connect.c

一个从网络上自动下载文件的自由工具

/* Establishing and handling network connections.
   Copyright (C) 1995, 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003,
   2004, 2005, 2006, 2007, 2008 Free Software Foundation, Inc.

This file is part of GNU Wget.

GNU Wget 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; either version 3 of the License, or
 (at your option) any later version.

GNU Wget 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 Wget.  If not, see <http://www.gnu.org/licenses/>.

Additional permission under GNU GPL version 3 section 7

If you modify this program, or any covered work, by linking or
combining it with the OpenSSL project's OpenSSL library (or a
modified version of that library), containing parts covered by the
terms of the OpenSSL or SSLeay licenses, the Free Software Foundation
grants you additional permission to convey the resulting work.
Corresponding Source for a non-source form of such a combination
shall include the source code for the parts of OpenSSL used as well
as that of the covered work.  */

#include <config.h>

#include <stdio.h>
#include <stdlib.h>
#ifdef HAVE_UNISTD_H
# include <unistd.h>
#endif
#include <assert.h>

#ifndef WINDOWS
# include <sys/socket.h>
# include <netdb.h>
# include <netinet/in.h>
# ifndef __BEOS__
#  include <arpa/inet.h>
# endif
#endif /* not WINDOWS */

#include <errno.h>
#include <string.h>
#ifdef HAVE_SYS_SELECT_H
# include <sys/select.h>
#endif /* HAVE_SYS_SELECT_H */

#include "wget.h"
#include "utils.h"
#include "host.h"
#include "connect.h"
#include "hash.h"

/* Define sockaddr_storage where unavailable (presumably on IPv4-only
   hosts).  */

#ifndef ENABLE_IPV6
# ifndef HAVE_STRUCT_SOCKADDR_STORAGE
#  define sockaddr_storage sockaddr_in
# endif
#endif /* ENABLE_IPV6 */

/* Fill SA as per the data in IP and PORT.  SA shoult point to struct
   sockaddr_storage if ENABLE_IPV6 is defined, to struct sockaddr_in
   otherwise.  */

static void
sockaddr_set_data (struct sockaddr *sa, const ip_address *ip, int port)
{
  switch (ip->family)
    {
    case AF_INET:
      {
        struct sockaddr_in *sin = (struct sockaddr_in *)sa;
        xzero (*sin);
        sin->sin_family = AF_INET;
        sin->sin_port = htons (port);
        sin->sin_addr = ip->data.d4;
        break;
      }
#ifdef ENABLE_IPV6
    case AF_INET6:
      {
        struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)sa;
        xzero (*sin6);
        sin6->sin6_family = AF_INET6;
        sin6->sin6_port = htons (port);
        sin6->sin6_addr = ip->data.d6;
#ifdef HAVE_SOCKADDR_IN6_SCOPE_ID
        sin6->sin6_scope_id = ip->ipv6_scope;
#endif
        break;
      }
#endif /* ENABLE_IPV6 */
    default:
      abort ();
    }
}

/* Get the data of SA, specifically the IP address and the port.  If
   you're not interested in one or the other information, pass NULL as
   the pointer.  */

static void
sockaddr_get_data (const struct sockaddr *sa, ip_address *ip, int *port)
{
  switch (sa->sa_family)
    {
    case AF_INET:
      {
        struct sockaddr_in *sin = (struct sockaddr_in *)sa;
        if (ip)
          {
            ip->family = AF_INET;
            ip->data.d4 = sin->sin_addr;
          }
        if (port)
          *port = ntohs (sin->sin_port);
        break;
      }
#ifdef ENABLE_IPV6
    case AF_INET6:
      {
        struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)sa;
        if (ip)
          {
            ip->family = AF_INET6;
            ip->data.d6 = sin6->sin6_addr;
#ifdef HAVE_SOCKADDR_IN6_SCOPE_ID
            ip->ipv6_scope = sin6->sin6_scope_id;
#endif
          }
        if (port)
          *port = ntohs (sin6->sin6_port);
        break;
      }
#endif
    default:
      abort ();
    }
}

/* Return the size of the sockaddr structure depending on its
   family.  */

static socklen_t
sockaddr_size (const struct sockaddr *sa)
{
  switch (sa->sa_family)
    {
    case AF_INET:
      return sizeof (struct sockaddr_in);
#ifdef ENABLE_IPV6
    case AF_INET6:
      return sizeof (struct sockaddr_in6);
#endif
    default:
      abort ();
    }
}

/* Resolve the bind address specified via --bind-address and store it
   to SA.  The resolved value is stored in a static variable and
   reused after the first invocation of this function.

