viosocket.c

/*
   Copyright (c) 2001, 2015, Oracle and/or its affiliates. All rights reserved.

   This program is free software; you can redistribute it and/or
   modify it under the terms of the GNU General Public License
   as published by the Free Software Foundation; version 2 of
   the License.

   This program is distributed in the hope that it will be useful,
   but WITHOUT ANY WARRANTY; without even the implied warranty of
   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
   GNU General Public License for more details.

   You should have received a copy of the GNU General Public License
   along with this program; if not, write to the Free Software
   Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA
   02110-1301  USA */

/*
  Note that we can't have assertion on file descriptors;  The reason for
  this is that during mysql shutdown, another thread can close a file
  we are working on.  In this case we should just return read errors from
  the file descriptior.
*/

#include "vio_priv.h"

#ifdef FIONREAD_IN_SYS_FILIO
# include <sys/filio.h>
#endif

int vio_errno(Vio *vio __attribute__((unused)))
{
  /* These transport types are not Winsock based. */
#ifdef _WIN32
  if (vio->type == VIO_TYPE_NAMEDPIPE ||
      vio->type == VIO_TYPE_SHARED_MEMORY)
    return GetLastError();
#endif

  /* Mapped to WSAGetLastError() on Win32. */
  return socket_errno;
}


/**
  Attempt to wait for an I/O event on a socket.

  @param vio      VIO object representing a connected socket.
  @param event    The type of I/O event (read or write) to wait for.

  @return Return value is -1 on failure, 0 on success.
*/

int vio_socket_io_wait(Vio *vio, enum enum_vio_io_event event)
{
  timeout_t timeout;
  int ret;
  DBUG_ENTER(__func__);

  DBUG_ASSERT(event == VIO_IO_EVENT_READ || event == VIO_IO_EVENT_WRITE);

  /* Choose an appropriate timeout. */
  if (event == VIO_IO_EVENT_READ)
    timeout= vio->read_timeout;
  else
    timeout= vio->write_timeout;

  /* Wait for input data to become available. */
  switch (vio_io_wait(vio, event, timeout))
  {
  case -1:
    /* Upon failure, vio_read/write() shall return -1. */
    ret= -1;
    break;
  case  0:
    /* The wait timed out. */
    ret= -1;
    break;
  default:
    /* A positive value indicates an I/O event. */
    ret= 0;
    break;
  }

  DBUG_PRINT("exit", ("ret: %d", ret));
  DBUG_RETURN(ret);
}


/*
  Define a stub MSG_DONTWAIT if unavailable. In this case, fcntl
  (or a equivalent) is used to enable non-blocking operations.
  The flag must be supported in both send and recv operations.
*/
#if defined(__linux__)
#define VIO_USE_DONTWAIT  1
#define VIO_DONTWAIT      MSG_DONTWAIT
#else
#define VIO_DONTWAIT 0
#endif


size_t vio_read(Vio *vio, uchar *buf, size_t size)
{
  ssize_t ret;
  int flags= 0;
  DBUG_ENTER("vio_read");

  /* Ensure nobody uses vio_read_buff and vio_read simultaneously. */
  DBUG_ASSERT(vio->read_end == vio->read_pos);

  /* If timeout is enabled, do not block if data is unavailable. */
  if (timeout_is_nonzero(vio->read_timeout))
    flags= VIO_DONTWAIT;

  while ((ret= mysql_socket_recv(vio->mysql_socket, (SOCKBUF_T *)buf, size, flags)) == -1)
  {
    int error= socket_errno;

    /* Error encountered that is unrelated to blocking; percolate it up. */
    if (error != SOCKET_EAGAIN && error != SOCKET_EWOULDBLOCK)
      break;

    /* non-blocking with either EAGAIN or EWOULDBLOCK -- don't call
     * io_wait. 0 bytes are available. */
    DBUG_ASSERT(error == SOCKET_EAGAIN || error == SOCKET_EWOULDBLOCK);
    if (!vio_is_blocking(vio)) {
      DBUG_PRINT("info", ("vio_read on nonblocking socket read no bytes"));
      DBUG_RETURN(-1);
    }

    /* Wait for input data to become available. */
    if ((ret= vio_socket_io_wait(vio, VIO_IO_EVENT_READ))) {
      break;
    }
  }

