        desc->mtd_cache = NULL;
    }
}
static void remove_multi_timer(tcp_descriptor *desc, ErlDrvPort port, MultiTimerData *p)
{
    if (p->prev != NULL) {
	p->prev->next = p->next;
    } else {
	driver_cancel_timer(port);
	desc->mtd = p->next;
	if (desc->mtd) {
	    ErlDrvTime ntmo = desc->mtd->when - erl_drv_monotonic_time(ERL_DRV_MSEC);
	    if (ntmo < 0)
		ntmo = 0;
	    driver_set_timer(port, (unsigned long) ntmo);
	}
    }
    if (p->next != NULL) {
	p->next->prev = p->prev;
    }
    if (desc->mtd_cache == NULL)
        desc->mtd_cache = p;
    else
        FREE(p);
}

/* Cancel a timer based on the timeout_fun */
static void cancel_multi_timer(tcp_descriptor *desc, ErlDrvPort port,
                               void (*timeout_fun)(ErlDrvData drv_data,
                                                   ErlDrvTermData caller))
{
    MultiTimerData *timer = desc->mtd;
    while(timer && timer->timeout_function != timeout_fun) {
        timer = timer->next;
    }
    if (timer) {
        remove_multi_timer(desc, port, timer);
    }
}

static MultiTimerData *add_multi_timer(tcp_descriptor *desc, ErlDrvPort port,
				       ErlDrvTermData caller, unsigned timeout,
				       void (*timeout_fun)(ErlDrvData drv_data,
							   ErlDrvTermData caller))
{
    MultiTimerData *mtd, *p, *s;

    /* Use cached timer if available */
    if (desc->mtd_cache != NULL) {
        mtd = desc->mtd_cache;
        desc->mtd_cache = NULL;
    } else
        mtd = ALLOC(sizeof(MultiTimerData));

    if (timeout)
        mtd->when = erl_drv_monotonic_time(ERL_DRV_MSEC) + ((ErlDrvTime) timeout);
    else
        mtd->when = INT64_MIN; /* Don't have to get the time for 0 msec timeouts */

    mtd->timeout_function = timeout_fun;
    mtd->caller = caller;
    mtd->next = mtd->prev = NULL;

    /* Find correct slot in timer linked list */
    for(p = desc->mtd,s = NULL; p != NULL; s = p, p = p->next) {
	if (p->when >= mtd->when) {
	    break;
	}
    }

    /* Insert in linked list */
    if (!p) {
	if (!s) {
	    desc->mtd = mtd;
	} else {
	    s->next = mtd;
	    mtd->prev = s;
	}
    } else {
	if (!s) {
	    desc->mtd = mtd;
	} else {
	    s->next = mtd;
	    mtd->prev = s;
	}
	mtd->next = p;
	p->prev = mtd;
    }
    /* Possibly set new timer */
    if (!s) {
	driver_set_timer(port,timeout);
    }
    return mtd;
}

/*-----------------------------------------------------------------------------

   Subscription

-----------------------------------------------------------------------------*/

static int
save_subscriber(subs, subs_pid)
subs_list *subs; ErlDrvTermData subs_pid;
{
  subs_list *tmp;

  if(NO_SUBSCRIBERS(subs)) {
    subs->subscriber = subs_pid;
    subs->next = NULL;
  }
  else {
    tmp = subs->next;
    subs->next = ALLOC(sizeof(subs_list));
    if(subs->next == NULL) {
      subs->next = tmp;
      return 0;
    }
    subs->next->subscriber = subs_pid;
    subs->next->next = tmp;
  }
  return 1;
}

static void
free_subscribers(subs)
subs_list *subs;
{
  subs_list *this;
  subs_list *next;

  this = subs->next;
  while(this) {
    next = this->next;
    FREE((void *) this);
    this = next;
  }

  subs->subscriber = NO_PROCESS;
  subs->next = NULL;
}

static void send_to_subscribers
(
    ErlDrvTermData port,
    subs_list	   *subs,
    int		   free_subs,
    ErlDrvTermData msg[],
    int msg_len
)
{
  subs_list *this;
  subs_list *next;
  int first = 1;

  if(NO_SUBSCRIBERS(subs))
    return;

  this = subs;
  while(this) {

    (void) erl_drv_send_term(port, this->subscriber, msg, msg_len);

    if(free_subs && !first) {
      next = this->next;
      FREE((void *) this);
      this = next;
    }
    else
      this = this->next;
    first = 0;
  }

  if(free_subs) {
    subs->subscriber = NO_PROCESS;
    subs->next = NULL;
  }

}

/*
 * A *very* limited socket interface. Used by the memory tracer
 * (erl_mtrace.c).
 */
#include "erl_sock.h"

erts_sock_t erts_sock_open(void)
{
    SOCKET s;

    if(!sock_init())
	return ERTS_SOCK_INVALID_SOCKET;

    s = sock_open(AF_INET, SOCK_STREAM, 0);

    if (s == INVALID_SOCKET)
	return ERTS_SOCK_INVALID_SOCKET;

    return (erts_sock_t) s;
}

void erts_sock_close(erts_sock_t socket)
{
    if (socket != ERTS_SOCK_INVALID_SOCKET)
	sock_close((SOCKET) socket);
}


int erts_sock_connect(erts_sock_t socket, byte *ip_addr, int len, Uint16 port)
{
    SOCKET s = (SOCKET) socket;
    char buf[2 + 4], *p;
    ErlDrvSizeT blen = 6;
    inet_address addr;

    if (socket == ERTS_SOCK_INVALID_SOCKET || len != 4)
	return 0;

    put_int16(port, buf);
    memcpy((void *) (buf + 2), (void *) ip_addr, 4);

    p = buf;
    if (inet_set_address(AF_INET, &addr, &p, &blen) != NULL)
	return 0;

    if (IS_SOCKET_ERROR
	(sock_connect(s, (struct sockaddr *) &addr, blen)))
	return 0;
    return 1;
}

Sint erts_sock_send(erts_sock_t socket, const void *buf, Sint len)
{
    Sint result = (Sint) sock_send((SOCKET) socket, buf, (size_t) len, 0);
    if (IS_SOCKET_ERROR(result))
	return SOCKET_ERROR;
    return result;
}


int erts_sock_gethostname(char *buf, int bufsz)
{
    if (IS_SOCKET_ERROR(sock_hostname(buf, bufsz)))
	return SOCKET_ERROR;
    return 0;
}


int erts_sock_errno()
{
    return sock_errno();
}