SRIOV: allow old sriov switch behavior that doesn't kill connection (#2294)

lisa/features/network_interface.py

@@ -21,7 +21,9 @@ class NetworkInterface(Feature):
     def enabled(self) -> bool:
         return True
 
-    def switch_sriov(self, enable: bool, wait: bool = True) -> None:
+    def switch_sriov(
+        self, enable: bool, wait: bool = True, reset_connections: bool = True
+    ) -> None:
         raise NotImplementedError
 
     def is_enabled_sriov(self) -> bool:

lisa/sut_orchestrator/aws/features.py

@@ -167,7 +167,9 @@ class NetworkInterface(AwsFeatureMixin, features.NetworkInterface):
             raise LisaException(f"fail to find primary nic for vm {self._node.name}")
         return nic
 
-    def switch_sriov(self, enable: bool, wait: bool = True) -> None:
+    def switch_sriov(
+        self, enable: bool, wait: bool = True, reset_connections: bool = True
+    ) -> None:
         aws_platform: AwsPlatform = self._platform  # type: ignore
         instance = boto3.resource("ec2").Instance(self._instance_id)
 

lisa/sut_orchestrator/azure/features.py

@@ -601,7 +601,9 @@ class NetworkInterface(AzureFeatureMixin, features.NetworkInterface):
             len(all_nics) - self.origin_extra_synthetic_nics_count - 1
         )
 
-    def switch_sriov(self, enable: bool, wait: bool = True) -> None:
+    def switch_sriov(
+        self, enable: bool, wait: bool = True, reset_connections: bool = True
+    ) -> None:
         azure_platform: AzurePlatform = self._platform  # type: ignore
         network_client = get_network_client(azure_platform)
         vm = get_vm(azure_platform, self._node)
@@ -639,7 +641,7 @@ class NetworkInterface(AzureFeatureMixin, features.NetworkInterface):
 
         # wait settings effective
         if wait:
-            self._check_sriov_enabled(enable)
+            self._check_sriov_enabled(enable, reset_connections)
 
     def is_enabled_sriov(self) -> bool:
         azure_platform: AzurePlatform = self._platform  # type: ignore
@@ -774,7 +776,10 @@ class NetworkInterface(AzureFeatureMixin, features.NetworkInterface):
         modprobe_tool.reload(["hv_netvsc"])
 
     @retry(tries=60, delay=10)
-    def _check_sriov_enabled(self, enabled: bool) -> None:
+    def _check_sriov_enabled(
+        self, enabled: bool, reset_connections: bool = True
+    ) -> None:
+        if reset_connections:
             self._node.close()
         self._node.nics.reload()
         default_nic = self._node.nics.get_nic_by_index(0)

microsoft/testsuites/dpdk/dpdksuite.py

@@ -665,4 +665,5 @@ class Dpdk(TestSuite):
             if modprobe.module_exists("uio_hv_generic"):
                 node.tools[Service].stop_service("vpp")
                 modprobe.remove(["uio_hv_generic"])
+                node.close()
                 modprobe.reload(["hv_netvsc"])

microsoft/testsuites/dpdk/dpdkutil.py

@@ -333,14 +333,18 @@ def run_testpmd_concurrent(
 
         # disable sriov (and wait for change to apply)
         for node_resources in [x for x in test_kits if x.switch_sriov]:
-            node_resources.nic_controller.switch_sriov(enable=False, wait=True)
+            node_resources.nic_controller.switch_sriov(
+                enable=False, wait=True, reset_connections=False
+            )
 
         # let run for a bit with SRIOV disabled
         time.sleep(10)
 
         # re-enable sriov
         for node_resources in [x for x in test_kits if x.switch_sriov]:
-            node_resources.nic_controller.switch_sriov(enable=True, wait=True)
+            node_resources.nic_controller.switch_sriov(
+                enable=True, wait=True, reset_connections=False
+            )
 
         # run for a bit with SRIOV re-enabled
         time.sleep(10)