http.c

一个从网络上自动下载文件的自由工具

      if (b < e && e[-1] == '\n')
        --e;
      if (b < e && e[-1] == '\r')
        --e;
      print_response_line(prefix, b, e);
    }
}

/* Parse the `Content-Range' header and extract the information it
   contains.  Returns true if successful, false otherwise.  */
static bool
parse_content_range (const char *hdr, wgint *first_byte_ptr,
                     wgint *last_byte_ptr, wgint *entity_length_ptr)
{
  wgint num;

  /* Ancient versions of Netscape proxy server, presumably predating
     rfc2068, sent out `Content-Range' without the "bytes"
     specifier.  */
  if (0 == strncasecmp (hdr, "bytes", 5))
    {
      hdr += 5;
      /* "JavaWebServer/1.1.1" sends "bytes: x-y/z", contrary to the
         HTTP spec. */
      if (*hdr == ':')
        ++hdr;
      while (ISSPACE (*hdr))
        ++hdr;
      if (!*hdr)
        return false;
    }
  if (!ISDIGIT (*hdr))
    return false;
  for (num = 0; ISDIGIT (*hdr); hdr++)
    num = 10 * num + (*hdr - '0');
  if (*hdr != '-' || !ISDIGIT (*(hdr + 1)))
    return false;
  *first_byte_ptr = num;
  ++hdr;
  for (num = 0; ISDIGIT (*hdr); hdr++)
    num = 10 * num + (*hdr - '0');
  if (*hdr != '/' || !ISDIGIT (*(hdr + 1)))
    return false;
  *last_byte_ptr = num;
  ++hdr;
  for (num = 0; ISDIGIT (*hdr); hdr++)
    num = 10 * num + (*hdr - '0');
  *entity_length_ptr = num;
  return true;
}

/* Read the body of the request, but don't store it anywhere and don't
   display a progress gauge.  This is useful for reading the bodies of
   administrative responses to which we will soon issue another
   request.  The response is not useful to the user, but reading it
   allows us to continue using the same connection to the server.

   If reading fails, false is returned, true otherwise.  In debug
   mode, the body is displayed for debugging purposes.  */

static bool
skip_short_body (int fd, wgint contlen)
{
  enum {
    SKIP_SIZE = 512,                /* size of the download buffer */
    SKIP_THRESHOLD = 4096        /* the largest size we read */
  };
  char dlbuf[SKIP_SIZE + 1];
  dlbuf[SKIP_SIZE] = '\0';        /* so DEBUGP can safely print it */

  /* We shouldn't get here with unknown contlen.  (This will change
     with HTTP/1.1, which supports "chunked" transfer.)  */
  assert (contlen != -1);

  /* If the body is too large, it makes more sense to simply close the
     connection than to try to read the body.  */
  if (contlen > SKIP_THRESHOLD)
    return false;

  DEBUGP (("Skipping %s bytes of body: [", number_to_static_string (contlen)));

  while (contlen > 0)
    {
      int ret = fd_read (fd, dlbuf, MIN (contlen, SKIP_SIZE), -1);
      if (ret <= 0)
        {
          /* Don't normally report the error since this is an
             optimization that should be invisible to the user.  */
          DEBUGP (("] aborting (%s).\n",
                   ret < 0 ? fd_errstr (fd) : "EOF received"));
          return false;
        }
      contlen -= ret;
      /* Safe even if %.*s bogusly expects terminating \0 because
         we've zero-terminated dlbuf above.  */
      DEBUGP (("%.*s", ret, dlbuf));
    }

  DEBUGP (("] done.\n"));
  return true;
}

/* Extract a parameter from the string (typically an HTTP header) at
   **SOURCE and advance SOURCE to the next parameter.  Return false
   when there are no more parameters to extract.  The name of the
   parameter is returned in NAME, and the value in VALUE.  If the
   parameter has no value, the token's value is zeroed out.

