Bug 12808:

Summary of changes:
  - Mutiple routing table for each public interface is added (previously there is only one routing table ). when the packet is send out of public interface corresponding per-interface routing table will be used. per-interface routing table will modified when ever ip/interface added/deleted.
  - New parameter is added to ipassoc command  to include the default gateway for every interface/ip. prevously it is using only one public interface to send out, default gateway is obtained at the boot up time.
  - In the DNAT case. In the revese path(from guest vm to outside, or when DNAT packet receives from the eth0) the public ip/source ip will not be available till POSTROUTING. to overcome this, DNAT connection are marked with routing table number at the time of connection creation, in the reverse path the routing table# from DNAT connection is used to detect per-interface routing table.

core/src/com/cloud/agent/resource/virtualnetwork/VirtualRoutingResource.java

@@ -635,6 +635,9 @@ public class VirtualRoutingResource implements Manager {
 
         String publicNic = "eth" + nicNum;
         command.add("-c", publicNic);
+        
+        command.add("-g", vlanGateway);
+        
 
         return command.execute();
     }

core/src/com/cloud/hypervisor/vmware/resource/VmwareResource.java

@@ -752,7 +752,10 @@ public class VmwareResource implements StoragePoolResource, ServerResource, Vmwa
         }
 
         args += " -c ";
-        args += "eth" + publicNicInfo.first();
+        args += "eth" + publicNicInfo.first();
+        
+        args += " -g ";
+        args += vlanGateway;
 
         if (s_logger.isDebugEnabled()) {
             s_logger.debug("Run command on domain router " + privateIpAddress + ", /root/ipassoc.sh " + args);

core/src/com/cloud/hypervisor/xen/resource/CitrixResourceBase.java

@@ -1639,7 +1639,10 @@ public abstract class CitrixResourceBase implements ServerResource, HypervisorRe
 
             args += " -c ";
             args += "eth" + correctVif.getDevice(conn);
-
+            
+            args += " -g ";
+            args += vlanGateway;
+            
 
             String result = callHostPlugin(conn, "vmops", "ipassoc", "args", args);
             if (result == null || result.isEmpty()) {

patches/systemvm/debian/config/etc/iptables/iptables-router

@@ -30,4 +30,5 @@ COMMIT
 :OUTPUT ACCEPT [0:0]
 :POSTROUTING ACCEPT [0:0]
 -A POSTROUTING -p udp --dport bootpc -j CHECKSUM --checksum-fill
+-A PREROUTING -m state --state ESTABLISHED,RELATED -j CONNMARK --restore-mark
 COMMIT

patches/systemvm/debian/config/root/firewall.sh

@@ -59,6 +59,7 @@ tcp_or_udp_entry() {
   local op=$5
   local proto=$6
   local cidrs=$7
+
   logger -t cloud "$(basename $0): creating port fwd entry for PAT: public ip=$publicIp \
   instance ip=$instIp proto=$proto port=$port dport=$dport op=$op"
 
@@ -67,11 +68,16 @@ tcp_or_udp_entry() {
   # the delete operation may have errored out but the only possible reason is 
   # that the rules didn't exist in the first place
   local dev=$(ip_to_dev $publicIp)
+  local tableNo=$(echo $dev | awk -F'eth' '{print $2}')
   # shortcircuit the process if error and it is an append operation
   # continue if it is delete
   (sudo iptables -t nat $op PREROUTING --proto $proto -i $dev -d $publicIp \
            --destination-port $port -j DNAT  \
            --to-destination $instIp:$dport &>> $OUTFILE || [ "$op" == "-D" ]) &&
+  (sudo iptables -t mangle $op PREROUTING --proto $proto -i $dev -d $publicIp \
+           --destination-port $port -j MARK --set-mark $tableNo) && 
+  (sudo iptables -t mangle $op PREROUTING --proto $proto -i $dev -d $publicIp \
+           --destination-port $port -m state --state NEW -j CONNMARK --save-mark) &&
   (sudo iptables -t nat $op OUTPUT  --proto $proto -d $publicIp  \
            --destination-port $port -j DNAT  \
            --to-destination $instIp:$dport &>> $OUTFILE || [ "$op" == "-D" ]) &&

