transfer.c

  *exc_fd_set = *exc_fd_set;
}


/*
 * Transfer()
 *
 * This function is what performs the actual transfer. It is capable of
 * doing both ways simultaneously.
 * The transfer must already have been setup by a call to Curl_Transfer().
 *
 * Note that headers are created in a preallocated buffer of a default size.
 * That buffer can be enlarged on demand, but it is never shrinken again.
 *
 * Parts of this function was once written by the friendly Mark Butler
 * <butlerm@xmission.com>.
 */

static CURLcode
Transfer(struct connectdata *conn)
{
  struct SessionHandle *data = conn->data;
  CURLcode result;
  struct Curl_transfer_keeper *k = &conn->keep;
  bool done=FALSE;

  Curl_readwrite_init(conn);

  if((conn->sockfd == -1) && (conn->writesockfd == -1))
    /* nothing to read, nothing to write, we're already OK! */
    return CURLE_OK;

  /* we want header and/or body, if neither then don't do this! */
  if(!conn->getheader && data->set.no_body)
    return CURLE_OK;

  while (!done) {
    struct timeval interval;
    k->readfd = k->rkeepfd;  /* set these every lap in the loop */
    k->writefd = k->wkeepfd;
    interval.tv_sec = 1;
    interval.tv_usec = 0;

    switch (select (k->maxfd, k->readfdp, k->writefdp, NULL, &interval)) {
    case -1: /* select() error, stop reading */
#ifdef EINTR
      /* The EINTR is not serious, and it seems you might get this more
         ofen when using the lib in a multi-threaded environment! */
      if(errno == EINTR)
        ;
      else
#endif
        done = TRUE; /* no more read or write */
      continue;
    case 0:  /* timeout */
      result = Curl_readwrite(conn, &done);
      break;

    default: /* readable descriptors */
      result = Curl_readwrite(conn, &done);
      break;
    }
    if(result)
      return result;
    
    /* "done" signals to us if the transfer(s) are ready */
  }

  return CURLE_OK;
}

CURLcode Curl_pretransfer(struct SessionHandle *data)
{
  if(!data->change.url)
    /* we can't do anything wihout URL */
    return CURLE_URL_MALFORMAT;

#ifdef USE_SSLEAY
  /* Init the SSL session ID cache here. We do it here since we want to
     do it after the *_setopt() calls (that could change the size) but
     before any transfer. */
  Curl_SSL_InitSessions(data, data->set.ssl.numsessions);
#endif

  data->set.followlocation=0; /* reset the location-follow counter */
  data->state.this_is_a_follow = FALSE; /* reset this */
  data->state.errorbuf = FALSE; /* no error has occurred */

 /* Allow data->set.use_port to set which port to use. This needs to be
  * disabled for example when we follow Location: headers to URLs using
  * different ports! */
  data->state.allow_port = TRUE;

#if defined(HAVE_SIGNAL) && defined(SIGPIPE)
  /*************************************************************
   * Tell signal handler to ignore SIGPIPE
   *************************************************************/
  data->state.prev_signal = signal(SIGPIPE, SIG_IGN);
#endif  

  Curl_initinfo(data); /* reset session-specific information "variables" */
  Curl_pgrsStartNow(data);

  return CURLE_OK;
}

CURLcode Curl_posttransfer(struct SessionHandle *data)
{
#if defined(HAVE_SIGNAL) && defined(SIGPIPE)
  /* restore the signal handler for SIGPIPE before we get back */
  signal(SIGPIPE, data->state.prev_signal);
#endif  

  return CURLE_OK;
}

CURLcode Curl_perform(struct SessionHandle *data)
{
  CURLcode res;
  CURLcode res2;
  struct connectdata *conn=NULL;
  char *newurl = NULL; /* possibly a new URL to follow to! */

  res = Curl_pretransfer(data);
  if(res)
    return res;

