r16582@catbus: nickm | 2007-11-08 13:14:16 -0500

Try to bulletproof the parse logic for router-stability.


svn:r12441

ChangeLog

@@ -42,6 +42,9 @@ Changes in version 0.2.0.10-alpha - 2007-11-07
       requests for all of them.
     - Correctly back off from failing certificate downloads.  Fixes bug
       546.
+    - Authorities don't vote on the Running flag if they have been running
+      for less than 30 minutes themselves.  Fixes bug 547, where a newly
+      started authority would vote that everyone was down.
 
   o New requirements:
     - Drop support for OpenSSL version 0.9.6. Just about nobody was using

src/or/dirserv.c

@@ -27,6 +27,8 @@ const char dirserv_c_id[] =
 /** If a v1 running-routers is older than this, discard it. */
 #define MAX_V1_RR_AGE (7*24*60*60)
 
+extern time_t time_of_process_start; /* from main.c */
+
 /** Do we need to regenerate the directory when someone asks for it? */
 static int the_directory_is_dirty = 1;
 static int runningrouters_is_dirty = 1;
@@ -1926,6 +1928,10 @@ set_routerstatus_from_routerinfo(routerstatus_t *rs,
   rs->dir_port = ri->dir_port;
 }
 
+/** If we've been around for less than this amount of time, our reachability
+ * information is not accurate. */
+#define DIRSERV_TIME_TO_GET_REACHABILITY_INFO (30*60)
+
 /** Return a new networkstatus_vote_t* containing our current opinion. (For v3
  * authorities */
 networkstatus_vote_t *
@@ -1949,10 +1955,14 @@ dirserv_generate_networkstatus_vote_obj(crypto_pk_env_t *private_key,
   networkstatus_voter_info_t *voter = NULL;
   vote_timing_t timing;
   digestmap_t *omit_as_sybil = NULL;
+  int vote_on_reachability = 1;
 
   tor_assert(private_key);
   tor_assert(cert);
 
+  if (now - time_of_process_start < DIRSERV_TIME_TO_GET_REACHABILITY_INFO)
+    vote_on_reachability = 0;
+
   if (resolve_my_address(LOG_WARN, options, &addr, &hostname)<0) {
     log_warn(LD_NET, "Couldn't resolve my hostname");
     return NULL;
@@ -2014,6 +2024,8 @@ dirserv_generate_networkstatus_vote_obj(crypto_pk_env_t *private_key,
           rs->is_running = rs->is_named = rs->is_valid = rs->is_v2_dir =
           rs->is_possible_guard = 0;
       }
+      if (!vote_on_reachability)
+        rs->is_running = 0;
 
       vrs->version = version_from_platform(ri->platform);
       smartlist_add(routerstatuses, vrs);
@@ -2057,8 +2069,10 @@ dirserv_generate_networkstatus_vote_obj(crypto_pk_env_t *private_key,
   v3_out->server_versions = server_versions;
   v3_out->known_flags = smartlist_create();
   smartlist_split_string(v3_out->known_flags,
-                "Authority Exit Fast Guard HSDir Running Stable V2Dir Valid",
+                "Authority Exit Fast Guard HSDir Stable V2Dir Valid",
                 0, SPLIT_SKIP_SPACE|SPLIT_IGNORE_BLANK, 0);
+  if (vote_on_reachability)
+    smartlist_add(v3_out->known_flags, tor_strdup("Running"));
   if (listbadexits)
     smartlist_add(v3_out->known_flags, tor_strdup("BadExit"));
   if (naming) {

src/or/main.c

@@ -53,7 +53,7 @@ static int stats_prev_global_write_bucket;
 static uint64_t stats_n_bytes_read = 0;
 static uint64_t stats_n_bytes_written = 0;
 /** What time did this process start up? */
-long time_of_process_start = 0;
+time_t time_of_process_start = 0;
 /** How many seconds have we been running? */
 long stats_n_seconds_working = 0;
 /** When do we next download a directory? */

src/or/rephist.c

@@ -684,7 +684,7 @@ rep_hist_record_mtbf_data(void)
 