   Returns true on success, false on failure.  */

static bool
resolve_bind_address (struct sockaddr *sa)
{
  struct address_list *al;

  /* Make sure this is called only once.  opt.bind_address doesn't
     change during a Wget run.  */
  static bool called, should_bind;
  static ip_address ip;
  if (called)
    {
      if (should_bind)
        sockaddr_set_data (sa, &ip, 0);
      return should_bind;
    }
  called = true;

  al = lookup_host (opt.bind_address, LH_BIND | LH_SILENT);
  if (!al)
    {
      /* #### We should be able to print the error message here. */
      logprintf (LOG_NOTQUIET,
                 _("%s: unable to resolve bind address `%s'; disabling bind.\n"),
                 exec_name, opt.bind_address);
      should_bind = false;
      return false;
    }

  /* Pick the first address in the list and use it as bind address.
     Perhaps we should try multiple addresses in succession, but I
     don't think that's necessary in practice.  */
  ip = *address_list_address_at (al, 0);
  address_list_release (al);

  sockaddr_set_data (sa, &ip, 0);
  should_bind = true;
  return true;
}

struct cwt_context {
  int fd;
  const struct sockaddr *addr;
  socklen_t addrlen;
  int result;
};

static void
connect_with_timeout_callback (void *arg)
{
  struct cwt_context *ctx = (struct cwt_context *)arg;
  ctx->result = connect (ctx->fd, ctx->addr, ctx->addrlen);
}

/* Like connect, but specifies a timeout.  If connecting takes longer
   than TIMEOUT seconds, -1 is returned and errno is set to
   ETIMEDOUT.  */

static int
connect_with_timeout (int fd, const struct sockaddr *addr, socklen_t addrlen,
                      double timeout)
{
  struct cwt_context ctx;
  ctx.fd = fd;
  ctx.addr = addr;
  ctx.addrlen = addrlen;

  if (run_with_timeout (timeout, connect_with_timeout_callback, &ctx))
    {
      errno = ETIMEDOUT;
      return -1;
    }
  if (ctx.result == -1 && errno == EINTR)
    errno = ETIMEDOUT;
  return ctx.result;
}

/* Connect via TCP to the specified address and port.

   If PRINT is non-NULL, it is the host name to print that we're
   connecting to.  */

int
connect_to_ip (const ip_address *ip, int port, const char *print)
{
  struct sockaddr_storage ss;
  struct sockaddr *sa = (struct sockaddr *)&ss;
  int sock;

  /* If PRINT is non-NULL, print the "Connecting to..." line, with
     PRINT being the host name we're connecting to.  */
  if (print)
    {
      const char *txt_addr = print_address (ip);
      if (print && 0 != strcmp (print, txt_addr))
        logprintf (LOG_VERBOSE, _("Connecting to %s|%s|:%d... "),
                   escnonprint (print), txt_addr, port);
      else
        logprintf (LOG_VERBOSE, _("Connecting to %s:%d... "), txt_addr, port);
    }

  /* Store the sockaddr info to SA.  */
  sockaddr_set_data (sa, ip, port);

  /* Create the socket of the family appropriate for the address.  */
  sock = socket (sa->sa_family, SOCK_STREAM, 0);
  if (sock < 0)
    goto err;

#if defined(ENABLE_IPV6) && defined(IPV6_V6ONLY)
  if (opt.ipv6_only) {
    int on = 1;
    /* In case of error, we will go on anyway... */
    int err = setsockopt (sock, IPPROTO_IPV6, IPV6_V6ONLY, &on, sizeof (on));
    IF_DEBUG
      if (err < 0) 
        DEBUGP (("Failed setting IPV6_V6ONLY: %s", strerror (errno)));
  }
#endif

  /* For very small rate limits, set the buffer size (and hence,
     hopefully, the kernel's TCP window size) to the per-second limit.
     That way we should never have to sleep for more than 1s between
     network reads.  */
  if (opt.limit_rate && opt.limit_rate < 8192)
    {
      int bufsize = opt.limit_rate;
      if (bufsize < 512)
        bufsize = 512;          /* avoid pathologically small values */
#ifdef SO_RCVBUF
      setsockopt (sock, SOL_SOCKET, SO_RCVBUF,
                  (void *)&bufsize, (socklen_t)sizeof (bufsize));
#endif
      /* When we add limit_rate support for writing, which is useful
         for POST, we should also set SO_SNDBUF here.  */
    }

  if (opt.bind_address)
    {
      /* Bind the client side of the socket to the requested
         address.  */
      struct sockaddr_storage bind_ss;
      struct sockaddr *bind_sa = (struct sockaddr *)&bind_ss;
      if (resolve_bind_address (bind_sa))
        {
          if (bind (sock, bind_sa, sockaddr_size (bind_sa)) < 0)
            goto err;
        }
    }