  DBUG_PRINT("exit", ("ret: %ld", ret));
  DBUG_RETURN(ret);
}


/*
  Buffered read: if average read size is small it may
  reduce number of syscalls.
*/

size_t vio_read_buff(Vio *vio, uchar* buf, size_t size)
{
  size_t rc;
#define VIO_UNBUFFERED_READ_MIN_SIZE 2048
  DBUG_ENTER("vio_read_buff");
  DBUG_PRINT("enter", ("sd: %d  buf: 0x%lx  size: %u",
             mysql_socket_getfd(vio->mysql_socket), (long)buf, (uint)size));

  if (vio->read_pos < vio->read_end)
  {
    rc= MY_MIN((size_t) (vio->read_end - vio->read_pos), size);
#ifdef DEBUG_DATA_PACKETS
    DBUG_DUMP("read bytes", vio->read_pos, rc);
#endif
    memcpy(buf, vio->read_pos, rc);
    DBUG_PRINT("buffer read", ("copied %lu bytes", rc));
    vio->read_pos+= rc;
    /*
      Do not try to read from the socket now even if rc < size:
      vio_read can return -1 due to an error or non-blocking mode, and
      the safest way to handle it is to move to a separate branch.
    */
  }
  else if (size < VIO_UNBUFFERED_READ_MIN_SIZE)
  {
    rc= vio_read(vio, (uchar*) vio->read_buffer, VIO_READ_BUFFER_SIZE);
    DBUG_PRINT("read result", ("req: %lu, got: %lu", size, rc));
    if (rc != 0 && rc != (size_t) -1)
    {
#ifdef DEBUG_DATA_PACKETS
      DBUG_DUMP("read bytes", vio->read_buffer, rc);
#endif
      if (rc > size)
      {
        vio->read_pos= vio->read_buffer + size;
        vio->read_end= vio->read_buffer + rc;
        rc= size;
      }
      memcpy(buf, vio->read_buffer, rc);
    }
  }
  else
  {
    rc= vio_read(vio, buf, size);
    DBUG_PRINT("read result", ("req: %lu, got: %lu", size, rc));
  }
  DBUG_RETURN(rc);
#undef VIO_UNBUFFERED_READ_MIN_SIZE
}


my_bool vio_buff_has_data(Vio *vio)
{
  return (vio->read_pos != vio->read_end);
}


size_t vio_write(Vio *vio, const uchar* buf, size_t size)
{
  ssize_t ret;
  int flags= 0;
  DBUG_ENTER("vio_write");

  /* If timeout is enabled, do not block. */
  if (timeout_is_nonzero(vio->write_timeout))
    flags= VIO_DONTWAIT;

  while ((ret= mysql_socket_send(vio->mysql_socket, (SOCKBUF_T *)buf, size, flags)) == -1)
  {
    int error= socket_errno;

    /* The operation would block? */
    if (error != SOCKET_EAGAIN && error != SOCKET_EWOULDBLOCK)
      break;

    /* Wait for the output buffer to become writable.*/
    if ((ret= vio_socket_io_wait(vio, VIO_IO_EVENT_WRITE)))
      break;
  }

  DBUG_RETURN(ret);
}

//WL#4896: Not covered
int vio_set_blocking(Vio *vio, my_bool status)
{
  DBUG_ENTER("vio_set_blocking");
  DBUG_PRINT("info", ("blocking: fd %d, %d -> %d",
                      mysql_socket_getfd(vio->mysql_socket),
                      vio->is_blocking_flag, status));

#ifdef _WIN32
  DBUG_ASSERT(vio->type != VIO_TYPE_NAMEDPIPE);
  DBUG_ASSERT(vio->type != VIO_TYPE_SHARED_MEMORY);
  {
    int ret;
    u_long arg= status ? 0 : 1;
    ret= ioctlsocket(mysql_socket_getfd(vio->mysql_socket), FIONBIO, &arg);
    DBUG_RETURN(ret);
  }
#else
  {
    int flags;

    if ((flags= fcntl(mysql_socket_getfd(vio->mysql_socket), F_GETFL, NULL)) < 0)
      DBUG_RETURN(-1);

    /*
      Always set/clear the flag to avoid inheritance issues. This is
      a issue mainly on Mac OS X Tiger (version 10.4) where although
      the O_NONBLOCK flag is inherited from the parent socket, the
      actual non-blocking behavior is not inherited.
    */
    if (status)
      flags&= ~O_NONBLOCK;
    else
      flags|= O_NONBLOCK;

    if (fcntl(mysql_socket_getfd(vio->mysql_socket), F_SETFL, flags) == -1)
      DBUG_RETURN(-1);

    vio->is_blocking_flag = status;
  }
#endif

  DBUG_RETURN(0);
}


int vio_socket_timeout(Vio *vio,
                       uint which __attribute__((unused)),
                       my_bool old_mode __attribute__((unused)))
{
  int ret= 0;
  DBUG_ENTER("vio_socket_timeout");

#if defined(_WIN32)
  {
    int optname;
    DWORD timeout= 0;
    const char *optval= (const char *) &timeout;

    /*
      The default socket timeout value is zero, which means an infinite
      timeout. Values less than 500 milliseconds are interpreted to be of
      500 milliseconds. Hence, the VIO behavior for zero timeout, which is
      intended to cause the send or receive operation to fail immediately
      if no data is available, is not supported on WIN32 and neither is
      necessary as it's not possible to set the VIO timeout value to zero.