patches/systemvm/debian/config/root/ipassoc.sh

@@ -134,6 +134,72 @@ convert_primary_to_32() {
   done
 }
 
+remove_routing() {
+  local pubIp=$1
+  local ipNoMask=$(echo $pubIp | awk -F'/' '{print $1}')
+  local mask=$(echo $pubIp | awk -F'/' '{print $2}')
+  local tableNo=$(echo $ethDev | awk -F'eth' '{print $2}')
+
+  local tableName="Table_$ethDev"
+  local ethMask=$(ip route show | grep $ethDev | grep src | awk '{print $1}')
+  if [ "$ethMask" != "" ]
+  then
+# rules and routes will be deleted for the last ip of the interface.
+     sudo ip route delete throw $ethMask  table $tableName proto static
+     sudo ip rule delete from $ethMask table $tableName
+     sudo ip rule delete fwmark $tableNo table $tableName
+     sudo ip route flush cache
+  fi
+}
+
+# copy eth0,eth1 and the current public interface
+copy_routes_from_main() {
+  local tableName=$1
+
+#get the network masks from the main table
+  local eth0Mask=$(ip route show | grep eth0 | grep src | awk '{print $1}')
+  local eth1Mask=$(ip route show | grep eth1 | grep src | awk '{print $1}')
+  local ethMask=$(ip route show | grep $ethDev | grep src | awk '{print $1}')
+
+# eth0,eth1 and other know routes will be skipped, so as main routing table will decide the route. This will be useful if the interface is down and up.  
+  sudo ip route add throw $eth0Mask table $tableName proto static 
+  sudo ip route add throw $eth1Mask table $tableName proto static 
+  sudo ip route add throw $ethMask  table $tableName proto static 
+  return 0;
+}
+
+add_routing() {
+  local pubIp=$1
+  local ipNoMask=$(echo $1 | awk -F'/' '{print $1}')
+  local mask=$(echo $1 | awk -F'/' '{print $2}')
+
+  local tableName="Table_$ethDev"
+  local tablePresent=$(grep $tableName /etc/iproute2/rt_tables)
+  local tableNo=$(echo $ethDev | awk -F'eth' '{print $2}')
+  if [ "$tablePresent" == "" ]
+  then
+     if [ "$tableNo" == ""] 
+     then
+       return 0;
+     fi
+     echo "$tableNo $tableName" >> /etc/iproute2/rt_tables
+  fi
+
+  copy_routes_from_main $tableName
+# NOTE: this  entry will be deleted if the interface is down without knowing to Management server, in that case all the outside traffic will be send through main routing table or it will be the first public NIC.
+  sudo ip route add default via $defaultGwIP table $tableName proto static
+  sudo ip route flush cache
+
+  local ethMask=$(ip route show | grep $ethDev | grep src | awk '{print $1}')
+  local rulePresent=$(ip rule show | grep $ethMask)
+  if [ "$rulePresent" == "" ]
+  then
+# rules will be added while adding the first ip of the interface 
+     sudo ip rule add from $ethMask table $tableName
+     sudo ip rule add fwmark $tableNo table $tableName
+  fi
+  return 0;
+}
 
 add_nat_entry() {
   local pubIp=$1
@@ -167,6 +233,7 @@ add_nat_entry() {
       sudo ip link set $ethDev up
       sudo arping -c 3 -I $ethDev -A -U -s $ipNoMask $ipNoMask;
   fi
+  add_routing $1 
 
   return 0
 }
@@ -182,6 +249,7 @@ del_nat_entry() {
   sudo iptables -t nat -D POSTROUTING   -j SNAT -o $ethDev --to-source $ipNoMask;
   sudo ip addr del dev $ethDev "$ipNoMask/$mask"
  
+  remove_routing $1
   if [ $? -gt 0  -a $? -ne 2 ]
   then
      return 1
@@ -205,7 +273,7 @@ add_an_ip () {
       sudo ip link set $ethDev up
       sudo arping -c 3 -I $ethDev -A -U -s $ipNoMask $ipNoMask;
   fi
-
+  add_routing $1 
   return $?
    
 }
@@ -236,6 +304,7 @@ remove_an_ip () {
         fi
     result=$?
   fi
+  remove_routing $1
   if [ $result -gt 0  -a $result -ne 2 ]
   then
      return 1
@@ -268,7 +337,7 @@ then
     if_keep_state=1
 fi
 
-while getopts 'fADa:l:c:' OPTION
+while getopts 'fADa:l:c:g:' OPTION
 do
   case $OPTION in
   A)	Aflag=1
@@ -285,6 +354,9 @@ do
   c)	cflag=1
   		ethDev="$OPTARG"
   		;;
+  g)	gflag=1
+  		defaultGwIP="$OPTARG"
+  		;;
   ?)	usage
                 unlock_exit 2 $lock $locked
 		;;