  /* Connect the socket to the remote endpoint.  */
  if (connect_with_timeout (sock, sa, sockaddr_size (sa),
                            opt.connect_timeout) < 0)
    goto err;

  /* Success. */
  assert (sock >= 0);
  if (print)
    logprintf (LOG_VERBOSE, _("connected.\n"));
  DEBUGP (("Created socket %d.\n", sock));
  return sock;

 err:
  {
    /* Protect errno from possible modifications by close and
       logprintf.  */
    int save_errno = errno;
    if (sock >= 0)
      fd_close (sock);
    if (print)
      logprintf (LOG_VERBOSE, _("failed: %s.\n"), strerror (errno));
    errno = save_errno;
    return -1;
  }
}

/* Connect via TCP to a remote host on the specified port.

   HOST is resolved as an Internet host name.  If HOST resolves to
   more than one IP address, they are tried in the order returned by
   DNS until connecting to one of them succeeds.  */

int
connect_to_host (const char *host, int port)
{
  int i, start, end;
  int sock;

  struct address_list *al = lookup_host (host, 0);

 retry:
  if (!al)
    {
      logprintf (LOG_NOTQUIET,
                 _("%s: unable to resolve host address `%s'\n"),
                 exec_name, host);
      return E_HOST;
    }

  address_list_get_bounds (al, &start, &end);
  for (i = start; i < end; i++)
    {
      const ip_address *ip = address_list_address_at (al, i);
      sock = connect_to_ip (ip, port, host);
      if (sock >= 0)
        {
          /* Success. */
          address_list_set_connected (al);
          address_list_release (al);
          return sock;
        }

      /* The attempt to connect has failed.  Continue with the loop
         and try next address. */

      address_list_set_faulty (al, i);
    }

  /* Failed to connect to any of the addresses in AL. */

  if (address_list_connected_p (al))
    {
      /* We connected to AL before, but cannot do so now.  That might
         indicate that our DNS cache entry for HOST has expired.  */
      address_list_release (al);
      al = lookup_host (host, LH_REFRESH);
      goto retry;
    }
  address_list_release (al);

  return -1;
}

/* Create a socket, bind it to local interface BIND_ADDRESS on port
   *PORT, set up a listen backlog, and return the resulting socket, or
   -1 in case of error.

   BIND_ADDRESS is the address of the interface to bind to.  If it is
   NULL, the socket is bound to the default address.  PORT should
   point to the port number that will be used for the binding.  If
   that number is 0, the system will choose a suitable port, and the
   chosen value will be written to *PORT.

   Calling accept() on such a socket waits for and accepts incoming
   TCP connections.  */

int
bind_local (const ip_address *bind_address, int *port)
{
  int sock;
  struct sockaddr_storage ss;
  struct sockaddr *sa = (struct sockaddr *)&ss;

  /* For setting options with setsockopt. */
  int setopt_val = 1;
  void *setopt_ptr = (void *)&setopt_val;
  socklen_t setopt_size = sizeof (setopt_val);

  sock = socket (bind_address->family, SOCK_STREAM, 0);
  if (sock < 0)
    return -1;

#ifdef SO_REUSEADDR
  setsockopt (sock, SOL_SOCKET, SO_REUSEADDR, setopt_ptr, setopt_size);
#endif

  xzero (ss);
  sockaddr_set_data (sa, bind_address, *port);
  if (bind (sock, sa, sockaddr_size (sa)) < 0)
    {
      fd_close (sock);
      return -1;
    }
  DEBUGP (("Local socket fd %d bound.\n", sock));

  /* If *PORT is 0, find out which port we've bound to.  */
  if (*port == 0)
    {
      socklen_t addrlen = sockaddr_size (sa);
      if (getsockname (sock, sa, &addrlen) < 0)
        {
          /* If we can't find out the socket's local address ("name"),
             something is seriously wrong with the socket, and it's
             unusable for us anyway because we must know the chosen
             port.  */
          fd_close (sock);
          return -1;
        }
      sockaddr_get_data (sa, NULL, port);
      DEBUGP (("binding to address %s using port %i.\n",
               print_address (bind_address), *port));
    }
  if (listen (sock, 1) < 0)
    {
      fd_close (sock);
      return -1;
    }
  return sock;
}

/* Like a call to accept(), but with the added check for timeout.

   In other words, accept a client connection on LOCAL_SOCK, and
   return the new socket used for communication with the client.
   LOCAL_SOCK should have been bound, e.g. using bind_local().

   The caller is blocked until a connection is established.  If no
   connection is established for opt.connect_timeout seconds, the
   function exits with an error status.  */