      Assert that the VIO timeout is either positive or set to infinite.
    */
    DBUG_ASSERT(which || vio->read_timeout);
    DBUG_ASSERT(!which || vio->write_timeout);

    if (which)
    {
      optname= SO_SNDTIMEO;
      if (vio->write_timeout > 0)
        timeout= vio->write_timeout;
    }
    else
    {
      optname= SO_RCVTIMEO;
      if (vio->read_timeout > 0)
        timeout= vio->read_timeout;
    }

    ret= mysql_socket_setsockopt(vio->mysql_socket, SOL_SOCKET, optname,
	                             optval, sizeof(timeout));
  }
#else
  /*
    The MSG_DONTWAIT trick is not used with SSL sockets as the send and
    receive I/O operations are wrapped through SSL-specific functions
    (SSL_read and SSL_write) which are not equivalent to the standard
    recv(2) and send(2) used in vio_read() and vio_write(). Hence, the
    socket blocking mode is changed and vio_io_wait() is used to wait
    for I/O or timeout.
  */
#ifdef VIO_USE_DONTWAIT
  if (vio->type == VIO_TYPE_SSL)
#endif
  {
    /* Deduce what should be the new blocking mode of the socket. */
    my_bool new_mode =
      timeout_is_infinite(vio->write_timeout) &&
      timeout_is_infinite(vio->read_timeout);

    /* If necessary, update the blocking mode. */
    if (new_mode != old_mode)
      ret= vio_set_blocking(vio, new_mode);
  }
#endif

  DBUG_RETURN(ret);
}

my_bool
vio_is_blocking(Vio * vio)
{
  DBUG_ENTER(__func__);
  DBUG_RETURN(vio->is_blocking_flag);
}

int vio_fastsend(Vio * vio __attribute__((unused)))
{
  int r=0;
  DBUG_ENTER("vio_fastsend");

#if defined(IPTOS_THROUGHPUT)
  {
    int tos = IPTOS_THROUGHPUT;
    r= mysql_socket_setsockopt(vio->mysql_socket, IPPROTO_IP, IP_TOS,
	                           (void *)&tos, sizeof(tos));
  }
#endif                                    /* IPTOS_THROUGHPUT */
  if (!r)
  {
#ifdef __WIN__
    BOOL nodelay= 1;
#else
    int nodelay = 1;
#endif

    r= mysql_socket_setsockopt(vio->mysql_socket, IPPROTO_TCP, TCP_NODELAY,
                  IF_WIN((const char*), (void*)) &nodelay,
                  sizeof(nodelay));

  }
  if (r)
  {
    DBUG_PRINT("warning", ("Couldn't set socket option for fast send"));
    r= -1;
  }
  DBUG_PRINT("exit", ("%d", r));
  DBUG_RETURN(r);
}

int vio_keepalive(Vio* vio, my_bool set_keep_alive)
{
  int r=0;
  uint opt = 0;
  DBUG_ENTER("vio_keepalive");
  DBUG_PRINT("enter", ("sd: %d  set_keep_alive: %d",
             mysql_socket_getfd(vio->mysql_socket), (int)set_keep_alive));
  if (vio->type != VIO_TYPE_NAMEDPIPE)
  {
    if (set_keep_alive)
      opt = 1;
    r = mysql_socket_setsockopt(vio->mysql_socket, SOL_SOCKET, SO_KEEPALIVE,
	                            (char *)&opt, sizeof(opt));
  }
  DBUG_RETURN(r);
}


/**
  Indicate whether a I/O operation must be retried later.

  @param vio  A VIO object

  @return Whether a I/O operation should be deferred.
  @retval TRUE    Temporary failure, retry operation.
  @retval FALSE   Indeterminate failure.
*/

my_bool
vio_should_retry(Vio *vio)
{
  return (vio_errno(vio) == SOCKET_EINTR);
}


/**
  Indicate whether a I/O operation timed out.

