Do not not cannibalize a circuit that has run out of RELAY_EARLY cells. Partial bug 878 fix.

svn:r17815

ChangeLog

@@ -18,6 +18,11 @@ Changes in version 0.2.1.10-alpha - 2009-01-??
       send on that circuit.  Otherwise we might violate the proposal-110
       limit.  Bugfix on 0.2.1.3-alpha.  Partial fix for Bug 878.  Diagnosis
       thanks to Karsten.
+    - When we're sending non-EXTEND cells to the first hop in a circuit
+      (for example in order to use an encrypted directory connection), we
+      don't need to use RELAY_EARLY cells: the first hop knows what kind of
+      cell it is, and nobody else can even see the cell type.  This makes
+      it easier to cannibalize circuits like this later.
 
   o Code simplifications and refactoring:
     - Change our header file guard macros to be less likely to conflict

src/or/relay.c

@@ -510,10 +510,13 @@ relay_send_command_from_edge(uint16_t stream_id, circuit_t *circ,
 
   if (cell_direction == CELL_DIRECTION_OUT) {
     origin_circuit_t *origin_circ = TO_ORIGIN_CIRCUIT(circ);
-    if (origin_circ->remaining_relay_early_cells > 0) {
-      /* If we've got any relay_early cells left, use one.  Don't worry
-       * about the conn protocol version: append_cell_to_circuit_queue will
-       * fix it up. */
+    if (origin_circ->remaining_relay_early_cells > 0 &&
+        (relay_command == RELAY_COMMAND_EXTEND ||
+         cpath_layer != origin_circ->cpath)) {
+      /* If we've got any relay_early cells left, and we're sending a relay
+       * cell or we're not talking to the first hop, use one of them.  Don't
+       * worry about the conn protocol version: append_cell_to_circuit_queue
+       * will fix it up. */
       cell.command = CELL_RELAY_EARLY;
       --origin_circ->remaining_relay_early_cells;
       log_debug(LD_OR, "Sending a RELAY_EARLY cell; %d remaining.",