Olsrd crash fixes (#1234)

* Simplify the OLSR watchdog

* Dont pull the routing table into LQM for supernodes.
LQM tracks routes on nodes to help keep leaf nodes connected even
when circumstances would probably prevent this. However on supernodes
the routing table is massive and pulling this into LQM will frequently
crash OLSRD. As we dont need this for supernode just dont do it for them.

files/usr/local/bin/mgr/lqm.lua

@@ -247,6 +247,7 @@ end
 
 local myhostname = canonical_hostname(aredn.info.get_nvram("node") or "localnode")
 local myip = uci.cursor():get("network", "wifi", "ipaddr")
+local is_supernode = uci.cursor():get("aredn", "@supernode[0]", "enable") == "1"
 
 local wgsupport = nixio.fs.stat("/usr/bin/wg")
 
@@ -803,8 +804,11 @@ function lqm()
 
         --
         -- Pull in the routing table to see how many node routes are associated with each tracker.
+        -- We dont do this if this is a supernode because the routes table is massive and can cause
+        -- crash olsrd.
         --
         total_node_route_count = 0
+        if not is_supernode then
             for _, route in ipairs(aredn.olsr.getOLSRRoutes())
             do
                 -- Count routes to nodes. There are two routes to most nodes, the node's primary address
@@ -817,6 +821,7 @@ function lqm()
                     end
                 end
             end
+        end
 
         --
         -- At this point we have gather all the data we need to determine which links are best to use and

files/usr/local/bin/mgr/olsrd_watchdog.lua

@@ -36,53 +36,22 @@
 --]]
 
 local watchdogfile = "/tmp/olsrd.watchdog"
-local pidfile = "/var/run/olsrd.pid"
-local logfile = "/tmp/olsrd.log"
-
-function olsrd_restart()
-    -- print "olsrd_restart"
-
-    os.execute("/etc/init.d/olsrd restart")
-
-    if nixio.fs.stat(logfile) then
-        local lines = read_all(logfile):splitNewLine()
-        lines[#lines + 1] = secondsToClock(nixio.sysinfo().uptime) .. " " .. os.date()
-        local start = 1
-        if #lines > 300 then
-            start = #lines - 275
-        end
-        local f = io.open(logfile, "w")
-        if f then
-            for i = start, #lines
-            do
-                f:write(lines[i] .. "\n")
-            end
-            f:close()
-        end
-    end
-end
+local sleeptime = 3 * 60 -- 3 minutes
+local timeout = 10 * 60 -- 10 minutes
 
 function olsrd_watchdog()
     while true
     do
-        wait_for_ticks(223)
-
-        local pid = read_all(pidfile)
-        if pid and nixio.fs.stat("/proc/" .. pid) then
+        wait_for_ticks(sleeptime)
         if nixio.fs.stat(watchdogfile) then
+            local watchtime = tonumber(read_all(watchdogfile))
+            -- If watchtime hasn't update recently then we restart OLSRD
+            if watchtime + timeout < os.time() then
+                nixio.syslog("err", "olsrd watchdog timeout - restarting")
                 os.remove(watchdogfile)
-            else
-                olsrd_restart()
-            end
-        else
-            local pids = capture("pidof olsrd"):splitWhiteSpace()
-            if #pids == 1 then
-                write_all(pidfile, pids[1]);
-            elseif #pids == 0 then
-                olsrd_restart()
+                os.execute("/etc/init.d/olsrd restart")
             end
         end
-
     end
 end