directory.c

关于tor匿名通信的源代码

      strlcpy(conn->rend_query, payload, sizeof(conn->rend_query));
      conn->rend_version = 2;
      payload = NULL;
      httpcommand = "GET";
      len = strlen(resource) + 32;
      url = tor_malloc(len);
      tor_snprintf(url, len, "/tor/rendezvous2/%s", resource);
      break;
    case DIR_PURPOSE_UPLOAD_RENDDESC:
      tor_assert(!resource);
      tor_assert(payload);
      httpcommand = "POST";
      url = tor_strdup("/tor/rendezvous/publish");
      break;
    case DIR_PURPOSE_UPLOAD_RENDDESC_V2:
      tor_assert(!resource);
      tor_assert(payload);
      httpcommand = "POST";
      url = tor_strdup("/tor/rendezvous2/publish");
      break;
    default:
      tor_assert(0);
      return;
  }

  if (strlen(proxystring) + strlen(url) >= 4096) {
    log_warn(LD_BUG,
             "Squid does not like URLs longer than 4095 bytes, and this "
             "one is %d bytes long: %s%s",
             (int)(strlen(proxystring) + strlen(url)), proxystring, url);
  }

  tor_snprintf(request, sizeof(request), "%s %s", httpcommand, proxystring);
  connection_write_to_buf(request, strlen(request), TO_CONN(conn));
  connection_write_to_buf(url, strlen(url), TO_CONN(conn));
  tor_free(url);

  if (!strcmp(httpcommand, "GET") && !payload) {
    tor_snprintf(request, sizeof(request),
                 " HTTP/1.0\r\nHost: %s%s%s\r\n\r\n",
                 hoststring,
                 imsstring,
                 proxyauthstring);
  } else {
    tor_snprintf(request, sizeof(request),
                 " HTTP/1.0\r\nContent-Length: %lu\r\nHost: %s%s%s\r\n\r\n",
                 payload ? (unsigned long)payload_len : 0,
                 hoststring,
                 imsstring,
                 proxyauthstring);
  }
  connection_write_to_buf(request, strlen(request), TO_CONN(conn));

  if (payload) {
    /* then send the payload afterwards too */
    connection_write_to_buf(payload, payload_len, TO_CONN(conn));
  }
}

/** Parse an HTTP request string <b>headers</b> of the form
 * \verbatim
 * "\%s [http[s]://]\%s HTTP/1..."
 * \endverbatim
 * If it's well-formed, strdup the second \%s into *<b>url</b>, and
 * nul-terminate it. If the url doesn't start with "/tor/", rewrite it
 * so it does. Return 0.
 * Otherwise, return -1.
 */
static int
parse_http_url(const char *headers, char **url)
{
  char *s, *start, *tmp;

  s = (char *)eat_whitespace_no_nl(headers);
  if (!*s) return -1;
  s = (char *)find_whitespace(s); /* get past GET/POST */
  if (!*s) return -1;
  s = (char *)eat_whitespace_no_nl(s);
  if (!*s) return -1;
  start = s; /* this is it, assuming it's valid */
  s = (char *)find_whitespace(start);
  if (!*s) return -1;

  /* tolerate the http[s] proxy style of putting the hostname in the url */
  if (s-start >= 4 && !strcmpstart(start,"http")) {
    tmp = start + 4;
    if (*tmp == 's')
      tmp++;
    if (s-tmp >= 3 && !strcmpstart(tmp,"://")) {
      tmp = strchr(tmp+3, '/');
      if (tmp && tmp < s) {
        log_debug(LD_DIR,"Skipping over 'http[s]://hostname' string");
        start = tmp;
      }
    }
  }

  if (s-start < 5 || strcmpstart(start,"/tor/")) { /* need to rewrite it */
    *url = tor_malloc(s - start + 5);
    strlcpy(*url,"/tor", s-start+5);
    strlcat((*url)+4, start, s-start+1);
  } else {
    *url = tor_strndup(start, s-start);
  }
  return 0;
}