  @param vio  A VIO object

  @return Whether a I/O operation timed out.
  @retval TRUE    Operation timed out.
  @retval FALSE   Not a timeout failure.
*/

my_bool
vio_was_timeout(Vio *vio)
{
  return (vio_errno(vio) == SOCKET_ETIMEDOUT);
}


int vio_shutdown(Vio * vio)
{
  int r=0;
  DBUG_ENTER("vio_shutdown");

 if (vio->inactive == FALSE)
  {
    DBUG_ASSERT(vio->type ==  VIO_TYPE_TCPIP ||
      vio->type == VIO_TYPE_SOCKET ||
      vio->type == VIO_TYPE_SSL);

    DBUG_ASSERT(mysql_socket_getfd(vio->mysql_socket) >= 0);
    if (mysql_socket_shutdown(vio->mysql_socket, SHUT_RDWR))
      r= -1;
    if (mysql_socket_close(vio->mysql_socket))
      r= -1;
  }
  if (r)
  {
    DBUG_PRINT("vio_error", ("close() failed, error: %d",socket_errno));
    /* FIXME: error handling (not critical for MySQL) */
  }
  vio->inactive= TRUE;
  vio->mysql_socket= MYSQL_INVALID_SOCKET;
  DBUG_RETURN(r);
}


const char *vio_description(Vio * vio)
{
  if (!vio->desc[0])
  {
    my_snprintf(vio->desc, VIO_DESCRIPTION_SIZE,
                (vio->type == VIO_TYPE_SOCKET ? "socket (%d)" : "TCP/IP (%d)"),
                mysql_socket_getfd(vio->mysql_socket));
  }
  return vio->desc;
}

enum enum_vio_type vio_type(Vio* vio)
{
  return vio->type;
}

my_socket vio_fd(Vio* vio)
{
  return mysql_socket_getfd(vio->mysql_socket);
}

/**
  Convert a sock-address (AF_INET or AF_INET6) into the "normalized" form,
  which is the IPv4 form for IPv4-mapped or IPv4-compatible IPv6 addresses.

  @note Background: when IPv4 and IPv6 are used simultaneously, IPv4
  addresses may be written in a form of IPv4-mapped or IPv4-compatible IPv6
  addresses. That means, one address (a.b.c.d) can be written in three forms:
    - IPv4: a.b.c.d;
    - IPv4-compatible IPv6: ::a.b.c.d;
    - IPv4-mapped IPv4: ::ffff:a.b.c.d;

  Having three forms of one address makes it a little difficult to compare
  addresses with each other (the IPv4-compatible IPv6-address of foo.bar
  will be different from the IPv4-mapped IPv6-address of foo.bar).

  @note This function can be made public when it's needed.

  @param src        [in] source IP address (AF_INET or AF_INET6).
  @param src_length [in] length of the src.
  @param dst        [out] a buffer to store normalized IP address
                          (sockaddr_storage).
  @param dst_length [out] actual length of the normalized IP address.
*/
static void vio_get_normalized_ip(const struct sockaddr *src,
                                  int src_length,
                                  struct sockaddr *dst,
                                  int *dst_length)
{
  switch (src->sa_family) {
  case AF_INET:
    memcpy(dst, src, src_length);
    *dst_length= src_length;
    break;

#ifdef HAVE_IPV6
  case AF_INET6:
  {
    const struct sockaddr_in6 *src_addr6= (const struct sockaddr_in6 *) src;
    const struct in6_addr *src_ip6= &(src_addr6->sin6_addr);
    const uint32 *src_ip6_int32= (uint32 *) src_ip6->s6_addr;

    if (IN6_IS_ADDR_V4MAPPED(src_ip6) || IN6_IS_ADDR_V4COMPAT(src_ip6))
    {
      struct sockaddr_in *dst_ip4= (struct sockaddr_in *) dst;

      /*
        This is an IPv4-mapped or IPv4-compatible IPv6 address. It should
        be converted to the IPv4 form.
      */

      *dst_length= sizeof (struct sockaddr_in);

      memset(dst_ip4, 0, *dst_length);
      dst_ip4->sin_family= AF_INET;
      dst_ip4->sin_port= src_addr6->sin6_port;

      /*
        In an IPv4 mapped or compatible address, the last 32 bits represent
        the IPv4 address. The byte orders for IPv6 and IPv4 addresses are
        the same, so a simple copy is possible.
      */
      dst_ip4->sin_addr.s_addr= src_ip6_int32[3];
    }
    else
    {
      /* This is a "native" IPv6 address. */

      memcpy(dst, src, src_length);
      *dst_length= src_length;
    }

    break;
  }
#endif /* HAVE_IPV6 */
  }
}


/**
  Return the normalized IP address string for a sock-address.

  The idea is to return an IPv4-address for an IPv4-mapped and
  IPv4-compatible IPv6 address.

  The function writes the normalized IP address to the given buffer.
  The buffer should have enough space, otherwise error flag is returned.
  The system constant INET6_ADDRSTRLEN can be used to reserve buffers of
  the right size.