  /*
   * It is important that there is NO 'return' from this function any any
   * other place than falling down the bottom! This is because we have cleanup
   * stuff that must be done before we get back, and that is only performed
   * after this do-while loop.
   */

  do {
    Curl_pgrsTime(data, TIMER_STARTSINGLE);
    res = Curl_connect(data, &conn);
    if(res == CURLE_OK) {
      res = Curl_do(&conn);

      if(res == CURLE_OK) {
        CURLcode res2; /* just a local extra result container */

        if(conn->protocol&PROT_FTPS)
          /* FTPS, disable ssl while transfering data */
          conn->ssl.use = FALSE;
        res = Transfer(conn); /* now fetch that URL please */
        if(conn->protocol&PROT_FTPS)
          /* FTPS, enable ssl again after havving transferred data */
          conn->ssl.use = TRUE;

        if(res == CURLE_OK)
          /*
           * We must duplicate the new URL here as the connection data
           * may be free()ed in the Curl_done() function.
           */
          newurl = conn->newurl?strdup(conn->newurl):NULL;
        else
          /* The transfer phase returned error, we mark the connection to get
           * closed to prevent being re-used. This is becasue we can't
           * possibly know if the connection is in a good shape or not now. */
          conn->bits.close = TRUE;

        /* Always run Curl_done(), even if some of the previous calls
           failed, but return the previous (original) error code */
        res2 = Curl_done(conn);

        if(CURLE_OK == res)
          res = res2;
      }

      /*
       * Important: 'conn' cannot be used here, since it may have been closed
       * in 'Curl_done' or other functions.
       */

      if((res == CURLE_OK) && newurl) {
        /* Location: redirect
 
           This is assumed to happen for HTTP(S) only!
        */
        char prot[16]; /* URL protocol string storage */
        char letter;   /* used for a silly sscanf */

	if (data->set.maxredirs && (data->set.followlocation >= data->set.maxredirs)) {
	  failf(data,"Maximum (%d) redirects followed", data->set.maxredirs);
          res=CURLE_TOO_MANY_REDIRECTS;
	  break;
	}

        /* mark the next request as a followed location: */
        data->state.this_is_a_follow = TRUE;

        data->set.followlocation++; /* count location-followers */

        if(data->set.http_auto_referer) {
          /* We are asked to automatically set the previous URL as the
             referer when we get the next URL. We pick the ->url field,
             which may or may not be 100% correct */

          if(data->change.referer_alloc)
            /* If we already have an allocated referer, free this first */
            free(data->change.referer);

          data->change.referer = strdup(data->change.url);
          data->change.referer_alloc = TRUE; /* yes, free this later */
        }

        if(2 != sscanf(newurl, "%15[^?&/:]://%c", prot, &letter)) {
          /***
           *DANG* this is an RFC 2068 violation. The URL is supposed
           to be absolute and this doesn't seem to be that!
           ***
           Instead, we have to TRY to append this new path to the old URL
           to the right of the host part. Oh crap, this is doomed to cause
           problems in the future...
          */
          char *protsep;
          char *pathsep;
          char *newest;

          /* we must make our own copy of the URL to play with, as it may
             point to read-only data */
          char *url_clone=strdup(data->change.url);

          if(!url_clone) {
            res = CURLE_OUT_OF_MEMORY;
            break; /* skip out of this loop NOW */
          }

          /* protsep points to the start of the host name */
          protsep=strstr(url_clone, "//");
          if(!protsep)
            protsep=url_clone;
          else
            protsep+=2; /* pass the slashes */

          if('/' != newurl[0]) {
            /* First we need to find out if there's a ?-letter in the URL,
               and cut it and the right-side of that off */
            pathsep = strrchr(protsep, '?');
            if(pathsep)
              *pathsep=0;

            /* we have a relative path to append to the last slash if
               there's one available */
            pathsep = strrchr(protsep, '/');
            if(pathsep)
              *pathsep=0;
          }
          else {
            /* We got a new absolute path for this server, cut off from the
               first slash */
            pathsep = strchr(protsep, '/');
            if(pathsep)
              *pathsep=0;
          }

          newest=(char *)malloc( strlen(url_clone) +
                                 1 + /* possible slash */
                                 strlen(newurl) + 1/* zero byte */);

          if(!newest) {
            res = CURLE_OUT_OF_MEMORY;
            break; /* go go go out from this loop */
          }
          sprintf(newest, "%s%s%s", url_clone, ('/' == newurl[0])?"":"/",
                  newurl);
          free(newurl);
          free(url_clone);
          newurl = newest;
        }
        else
          /* This is an absolute URL, don't allow the custom port number */
          data->state.allow_port = FALSE;

        if(data->change.url_alloc)
          free(data->change.url);
        else
          data->change.url_alloc = TRUE; /* the URL is allocated */
      
        /* TBD: set the URL with curl_setopt() */
        data->change.url = newurl;
        newurl = NULL; /* don't free! */

        infof(data, "Follows Location: to new URL: '%s'\n", data->change.url);