/** Return a copy of the first HTTP header in <b>headers</b> whose key is
 * <b>which</b>.  The key should be given with a terminating colon and space;
 * this function copies everything after, up to but not including the
 * following \\r\\n. */
static char *
http_get_header(const char *headers, const char *which)
{
  const char *cp = headers;
  while (cp) {
    if (!strcasecmpstart(cp, which)) {
      char *eos;
      cp += strlen(which);
      if ((eos = strchr(cp,'\r')))
        return tor_strndup(cp, eos-cp);
      else
        return tor_strdup(cp);
    }
    cp = strchr(cp, '\n');
    if (cp)
      ++cp;
  }
  return NULL;
}

/** If <b>headers</b> indicates that a proxy was involved, then rewrite
 * <b>conn</b>-\>address to describe our best guess of the address that
 * originated this HTTP request. */
static void
http_set_address_origin(const char *headers, connection_t *conn)
{
  char *fwd;

  fwd = http_get_header(headers, "Forwarded-For: ");
  if (!fwd)
    fwd = http_get_header(headers, "X-Forwarded-For: ");
  if (fwd) {
    struct in_addr in;
    if (!tor_inet_aton(fwd, &in) || is_internal_IP(ntohl(in.s_addr), 0)) {
      log_debug(LD_DIR, "Ignoring unrecognized or internal IP %s",
                escaped(fwd));
      tor_free(fwd);
      return;
    }
    tor_free(conn->address);
    conn->address = tor_strdup(fwd);
    tor_free(fwd);
  }
}

/** Parse an HTTP response string <b>headers</b> of the form
 * \verbatim
 * "HTTP/1.\%d \%d\%s\r\n...".
 * \endverbatim
 *
 * If it's well-formed, assign the status code to *<b>code</b> and
 * return 0.  Otherwise, return -1.
 *
 * On success: If <b>date</b> is provided, set *date to the Date
 * header in the http headers, or 0 if no such header is found.  If
 * <b>compression</b> is provided, set *<b>compression</b> to the
 * compression method given in the Content-Encoding header, or 0 if no
 * such header is found, or -1 if the value of the header is not
 * recognized.  If <b>reason</b> is provided, strdup the reason string
 * into it.
 */
int
parse_http_response(const char *headers, int *code, time_t *date,
                    compress_method_t *compression, char **reason)
{
  int n1, n2;
  char datestr[RFC1123_TIME_LEN+1];
  smartlist_t *parsed_headers;
  tor_assert(headers);
  tor_assert(code);

  while (TOR_ISSPACE(*headers)) headers++; /* tolerate leading whitespace */

  if (sscanf(headers, "HTTP/1.%d %d", &n1, &n2) < 2 ||
      (n1 != 0 && n1 != 1) ||
      (n2 < 100 || n2 >= 600)) {
    log_warn(LD_HTTP,"Failed to parse header %s",escaped(headers));
    return -1;
  }
  *code = n2;

  parsed_headers = smartlist_create();
  smartlist_split_string(parsed_headers, headers, "\n",
                         SPLIT_SKIP_SPACE|SPLIT_IGNORE_BLANK, -1);
  if (reason) {
    smartlist_t *status_line_elements = smartlist_create();
    tor_assert(smartlist_len(parsed_headers));
    smartlist_split_string(status_line_elements,
                           smartlist_get(parsed_headers, 0),
                           " ", SPLIT_SKIP_SPACE|SPLIT_IGNORE_BLANK, 3);
    tor_assert(smartlist_len(status_line_elements) <= 3);
    if (smartlist_len(status_line_elements) == 3) {
      *reason = smartlist_get(status_line_elements, 2);
      smartlist_set(status_line_elements, 2, NULL); /* Prevent free */
    }
    SMARTLIST_FOREACH(status_line_elements, char *, cp, tor_free(cp));
    smartlist_free(status_line_elements);
  }
  if (date) {
    *date = 0;
    SMARTLIST_FOREACH(parsed_headers, const char *, s,
      if (!strcmpstart(s, "Date: ")) {
        strlcpy(datestr, s+6, sizeof(datestr));
        /* This will do nothing on failure, so we don't need to check
           the result.   We shouldn't warn, since there are many other valid
           date formats besides the one we use. */
        parse_rfc1123_time(datestr, date);
        break;
      });
  }
  if (compression) {
    const char *enc = NULL;
    SMARTLIST_FOREACH(parsed_headers, const char *, s,
      if (!strcmpstart(s, "Content-Encoding: ")) {
        enc = s+18; break;
      });
    if (!enc || !strcmp(enc, "identity")) {
      *compression = NO_METHOD;
    } else if (!strcmp(enc, "deflate") || !strcmp(enc, "x-deflate")) {
      *compression = ZLIB_METHOD;
    } else if (!strcmp(enc, "gzip") || !strcmp(enc, "x-gzip")) {
      *compression = GZIP_METHOD;
    } else {
      log_info(LD_HTTP, "Unrecognized content encoding: %s. Trying to deal.",
               escaped(enc));
      *compression = UNKNOWN_METHOD;
    }
  }
  SMARTLIST_FOREACH(parsed_headers, char *, s, tor_free(s));
  smartlist_free(parsed_headers);

