If the controller claimed responsibility for a stream, but that

stream never finished making its connection, it would live
forever in circuit_wait state. Now we close it after SocksTimeout
seconds. Bugfix on 0.1.2.7-alpha; reported by Mike Perry.


svn:r18516

ChangeLog

@@ -1,4 +1,4 @@
-Changes in version 0.2.1.13-????? - 2009-0?-??
+Changes in version 0.2.1.13-????? - 2009-02-??
   o Minor bugfixes:
     - Automatically detect MacOSX versions earlier than 10.4.0, and
       disable kqueue from inside Tor when running with these versions.
@@ -19,6 +19,10 @@ Changes in version 0.2.1.13-????? - 2009-0?-??
     - When we can't transmit a DNS request due to a network error, retry
       it after a while, and eventually transmit a failing response to the
       RESOLVED cell.  Bugfix on 0.1.2.5-alpha.
+    - If the controller claimed responsibility for a stream, but that
+      stream never finished making its connection, it would live
+      forever in circuit_wait state. Now we close it after SocksTimeout
+      seconds. Bugfix on 0.1.2.7-alpha; reported by Mike Perry.
 
   o Minor features:
     - On Linux, use the prctl call to re-enable core dumps when the user

src/or/connection_edge.c

@@ -397,7 +397,7 @@ connection_ap_expire_beginning(void)
   or_options_t *options = get_options();
   int severity;
   int cutoff;
-  int seconds_idle;
+  int seconds_idle, seconds_since_born;
   smartlist_t *conns = get_connection_array();
 
   SMARTLIST_FOREACH_BEGIN(conns, connection_t *, c) {
@@ -408,18 +408,21 @@ connection_ap_expire_beginning(void)
     severity = (tor_addr_is_null(&conn->_base.addr) && !conn->_base.port)
       ? LOG_INFO : LOG_NOTICE;
     seconds_idle = (int)( now - conn->_base.timestamp_lastread );
+    seconds_since_born = (int)( now - conn->_base.timestamp_created );
 
-    /* XXX021 this clause was originally thought redundant with the
-     * clause in connection_ap_handshake_attach_circuit(). But actually,
-     * we need it because controllers that put streams in controller_wait
-     * state never go to the other clause. we should fix so it compares
-     * seconds since timestamp_created, not since last read. -RD */
+    if (conn->_base.state == AP_CONN_STATE_OPEN)
+      continue;
+
+    /* We already consider SocksTimeout in
+     * connection_ap_handshake_attach_circuit(), but we need to consider
+     * it here too because controllers that put streams in controller_wait
+     * state never ask Tor to attach the circuit. */
     if (AP_CONN_STATE_IS_UNATTACHED(conn->_base.state)) {
-      if (seconds_idle >= options->SocksTimeout) {
+      if (seconds_since_born >= options->SocksTimeout) {
         log_fn(severity, LD_APP,
             "Tried for %d seconds to get a connection to %s:%d. "
             "Giving up. (%s)",
-            seconds_idle, safe_str(conn->socks_request->address),
+            seconds_since_born, safe_str(conn->socks_request->address),
             conn->socks_request->port,
             conn_state_to_string(CONN_TYPE_AP, conn->_base.state));
         connection_mark_unattached_ap(conn, END_STREAM_REASON_TIMEOUT);
@@ -427,9 +430,6 @@ connection_ap_expire_beginning(void)
       continue;
     }
 
-    if (conn->_base.state == AP_CONN_STATE_OPEN)
-      continue;
-
     /* We're in state connect_wait or resolve_wait now -- waiting for a
      * reply to our relay cell. See if we want to retry/give up. */