  @param addr           [in]  sockaddr object (AF_INET or AF_INET6).
  @param addr_length    [in]  length of the addr.
  @param ip_string      [out] buffer to write normalized IP address.
  @param ip_string_size [in]  size of the ip_string.

  @return Error status.
  @retval TRUE in case of error (the ip_string buffer is not enough).
  @retval FALSE on success.
*/

my_bool vio_get_normalized_ip_string(const struct sockaddr *addr,
                                     int addr_length,
                                     char *ip_string,
                                     size_t ip_string_size)
{
  struct sockaddr_storage norm_addr_storage;
  struct sockaddr *norm_addr= (struct sockaddr *) &norm_addr_storage;
  int norm_addr_length;
  int err_code;

  vio_get_normalized_ip(addr, addr_length, norm_addr, &norm_addr_length);

  err_code= vio_getnameinfo(norm_addr, ip_string, ip_string_size, NULL, 0,
                            NI_NUMERICHOST);

  if (!err_code)
    return FALSE;

  DBUG_PRINT("error", ("getnameinfo() failed with %d (%s).",
                       (int) err_code,
                       (const char *) gai_strerror(err_code)));
  return TRUE;
}


/**
  Return IP address and port of a VIO client socket.

  The function returns an IPv4 address if IPv6 support is disabled.

  The function returns an IPv4 address if the client socket is associated
  with an IPv4-compatible or IPv4-mapped IPv6 address. Otherwise, the native
  IPv6 address is returned.
*/

my_bool vio_peer_addr(Vio *vio, char *ip_buffer, uint16 *port,
                      size_t ip_buffer_size)
{
  DBUG_ENTER("vio_peer_addr");
  DBUG_PRINT("enter", ("Client socked fd: %d",
            (int)mysql_socket_getfd(vio->mysql_socket)));

  if (vio->localhost)
  {
    /*
      Initialize vio->remote and vio->addLen. Set vio->remote to IPv4 loopback
      address.
    */
    struct in_addr *ip4= &((struct sockaddr_in *) &(vio->remote))->sin_addr;

    vio->remote.ss_family= AF_INET;
    vio->addrLen= sizeof (struct sockaddr_in);

    ip4->s_addr= htonl(INADDR_LOOPBACK);

    /* Initialize ip_buffer and port. */

    strmov(ip_buffer, "127.0.0.1");
    *port= 0;
  }
  else
  {
    int err_code;
    char port_buffer[NI_MAXSERV];

    struct sockaddr_storage addr_storage;
    struct sockaddr *addr= (struct sockaddr *) &addr_storage;
    size_socket addr_length= sizeof (addr_storage);

    /* Get sockaddr by socked fd. */

    err_code= mysql_socket_getpeername(vio->mysql_socket, addr, &addr_length);

    if (err_code)
    {
      DBUG_PRINT("exit", ("getpeername() gave error: %d", socket_errno));
      DBUG_RETURN(TRUE);
    }

    /* Normalize IP address. */

    vio_get_normalized_ip(addr, addr_length,
                          (struct sockaddr *) &vio->remote, &vio->addrLen);

    /* Get IP address & port number. */

    err_code= vio_getnameinfo((struct sockaddr *) &vio->remote,
                              ip_buffer, ip_buffer_size,
                              port_buffer, NI_MAXSERV,
                              NI_NUMERICHOST | NI_NUMERICSERV);

    if (err_code)
    {
      DBUG_PRINT("exit", ("getnameinfo() gave error: %s",
                          gai_strerror(err_code)));
      DBUG_RETURN(TRUE);
    }

    *port= (uint16) strtol(port_buffer, NULL, 10);
  }

  DBUG_PRINT("exit", ("Client IP address: %s; port: %d",
                      (const char *) ip_buffer,
                      (int) *port));
  DBUG_RETURN(FALSE);
}


/**
  Retrieve the amount of data that can be read from a socket.

  @param vio          A VIO object.
  @param bytes[out]   The amount of bytes available.

  @retval FALSE   Success.
  @retval TRUE    Failure.
*/
// WL#4896: Not covered
static my_bool socket_peek_read(Vio *vio, uint *bytes)
{
  my_socket sd= mysql_socket_getfd(vio->mysql_socket);
#if defined(_WIN32)
  int len;
  if (ioctlsocket(sd, FIONREAD, &len))
    return TRUE;
  *bytes= len;
  return FALSE;
#elif defined(FIONREAD_IN_SYS_IOCTL) || defined(FIONREAD_IN_SYS_FILIO)
  int len;
  if (ioctl(sd, FIONREAD, &len) < 0)
    return TRUE;
  *bytes= len;
  return FALSE;
#else
  char buf[1024];
  ssize_t res= recv(sd, &buf, sizeof(buf), MSG_PEEK);
  if (res < 0)
    return TRUE;
  *bytes= res;
  return FALSE;
#endif
}

