Add wait for lag to netplanner rebind

netplanner/sriov/__main__.py

@@ -21,6 +21,7 @@
 import logging
 from pathlib import Path
 import subprocess
+import time
 
 from jinja2 import Environment
 
@@ -115,5 +116,10 @@ def config(configuration: NetplannerConfig, queue_rebind: bool = False):
 def rebind(pci_addresses: list[str]):
     for pci_address in pci_addresses:
         device = pci.PCIDevice(pci_address)
+        # We always wait for link_aggregation_state to become active because
+        # delayed rebind is only used for this very specific case anyway
+        while device.link_aggregation_state != "active":
+            print("Waiting for lag state of device {} to become active".format(pci_address))
+            time.sleep(1)
         device.sriov_drivers_autoprobe = True
         pci.bind_vfs(device.vfs)

netplanner/sriov/pci.py

@@ -185,6 +185,12 @@ def sriov_drivers_autoprobe(self, value: bool):
         with open(sriov_drivers_autoprobe_file, "w") as f:
             f.write("1" if value else "0")
 
+    @property
+    def link_aggregation_state(self) -> str:
+        link_aggregation_file = "/sys/kernel/debug/mlx5/{}/lag/state".format(self.pci_addr)
+        with open(link_aggregation_file, "r") as f:
+            return f.read().strip()
+
     def __str__(self) -> str:
         """String represenation of object