Remove workaround for ipconfig issues
authorCharles (Chas) Williams <ciwillia@mail.eng.vyatta.net>
Mon, 12 Mar 2018 22:44:01 +0000 (18:44 -0400)
committerCharles (Chas) Williams <ciwillia@mail.eng.vyatta.net>
Tue, 13 Mar 2018 13:14:12 +0000 (09:14 -0400)
If a system has more than a few interfaces, the enforced delay for
each interface can lead to excruciatingly long boot times.  Even with
one interface, there is always a 15s penalty.  The appropriate fix for
issues with ipconfig (and/or klibc) isn't here.

components/9990-networking.sh

index dd1d0f0..8e059c9 100755 (executable)
@@ -89,20 +89,10 @@ do_netsetup ()
                        devlist="$devlist $device"
                done
 
-               # this is tricky (and ugly) because ipconfig sometimes just hangs/runs into
-               # an endless loop; if execution fails give it two further tries, that's
-               # why we use '$devlist $devlist $devlist' for the other for loop
-               for dev in $devlist $devlist $devlist
+               for dev in $devlist
                do
                        echo "Executing ipconfig -t $ETHDEV_TIMEOUT $dev"
-                       ipconfig -t "$ETHDEV_TIMEOUT" $dev | tee -a /netboot.config &
-                       jobid=$!
-                       sleep "$ETHDEV_TIMEOUT" ; sleep 1
-                       if [ -r /proc/"$jobid"/status ]
-                       then
-                               echo "Killing job $jobid for device $dev as ipconfig ran into recursion..."
-                               kill -9 $jobid
-                       fi
+                       ipconfig -t "$ETHDEV_TIMEOUT" $dev | tee -a /netboot.config
 
                        # if configuration of device worked we should have an assigned
                        # IP address, if so let's use the device as $DEVICE for later usage.