#ifndef _WIN32

/**
  Set of event flags grouped by operations.
*/

/*
  Linux specific flag used to detect connection shutdown. The flag is
  also used for half-closed notification, which here is interpreted as
  if there is data available to be read from the socket.
*/
#ifndef POLLRDHUP
#define POLLRDHUP 0
#endif

/* Data may be read. */
#define MY_POLL_SET_IN      (POLLIN | POLLPRI)
/* Data may be written. */
#define MY_POLL_SET_OUT     (POLLOUT)
/* An error or hangup. */
#define MY_POLL_SET_ERR     (POLLERR | POLLHUP | POLLNVAL)

#endif

/**
  Wait for an I/O event on a VIO socket.

  @param vio      VIO object representing a connected socket.
  @param event    The type of I/O event to wait for.
  @param timeout  Interval (in milliseconds) to wait for an I/O event.
                  A negative timeout value means an infinite timeout.

  @remark sock_errno is set to SOCKET_ETIMEDOUT on timeout.

  @return A three-state value which indicates the operation status.
  @retval -1  Failure, socket_errno indicates the error.
  @retval  0  The wait has timed out.
  @retval  1  The requested I/O event has occurred.
*/

#if !defined(_WIN32) && !defined(__APPLE__)
int vio_io_wait(Vio *vio, enum enum_vio_io_event event, timeout_t timeout)
{
  int ret;
  short revents __attribute__((unused)) = 0;
  struct pollfd pfd;
  my_socket sd= mysql_socket_getfd(vio->mysql_socket);
  MYSQL_SOCKET_WAIT_VARIABLES(locker, state) /* no ';' */
  DBUG_ENTER("vio_io_wait");

  memset(&pfd, 0, sizeof(pfd));

  pfd.fd= sd;

  /*
    Set the poll bitmask describing the type of events.
    The error flags are only valid in the revents bitmask.
  */
  switch (event)
  {
  case VIO_IO_EVENT_READ:
    pfd.events= MY_POLL_SET_IN;
    revents= MY_POLL_SET_IN | MY_POLL_SET_ERR | POLLRDHUP;
    break;
  case VIO_IO_EVENT_WRITE:
  case VIO_IO_EVENT_CONNECT:
    pfd.events= MY_POLL_SET_OUT;
    revents= MY_POLL_SET_OUT | MY_POLL_SET_ERR;
    break;
  }

  MYSQL_START_SOCKET_WAIT(locker, &state, vio->mysql_socket, PSI_SOCKET_SELECT, 0);

  /*
    Wait for the I/O event and return early in case of
    error or timeout.
  */
  switch ((ret = poll(&pfd, 1, timeout_to_millis(timeout))))
  {
  case -1:
    /* On error, -1 is returned. */
    break;
  case 0:
    /*
      Set errno to indicate a timeout error.
      (This is not compiled in on WIN32.)
    */
    errno= SOCKET_ETIMEDOUT;
    break;
  default:
    /* Ensure that the requested I/O event has completed. */
    DBUG_ASSERT(pfd.revents & revents);
    break;
  }

  MYSQL_END_SOCKET_WAIT(locker, 0);
  DBUG_RETURN(ret);
}

#else

int vio_io_wait(Vio *vio, enum enum_vio_io_event event, timeout_t timeout)
{
  int ret;
  struct timeval tm;
  my_socket fd;
  fd_set readfds, writefds, exceptfds;
  MYSQL_SOCKET_WAIT_VARIABLES(locker, state) /* no ';' */
  DBUG_ENTER("vio_io_wait");

  fd= mysql_socket_getfd(vio->mysql_socket);

  if (fd == INVALID_SOCKET)
    DBUG_RETURN(-1);

  /* Convert the timeout, in milliseconds, to seconds and microseconds. */
  if (!timeout_is_infinite(timeout))
  {
    uint timeout_ms = timeout_to_millis(timeout);
    tm.tv_sec= timeout_ms / 1000;
    tm.tv_usec= (timeout_ms % 1000) * 1000;
  }

  FD_ZERO(&readfds);
  FD_ZERO(&writefds);
  FD_ZERO(&exceptfds);

  /* Always receive notification of exceptions. */
  FD_SET(fd, &exceptfds);

  switch (event)
  {
  case VIO_IO_EVENT_READ:
    /* Readiness for reading. */
    FD_SET(fd, &readfds);
    break;
  case VIO_IO_EVENT_WRITE:
  case VIO_IO_EVENT_CONNECT:
    /* Readiness for writing. */
    FD_SET(fd, &writefds);
    break;
  }