     base16_encode(dbuf, sizeof(dbuf), digest, DIGEST_LEN);
     PRINTF((f, "R %s\n", dbuf));
-    if (hist->start_of_run) {
+    if (hist->start_of_run > 0) {
       format_iso_time(time_buf, hist->start_of_run);
       t = time_buf;
     }
@@ -692,7 +692,7 @@ rep_hist_record_mtbf_data(void)
             hist->weighted_run_length, hist->total_run_weights,
             t ? " S=" : "", t ? t : ""));
     t = NULL;
-    if (hist->start_of_downtime) {
+    if (hist->start_of_downtime > 0) {
       format_iso_time(time_buf, hist->start_of_downtime);
       t = time_buf;
     }
@@ -728,6 +728,43 @@ find_next_with(smartlist_t *sl, int i, const char *prefix)
   return -1;
 }
 
+static int n_bogus_times = 0;
+/** DOCDOC */
+static int
+parse_possibly_bad_iso_time(const char *s, time_t *time_out)
+{
+  int year;
+  char b[5];
+  strlcpy(b, s, sizeof(b));
+  b[4] = '\0';
+  year = atoi(b);
+  if (year < 1970) {
+    *time_out = 0;
+    ++n_bogus_times;
+    return 0;
+  } else
+    return parse_iso_time(s, time_out);
+}
+
+/** DOCDOC */
+static INLINE time_t
+correct_time(time_t t, time_t now, time_t stored_at, time_t started_measuring)
+{
+  if (t < started_measuring - 24*60*60*365)
+    return 0;
+  else if (t < started_measuring)
+    return started_measuring;
+  else if (t > stored_at)
+    return 0;
+  else {
+    long run_length = stored_at - t;
+    t = now - run_length;
+    if (t < started_measuring)
+      t = started_measuring;
+    return t;
+  }
+}
+
 /** Load MTBF data from disk.  Returns 0 on success or recoverable error, -1
  * on failure. */
 int
@@ -799,6 +836,8 @@ rep_hist_load_mtbf_data(time_t now)
   if (line && !strcmp(line, "data"))
     ++i;
 
+  n_bogus_times = 0;
+
   for (; i < smartlist_len(lines); ++i) {
     char digest[DIGEST_LEN];
     char hexbuf[HEX_DIGEST_LEN+1];
@@ -874,16 +913,12 @@ rep_hist_load_mtbf_data(time_t now)
     if (have_mtbf) {
       if (mtbf_timebuf[0]) {
         mtbf_timebuf[10] = ' ';
-        if (parse_iso_time(mtbf_timebuf, &start_of_run)<0)
+        if (parse_possibly_bad_iso_time(mtbf_timebuf, &start_of_run)<0)
           log_warn(LD_GENERAL, "Couldn't parse time %s",
                    escaped(mtbf_timebuf));
       }
-      if (!start_of_run || start_of_run > stored_at) {
-        hist->start_of_run = 0;
-      } else {
-        long run_length = stored_at - start_of_run;
-        hist->start_of_run = now - run_length;
-      }
+      hist->start_of_run = correct_time(start_of_run, now, stored_at,
+                                        tracked_since);
       if (hist->start_of_run < latest_possible_start + wrl)
         latest_possible_start = hist->start_of_run - wrl;
 
@@ -893,16 +928,12 @@ rep_hist_load_mtbf_data(time_t now)
     if (have_wfu) {
       if (wfu_timebuf[0]) {
         wfu_timebuf[10] = ' ';
-        if (parse_iso_time(wfu_timebuf, &start_of_downtime)<0)
+        if (parse_possibly_bad_iso_time(wfu_timebuf, &start_of_downtime)<0)
           log_warn(LD_GENERAL, "Couldn't parse time %s", escaped(wfu_timebuf));
       }
     }
-    if (!start_of_downtime || start_of_downtime > stored_at) {
-      hist->start_of_downtime = 0;
-    } else {
-      long down_length = stored_at - start_of_downtime;
-      hist->start_of_downtime = start_of_downtime - down_length;
-    }
+    hist->start_of_downtime = correct_time(start_of_downtime, now, stored_at,
+                                           tracked_since);
     hist->weighted_uptime = wt_uptime;
     hist->total_weighted_time = total_wt_time;
   }