   For example, if *SOURCE points to the string "attachment;
   filename=\"foo bar\"", the first call to this function will return
   the token named "attachment" and no value, and the second call will
   return the token named "filename" and value "foo bar".  The third
   call will return false, indicating no more valid tokens.  */

bool
extract_param (const char **source, param_token *name, param_token *value,
               char separator)
{
  const char *p = *source;

  while (ISSPACE (*p)) ++p;
  if (!*p)
    {
      *source = p;
      return false;             /* no error; nothing more to extract */
    }

  /* Extract name. */
  name->b = p;
  while (*p && !ISSPACE (*p) && *p != '=' && *p != separator) ++p;
  name->e = p;
  if (name->b == name->e)
    return false;               /* empty name: error */
  while (ISSPACE (*p)) ++p;
  if (*p == separator || !*p)           /* no value */
    {
      xzero (*value);
      if (*p == separator) ++p;
      *source = p;
      return true;
    }
  if (*p != '=')
    return false;               /* error */

  /* *p is '=', extract value */
  ++p;
  while (ISSPACE (*p)) ++p;
  if (*p == '"')                /* quoted */
    {
      value->b = ++p;
      while (*p && *p != '"') ++p;
      if (!*p)
        return false;
      value->e = p++;
      /* Currently at closing quote; find the end of param. */
      while (ISSPACE (*p)) ++p;
      while (*p && *p != separator) ++p;
      if (*p == separator)
        ++p;
      else if (*p)
        /* garbage after closed quote, e.g. foo="bar"baz */
        return false;
    }
  else                          /* unquoted */
    {
      value->b = p;
      while (*p && *p != separator) ++p;
      value->e = p;
      while (value->e != value->b && ISSPACE (value->e[-1]))
        --value->e;
      if (*p == separator) ++p;
    }
  *source = p;
  return true;
}

#undef MAX
#define MAX(p, q) ((p) > (q) ? (p) : (q))

/* Parse the contents of the `Content-Disposition' header, extracting
   the information useful to Wget.  Content-Disposition is a header
   borrowed from MIME; when used in HTTP, it typically serves for
   specifying the desired file name of the resource.  For example:

       Content-Disposition: attachment; filename="flora.jpg"

   Wget will skip the tokens it doesn't care about, such as
   "attachment" in the previous example; it will also skip other
   unrecognized params.  If the header is syntactically correct and
   contains a file name, a copy of the file name is stored in
   *filename and true is returned.  Otherwise, the function returns
   false.

   The file name is stripped of directory components and must not be
   empty.  */

static bool
parse_content_disposition (const char *hdr, char **filename)
{
  param_token name, value;
  while (extract_param (&hdr, &name, &value, ';'))
    if (BOUNDED_EQUAL_NO_CASE (name.b, name.e, "filename") && value.b != NULL)
      {
        /* Make the file name begin at the last slash or backslash. */
        const char *last_slash = memrchr (value.b, '/', value.e - value.b);
        const char *last_bs = memrchr (value.b, '\\', value.e - value.b);
        if (last_slash && last_bs)
          value.b = 1 + MAX (last_slash, last_bs);
        else if (last_slash || last_bs)
          value.b = 1 + (last_slash ? last_slash : last_bs);
        if (value.b == value.e)
          continue;
        /* Start with the directory prefix, if specified. */
        if (opt.dir_prefix)
          {
            int prefix_length = strlen (opt.dir_prefix);
            bool add_slash = (opt.dir_prefix[prefix_length - 1] != '/');
            int total_length;

            if (add_slash) 
              ++prefix_length;
            total_length = prefix_length + (value.e - value.b);            
            *filename = xmalloc (total_length + 1);
            strcpy (*filename, opt.dir_prefix);
            if (add_slash) 
              (*filename)[prefix_length - 1] = '/';
            memcpy (*filename + prefix_length, value.b, (value.e - value.b));
            (*filename)[total_length] = '\0';
          }
        else
          *filename = strdupdelim (value.b, value.e);
        return true;
      }
  return false;
}

/* Persistent connections.  Currently, we cache the most recently used
   connection as persistent, provided that the HTTP server agrees to
   make it such.  The persistence data is stored in the variables
   below.  Ideally, it should be possible to cache an arbitrary fixed
   number of these connections.  */

/* Whether a persistent connection is active. */
static bool pconn_active;

static struct {
  /* The socket of the connection.  */
  int socket;

  /* Host and port of the currently active persistent connection. */
  char *host;
  int port;

  /* Whether a ssl handshake has occoured on this connection.  */
  bool ssl;

  /* Whether the connection was authorized.  This is only done by
     NTLM, which authorizes *connections* rather than individual
     requests.  (That practice is peculiar for HTTP, but it is a
     useful optimization.)  */
  bool authorized;