  MYSQL_START_SOCKET_WAIT(locker, &state, vio->mysql_socket, PSI_SOCKET_SELECT, 0);

  /* The first argument is ignored on Windows. */
  ret= select(fd + 1, &readfds, &writefds, &exceptfds, 
              timeout_is_infinite(timeout) ? NULL : &tm);

  MYSQL_END_SOCKET_WAIT(locker, 0);

  /* Set error code to indicate a timeout error. */
  if (ret == 0)
#if defined(_WIN32)
    WSASetLastError(SOCKET_ETIMEDOUT);
#elif defined(__APPLE__)
    errno= SOCKET_ETIMEDOUT;
#else
#error Oops...Wrong OS
#endif

  /* Error or timeout? */
  if (ret <= 0)
    DBUG_RETURN(ret);

  /* The requested I/O event is ready? */
  switch (event)
  {
  case VIO_IO_EVENT_READ:
    ret= MY_TEST(FD_ISSET(fd, &readfds));
    break;
  case VIO_IO_EVENT_WRITE:
  case VIO_IO_EVENT_CONNECT:
    ret= MY_TEST(FD_ISSET(fd, &writefds));
    break;
  }

  /* Error conditions pending? */
  ret|= MY_TEST(FD_ISSET(fd, &exceptfds));

  /* Not a timeout, ensure that a condition was met. */
  DBUG_ASSERT(ret);

  DBUG_RETURN(ret);
}

#endif /* _WIN32 */


/**
  Connect to a peer address.

  @param vio       A VIO object.
  @param addr      Socket address containing the peer address.
  @param len       Length of socket address.
  @param timeout   Interval (in milliseconds) to wait until a
                   connection is established.

  @retval FALSE   A connection was successfully established.
  @retval TRUE    A fatal error. See socket_errno.
*/

my_bool
vio_socket_connect(Vio *vio, struct sockaddr *addr, socklen_t len,
                   my_bool nonblocking, timeout_t timeout)
{
  int ret, wait;
  DBUG_ENTER("vio_socket_connect");

  /* Only for socket-based transport types. */
  DBUG_ASSERT(vio->type == VIO_TYPE_SOCKET || vio->type == VIO_TYPE_TCPIP);

  /* If timeout is not infinite, set socket to non-blocking mode. */
  if ((!timeout_is_infinite(timeout) || nonblocking) &&
      vio_set_blocking(vio, FALSE))
    DBUG_RETURN(TRUE);

  /* Initiate the connection. */
  ret= mysql_socket_connect(vio->mysql_socket, addr, len);

#ifdef _WIN32
  wait= (ret == SOCKET_ERROR) &&
        (WSAGetLastError() == WSAEINPROGRESS ||
         WSAGetLastError() == WSAEWOULDBLOCK);
#else
  wait= (ret == -1) && (errno == EINPROGRESS || errno == EALREADY);
#endif

  /*
    The connection is in progress. The vio_io_wait() call can be used
    to wait up to a specified period of time for the connection to
    succeed.

    If vio_io_wait() returns 0 (after waiting however many seconds),
    the socket never became writable (host is probably unreachable.)
    Otherwise, if vio_io_wait() returns 1, then one of two conditions
    exist:

    1. An error occurred. Use getsockopt() to check for this.
    2. The connection was set up successfully: getsockopt() will
       return 0 as an error.
  */
  if (!nonblocking && wait && (vio_io_wait(vio, VIO_IO_EVENT_CONNECT, timeout) == 1))
  {
    int error;
    IF_WIN(int, socklen_t) optlen= sizeof(error);
    IF_WIN(char, void) *optval= (IF_WIN(char, void) *) &error;

    /*
      At this point, we know that something happened on the socket.
      But this does not means that everything is alright. The connect
      might have failed. We need to retrieve the error code from the
      socket layer. We must return success only if we are sure that
      it was really a success. Otherwise we might prevent the caller
      from trying another address to connect to.
    */
    if (!(ret= mysql_socket_getsockopt(vio->mysql_socket, SOL_SOCKET, SO_ERROR, optval, &optlen)))
    {
#ifdef _WIN32
      WSASetLastError(error);
#else
      errno= error;
#endif
      ret= MY_TEST(error);
    }
  }

  /* If necessary, restore the blocking mode, but only if connect succeeded. */
  if ((!nonblocking && timeout_is_nonzero(timeout)) && (ret == 0))
  {
    if (vio_set_blocking(vio, TRUE))
      DBUG_RETURN(TRUE);
  }

  if (nonblocking && wait) {
    DBUG_RETURN(FALSE);
  } else {
    DBUG_RETURN(MY_TEST(ret));
  }
}


/**
  Determine if the endpoint of a connection is still available.