  return 0;
}

/** Return true iff <b>body</b> doesn't start with a plausible router or
 * running-list or directory opening.  This is a sign of possible compression.
 **/
static int
body_is_plausible(const char *body, size_t len, int purpose)
{
  int i;
  if (len == 0)
    return 1; /* empty bodies don't need decompression */
  if (len < 32)
    return 0;
  if (purpose != DIR_PURPOSE_FETCH_RENDDESC) {
    if (!strcmpstart(body,"router") ||
        !strcmpstart(body,"signed-directory") ||
        !strcmpstart(body,"network-status") ||
        !strcmpstart(body,"running-routers"))
    return 1;
    for (i=0;i<32;++i) {
      if (!TOR_ISPRINT(body[i]) && !TOR_ISSPACE(body[i]))
        return 0;
    }
    return 1;
  } else {
    return 1;
  }
}

/** Called when we've just fetched a bunch of router descriptors in
 * <b>body</b>.  The list <b>which</b>, if present, holds digests for
 * descriptors we requested: descriptor digests if <b>descriptor_digests</b>
 * is true, or identity digests otherwise.  Parse the descriptors, validate
 * them, and annotate them as having purpose <b>purpose</b> and as having been
 * downloaded from <b>source</b>. */
static void
load_downloaded_routers(const char *body, smartlist_t *which,
                        int descriptor_digests,
                        int router_purpose,
                        const char *source)
{
  char buf[256];
  char time_buf[ISO_TIME_LEN+1];
  int general = router_purpose == ROUTER_PURPOSE_GENERAL;
  format_iso_time(time_buf, time(NULL));
  tor_assert(source);

  if (tor_snprintf(buf, sizeof(buf),
                   "@downloaded-at %s\n"
                   "@source %s\n"
                   "%s%s%s", time_buf, escaped(source),
                   !general ? "@purpose " : "",
                   !general ? router_purpose_to_string(router_purpose) : "",
                   !general ? "\n" : "")<0)
    return;

  router_load_routers_from_string(body, NULL, SAVED_NOWHERE, which,
                                  descriptor_digests, buf);
}

/** We are a client, and we've finished reading the server's
 * response. Parse it and act appropriately.
 *
 * If we're still happy with using this directory server in the future, return
 * 0. Otherwise return -1; and the caller should consider trying the request
 * again.
 *
 * The caller will take care of marking the connection for close.
 */
static int
connection_dir_client_reached_eof(dir_connection_t *conn)
{
  char *body;
  char *headers;
  char *reason = NULL;
  size_t body_len=0, orig_len=0;
  int status_code;
  time_t date_header=0;
  long delta;
  compress_method_t compression;
  int plausible;
  int skewed=0;
  int allow_partial = (conn->_base.purpose == DIR_PURPOSE_FETCH_SERVERDESC ||
                       conn->_base.purpose == DIR_PURPOSE_FETCH_EXTRAINFO);
  int was_compressed=0;
  time_t now = time(NULL);

  switch (fetch_from_buf_http(conn->_base.inbuf,
                              &headers, MAX_HEADERS_SIZE,
                              &body, &body_len, MAX_DIR_DL_SIZE,
                              allow_partial)) {
    case -1: /* overflow */