[or-cvs] r10817: Backport r10730: Do not choose guard nodes that appear after (in tor/branches/tor-0_1_2-patches: . doc src/or)

[nickm@xxxxxxxx]

Author: nickm
Date: 2007-07-12 13:00:44 -0400 (Thu, 12 Jul 2007)
New Revision: 10817

Modified:
   tor/branches/tor-0_1_2-patches/
   tor/branches/tor-0_1_2-patches/ChangeLog
   tor/branches/tor-0_1_2-patches/doc/TODO.012
   tor/branches/tor-0_1_2-patches/src/or/circuitbuild.c
Log:
 r13738@catbus:  nickm | 2007-07-12 13:00:35 -0400
 Backport r10730: Do not choose guard nodes that appear after any never-before-connected-to guard.



Property changes on: tor/branches/tor-0_1_2-patches
___________________________________________________________________
 svk:merge ticket from /tor/012 [r13738] on 8246c3cf-6607-4228-993b-4d95d33730f1

Modified: tor/branches/tor-0_1_2-patches/ChangeLog
===================================================================
--- tor/branches/tor-0_1_2-patches/ChangeLog	2007-07-12 17:00:42 UTC (rev 10816)
+++ tor/branches/tor-0_1_2-patches/ChangeLog	2007-07-12 17:00:44 UTC (rev 10817)
@@ -9,12 +9,17 @@
       routerlist while inserting a new router.
     - Fix eventdns.c behavior on Solaris: It is critical to include
       orconfig.h _before_ sys/types.h, so that we can get the expected
-      definition of _FILE_OFFSET_BITS.  [Bugfix on 0.1.2.x]
+      definition of _FILE_OFFSET_BITS.
 
   o Major bugfixes (security):
     - Fix a possible buffer overrun when using BSD natd support.  Bug found
       by "Mr. Croup."
 
+  o Minor bugfixes (guard nodes):
+    - If there's a never-before-connected-to guard node in our list,
+      never choose any guards past it. This way we don't expand our
+      guard list unless we need to.
+
   o Minor bugfixes (security):
     - When sending destroy cells from a circuit's origin, don't include
       the reason for tearing down the circuit. The spec says we didn't,

Modified: tor/branches/tor-0_1_2-patches/doc/TODO.012
===================================================================
--- tor/branches/tor-0_1_2-patches/doc/TODO.012	2007-07-12 17:00:42 UTC (rev 10816)
+++ tor/branches/tor-0_1_2-patches/doc/TODO.012	2007-07-12 17:00:44 UTC (rev 10817)
@@ -13,7 +13,7 @@
   o r10563: use correct types with desc_digest_map.
   o r10566: build correctly on systems where size_t is bigger than ulong. 
   o r10643: eventdns.c behavior fix for solaris.
-  - r10730: Don't choose guards after any never-connected-to guard. (??)
+  o r10730: Don't choose guards after any never-connected-to guard.
   o r10760: fix possible buffer overrun in old BSD natd code
   o r10790: Don't include reasons in destroy cells from the origin.
   - Some fix for bug 455.

Modified: tor/branches/tor-0_1_2-patches/src/or/circuitbuild.c
===================================================================
--- tor/branches/tor-0_1_2-patches/src/or/circuitbuild.c	2007-07-12 17:00:42 UTC (rev 10816)
+++ tor/branches/tor-0_1_2-patches/src/or/circuitbuild.c	2007-07-12 17:00:44 UTC (rev 10817)
@@ -2354,6 +2354,13 @@
       r = entry_is_live(entry, need_uptime, need_capacity, 0);
       if (r && !smartlist_isin(exit_family, r)) {
         smartlist_add(live_entry_guards, r);
+        if (!entry->made_contact) {
+          /* Always start with the first not-yet-contacted entry
+           * guard. Otherwise we might add several new ones, pick
+           * the second new one, and now we've expanded our entry
+           * guard list without needing to. */
+          goto choose_and_finish;
+        }
         if (smartlist_len(live_entry_guards) >= options->NumEntryGuards)
           break; /* we have enough */
       }
@@ -2387,6 +2394,7 @@
     /* live_entry_guards will be empty below. Oh well, we tried. */
   }
 
+ choose_and_finish:
   r = smartlist_choose(live_entry_guards);
   smartlist_free(live_entry_guards);
   smartlist_free(exit_family);