        /*
         * We get here when the HTTP code is 300-399. We need to perform
         * differently based on exactly what return code there was.
         * Discussed on the curl mailing list and posted about on the 26th
         * of January 2001.
         */
        switch(data->info.httpcode) {
        case 300: /* Multiple Choices */
        case 306: /* Not used */
        case 307: /* Temporary Redirect */
        default:  /* for all unknown ones */
          /* These are explicitly mention since I've checked RFC2616 and they
           * seem to be OK to POST to.
           */
          break;
        case 301: /* Moved Permanently */
          /* (quote from RFC2616, section 10.3.2):
           * 
           *  Note: When automatically redirecting a POST request after
           *  receiving a 301 status code, some existing HTTP/1.0 user agents
           *  will erroneously change it into a GET request.
           *
           * ----
           * Warning: Because most of importants user agents do this clear
           * RFC2616 violation, many webservers expect this misbehavior. So
           * these servers often answers to a POST request with an error page.
           * To be sure that libcurl gets the page that most user agents
           * would get, libcurl has to force GET:
           */
          if( data->set.httpreq == HTTPREQ_POST
              || data->set.httpreq == HTTPREQ_POST_FORM) {
            infof(data,
                  "Violate RFC 2616/10.3.2 and switch from POST to GET\n");
            data->set.httpreq = HTTPREQ_GET;
          }
          break;
        case 302: /* Found */
          /* (From 10.3.3)

            Note: RFC 1945 and RFC 2068 specify that the client is not allowed
            to change the method on the redirected request.  However, most
            existing user agent implementations treat 302 as if it were a 303
            response, performing a GET on the Location field-value regardless
            of the original request method. The status codes 303 and 307 have
            been added for servers that wish to make unambiguously clear which
            kind of reaction is expected of the client.

            (From 10.3.4)

            Note: Many pre-HTTP/1.1 user agents do not understand the 303
            status. When interoperability with such clients is a concern, the
            302 status code may be used instead, since most user agents react
            to a 302 response as described here for 303.             
          */
        case 303: /* See Other */
          /* Disable both types of POSTs, since doing a second POST when
           * following isn't what anyone would want! */
          if(data->set.httpreq != HTTPREQ_GET) {
            data->set.httpreq = HTTPREQ_GET; /* enforce GET request */
            infof(data, "Disables POST, goes with %s\n",
                  data->set.no_body?"HEAD":"GET");
          }
          break;
        case 304: /* Not Modified */
          /* 304 means we did a conditional request and it was "Not modified".
           * We shouldn't get any Location: header in this response!
           */
          break;
        case 305: /* Use Proxy */
          /* (quote from RFC2616, section 10.3.6):
           * "The requested resource MUST be accessed through the proxy given
           * by the Location field. The Location field gives the URI of the
           * proxy.  The recipient is expected to repeat this single request
           * via the proxy. 305 responses MUST only be generated by origin
           * servers."
           */
          break;
        }
        Curl_pgrsTime(data, TIMER_REDIRECT);
        Curl_pgrsResetTimes(data);
        continue;
      }
    }
    break; /* it only reaches here when this shouldn't loop */

  } while(1); /* loop if Location: */

  if(newurl)
    free(newurl);

  /* run post-transfer uncondionally, but don't clobber the return code if
     we already have an error code recorder */
  res2 = Curl_posttransfer(data);
  if(!res && res2)
    res = res2;

  return res;
}

CURLcode 
Curl_Transfer(struct connectdata *c_conn, /* connection data */
              int sockfd,	/* socket to read from or -1 */
              int size,		/* -1 if unknown at this point */
              bool getheader,	/* TRUE if header parsing is wanted */
              long *bytecountp,	/* return number of bytes read or NULL */
              int writesockfd,  /* socket to write to, it may very well be
                                   the same we read from. -1 disables */
              long *writebytecountp /* return number of bytes written or
                                       NULL */
              )
{
  struct connectdata *conn = (struct connectdata *)c_conn;
  if(!conn)
    return CURLE_BAD_FUNCTION_ARGUMENT;

  /* now copy all input parameters */
  conn->sockfd = sockfd;
  conn->size = size;
  conn->getheader = getheader;
  conn->bytecountp = bytecountp;
  conn->writesockfd = writesockfd;
  conn->writebytecountp = writebytecountp;

  return CURLE_OK;

}
          
/*
 * local variables:
 * eval: (load-file "../curl-mode.el")
 * end:
 * vim600: fdm=marker
 * vim: et sw=2 ts=2 sts=2 tw=78
 */