  @remark The socket is assumed to be disconnected if an EOF
          condition is encountered.

  @param vio      The VIO object.

  @retval TRUE    EOF condition not found.
  @retval FALSE   EOF condition is signaled.
*/

my_bool vio_is_connected(Vio *vio)
{
  uint bytes= 0;
  DBUG_ENTER("vio_is_connected");

  /*
    The first step of detecting an EOF condition is verifying
    whether there is data to read. Data in this case would be
    the EOF. An exceptional condition event and/or errors are
    interpreted as if there is data to read.
  */
  if (!vio_io_wait(vio, VIO_IO_EVENT_READ, timeout_from_seconds(0)))
    DBUG_RETURN(TRUE);

  /*
    The second step is read() or recv() from the socket returning
    0 (EOF). Unfortunately, it's not possible to call read directly
    as we could inadvertently read meaningful connection data.
    Simulate a read by retrieving the number of bytes available to
    read -- 0 meaning EOF. In the presence of unrecoverable errors,
    the socket is assumed to be disconnected.
  */
  while (socket_peek_read(vio, &bytes))
  {
    if (socket_errno != SOCKET_EINTR)
      DBUG_RETURN(FALSE);
  }

#ifdef HAVE_OPENSSL
  /* There might be buffered data at the SSL layer. */
  if (!bytes && vio->type == VIO_TYPE_SSL)
    bytes= SSL_pending((SSL*) vio->ssl_arg);
#endif

  DBUG_RETURN(bytes ? TRUE : FALSE);
}

#ifndef DBUG_OFF

/**
  Number of bytes in the read or socket buffer

  @remark An EOF condition might count as one readable byte.

  @return number of bytes in one of the buffers or < 0 if error.
*/

ssize_t vio_pending(Vio *vio)
{
  DBUG_ENTER(__func__);
  uint bytes= 0;

  /* Data pending on the read buffer. */
  if (vio->read_pos < vio->read_end)
    DBUG_RETURN(vio->read_end - vio->read_pos);

  /* Skip non-socket based transport types. */
  if (vio->type == VIO_TYPE_TCPIP || vio->type == VIO_TYPE_SOCKET)
  {
    /* Obtain number of readable bytes in the socket buffer. */
    if (socket_peek_read(vio, &bytes))
      DBUG_RETURN(-1);
  }

  /*
    SSL not checked due to a yaSSL bug in SSL_pending that
    causes it to attempt to read from the socket.
  */

  DBUG_PRINT("bytes left", ("%u", bytes));
  DBUG_RETURN((ssize_t) bytes);
}

#endif

/**
  Checks if the error code, returned by vio_getnameinfo(), means it was the
  "No-name" error.

  Windows-specific note: getnameinfo() returns WSANO_DATA instead of
  EAI_NODATA or EAI_NONAME when no reverse mapping is available at the host
  (i.e. Windows can't get hostname by IP-address). This error should be
  treated as EAI_NONAME.

  @return if the error code is actually EAI_NONAME.
  @retval true if the error code is EAI_NONAME.
  @retval false otherwise.
*/

my_bool vio_is_no_name_error(int err_code)
{
#ifdef _WIN32

  return err_code == WSANO_DATA || err_code == EAI_NONAME;

#else

  return err_code == EAI_NONAME;

#endif
}


/**
  This is a wrapper for the system getnameinfo(), because different OS
  differ in the getnameinfo() implementation:
    - Solaris 10 requires that the 2nd argument (salen) must match the
      actual size of the struct sockaddr_storage passed to it;
    - Mac OS X has sockaddr_in::sin_len and sockaddr_in6::sin6_len and
      requires them to be filled.
*/

int vio_getnameinfo(const struct sockaddr *sa,
                    char *hostname, size_t hostname_size,
                    char *port, size_t port_size,
                    int flags)
{
  int sa_length= 0;

  switch (sa->sa_family) {
  case AF_INET:
    sa_length= sizeof (struct sockaddr_in);
#ifdef HAVE_SOCKADDR_IN_SIN_LEN
    ((struct sockaddr_in *) sa)->sin_len= sa_length;
#endif /* HAVE_SOCKADDR_IN_SIN_LEN */
    break;

#ifdef HAVE_IPV6
  case AF_INET6:
    sa_length= sizeof (struct sockaddr_in6);
# ifdef HAVE_SOCKADDR_IN6_SIN6_LEN
    ((struct sockaddr_in6 *) sa)->sin6_len= sa_length;
# endif /* HAVE_SOCKADDR_IN6_SIN6_LEN */
    break;
#endif /* HAVE_IPV6 */
  }

  return getnameinfo(sa, sa_length,
                     hostname, hostname_size,
                     port, port_size,
                     flags);
}