#ifdef ENABLE_NTLM
  /* NTLM data of the current connection.  */
  struct ntlmdata ntlm;
#endif
} pconn;

/* Mark the persistent connection as invalid and free the resources it
   uses.  This is used by the CLOSE_* macros after they forcefully
   close a registered persistent connection.  */

static void
invalidate_persistent (void)
{
  DEBUGP (("Disabling further reuse of socket %d.\n", pconn.socket));
  pconn_active = false;
  fd_close (pconn.socket);
  xfree (pconn.host);
  xzero (pconn);
}

/* Register FD, which should be a TCP/IP connection to HOST:PORT, as
   persistent.  This will enable someone to use the same connection
   later.  In the context of HTTP, this must be called only AFTER the
   response has been received and the server has promised that the
   connection will remain alive.

   If a previous connection was persistent, it is closed. */

static void
register_persistent (const char *host, int port, int fd, bool ssl)
{
  if (pconn_active)
    {
      if (pconn.socket == fd)
        {
          /* The connection FD is already registered. */
          return;
        }
      else
        {
          /* The old persistent connection is still active; close it
             first.  This situation arises whenever a persistent
             connection exists, but we then connect to a different
             host, and try to register a persistent connection to that
             one.  */
          invalidate_persistent ();
        }
    }

  pconn_active = true;
  pconn.socket = fd;
  pconn.host = xstrdup (host);
  pconn.port = port;
  pconn.ssl = ssl;
  pconn.authorized = false;

  DEBUGP (("Registered socket %d for persistent reuse.\n", fd));
}

/* Return true if a persistent connection is available for connecting
   to HOST:PORT.  */

static bool
persistent_available_p (const char *host, int port, bool ssl,
                        bool *host_lookup_failed)
{
  /* First, check whether a persistent connection is active at all.  */
  if (!pconn_active)
    return false;

  /* If we want SSL and the last connection wasn't or vice versa,
     don't use it.  Checking for host and port is not enough because
     HTTP and HTTPS can apparently coexist on the same port.  */
  if (ssl != pconn.ssl)
    return false;

  /* If we're not connecting to the same port, we're not interested. */
  if (port != pconn.port)
    return false;

  /* If the host is the same, we're in business.  If not, there is
     still hope -- read below.  */
  if (0 != strcasecmp (host, pconn.host))
    {
      /* Check if pconn.socket is talking to HOST under another name.
         This happens often when both sites are virtual hosts
         distinguished only by name and served by the same network
         interface, and hence the same web server (possibly set up by
         the ISP and serving many different web sites).  This
         admittedly unconventional optimization does not contradict
         HTTP and works well with popular server software.  */

      bool found;
      ip_address ip;
      struct address_list *al;

      if (ssl)
        /* Don't try to talk to two different SSL sites over the same
           secure connection!  (Besides, it's not clear that
           name-based virtual hosting is even possible with SSL.)  */
        return false;

      /* If pconn.socket's peer is one of the IP addresses HOST
         resolves to, pconn.socket is for all intents and purposes
         already talking to HOST.  */

      if (!socket_ip_address (pconn.socket, &ip, ENDPOINT_PEER))
        {
          /* Can't get the peer's address -- something must be very
             wrong with the connection.  */
          invalidate_persistent ();
          return false;
        }
      al = lookup_host (host, 0);
      if (!al)
        {
          *host_lookup_failed = true;
          return false;
        }

      found = address_list_contains (al, &ip);
      address_list_release (al);

      if (!found)
        return false;

      /* The persistent connection's peer address was found among the
         addresses HOST resolved to; therefore, pconn.sock is in fact
         already talking to HOST -- no need to reconnect.  */
    }

  /* Finally, check whether the connection is still open.  This is
     important because most servers implement liberal (short) timeout
     on persistent connections.  Wget can of course always reconnect
     if the connection doesn't work out, but it's nicer to know in
     advance.  This test is a logical followup of the first test, but
     is "expensive" and therefore placed at the end of the list.

     (Current implementation of test_socket_open has a nice side
     effect that it treats sockets with pending data as "closed".
     This is exactly what we want: if a broken server sends message
     body in response to HEAD, or if it sends more than conent-length
     data, we won't reuse the corrupted connection.)  */

  if (!test_socket_open (pconn.socket))
    {
      /* Oops, the socket is no longer open.  Now that we know that,
         let's invalidate the persistent connection before returning
         0.  */