Browse Source

Move entry-guard-is-up notification later into dirguard path.

Previously we were marking directory guards up in
..._process_inbuf(), but that's wrong: we call that function on
close as well as on success.  Instead, we're marking the dirguard up
only after we parse the HTTP headers. Closes 20974.
Nick Mathewson 2 years ago
parent
commit
9d37449fb0
2 changed files with 20 additions and 15 deletions
  1. 5 0
      changes/bug20974
  2. 15 15
      src/or/directory.c

+ 5 - 0
changes/bug20974

@@ -0,0 +1,5 @@
+  o Minor bugfixes (guards, bootstrapping):
+    - When connecting to a directory guard during bootstrap, do not
+      mark the guard as successful until we receive a good-looking
+      directory response from it. Fixes bug 20974; bugfix on 0.3.0.1-alpha.
+

+ 15 - 15
src/or/directory.c

@@ -1967,6 +1967,21 @@ connection_dir_client_reached_eof(dir_connection_t *conn)
             escaped(reason),
             conn->base_.purpose);
 
+  if (conn->guard_state) {
+    /* we count the connection as successful once we can read from it.  We do
+     * not, however, delay use of the circuit here, since it's just for a
+     * one-hop directory request. */
+    /* XXXXprop271 note that this will not do the right thing for other
+     * waiting circuits that would be triggered by this circuit becoming
+     * complete/usable. But that's ok, I think.
+     */
+    /* XXXXprop271 should we count this as only a partial success somehow?
+    */
+    entry_guard_succeeded(&conn->guard_state);
+    circuit_guard_state_free(conn->guard_state);
+    conn->guard_state = NULL;
+  }
+
   /* now check if it's got any hints for us about our IP address. */
   if (conn->dirconn_direct) {
     char *guess = http_get_header(headers, X_ADDRESS_HEADER);
@@ -2578,21 +2593,6 @@ connection_dir_process_inbuf(dir_connection_t *conn)
   tor_assert(conn);
   tor_assert(conn->base_.type == CONN_TYPE_DIR);
 
-  if (conn->guard_state) {
-    /* we count the connection as successful once we can read from it.  We do
-     * not, however, delay use of the circuit here, since it's just for a
-     * one-hop directory request. */
-    /* XXXXprop271 note that this will not do the right thing for other
-     * waiting circuits that would be triggered by this circuit becoming
-     * complete/usable. But that's ok, I think.
-     */
-    /* XXXXprop271 should we count this as only a partial success somehow?
-    */
-    entry_guard_succeeded(&conn->guard_state);
-    circuit_guard_state_free(conn->guard_state);
-    conn->guard_state = NULL;
-  }
-
   /* Directory clients write, then read data until they receive EOF;
    * directory servers read data until they get an HTTP command, then
    * write their response (when it's finished flushing, they mark for