Return error on mlx plugin on firmware reset check

This commit makes the functions to return error if we are not able to
read from the configuration cache on the node.

Also this commit fix a bug when checking if a MLX card is controlled by
the operator

Signed-off-by: Sebastian Sch <sebassch@gmail.com>

pkg/plugins/mellanox/mellanox_plugin.go

@@ -2,13 +2,14 @@ package mellanox
 
 import (
 	"fmt"
+	"github.com/k8snetworkplumbingwg/sriov-network-operator/pkg/consts"
+	mlx "github.com/k8snetworkplumbingwg/sriov-network-operator/pkg/vendors/mellanox"
 
 	"sigs.k8s.io/controller-runtime/pkg/log"
 
 	sriovnetworkv1 "github.com/k8snetworkplumbingwg/sriov-network-operator/api/v1"
 	"github.com/k8snetworkplumbingwg/sriov-network-operator/pkg/helper"
 	plugin "github.com/k8snetworkplumbingwg/sriov-network-operator/pkg/plugins"
-	mlx "github.com/k8snetworkplumbingwg/sriov-network-operator/pkg/vendors/mellanox"
 )
 
 var PluginName = "mellanox"
@@ -144,12 +145,24 @@ func (p *MellanoxPlugin) OnNodeStateChange(new *sriovnetworkv1.SriovNetworkNodeS
 		pciAddress := pciPrefix + "0"
 
 		// Skip devices not configured by the operator
-		if p.nicNotConfiguredByOperator(portsMap) {
+		isConfigured, err := p.nicConfiguredByOperator(portsMap)
+		if err != nil {
+			return false, false, err
+		}
+		if !isConfigured {
+			log.Log.V(2).Info("None of the ports are configured by the operator skipping firmware reset",
+				"portMap", portsMap)
 			continue
 		}
 
 		// Skip externally managed NICs
-		if p.nicHasExternallyManagedPFs(portsMap) {
+		hasExternally, err := p.nicHasExternallyManagedPFs(portsMap)
+		if err != nil {
+			return false, false, err
+		}
+		if hasExternally {
+			log.Log.V(2).Info("One of the ports is configured as externally managed skipping firmware reset",
+				"portMap", portsMap)
 			continue
 		}
 
@@ -194,36 +207,38 @@ func (p *MellanoxPlugin) Apply() error {
 
 // nicHasExternallyManagedPFs returns true if one of the ports(interface) of the NIC is marked as externally managed
 // in StoreManagerInterface.
-func (p *MellanoxPlugin) nicHasExternallyManagedPFs(nicPortsMap map[string]sriovnetworkv1.InterfaceExt) bool {
+func (p *MellanoxPlugin) nicHasExternallyManagedPFs(nicPortsMap map[string]sriovnetworkv1.InterfaceExt) (bool, error) {
 	for _, iface := range nicPortsMap {
 		pfStatus, exist, err := p.helpers.LoadPfsStatus(iface.PciAddress)
 		if err != nil {
-			log.Log.Error(err, "failed to load PF status from disk", "address", iface.PciAddress)
-			continue
+			log.Log.Error(err, "This should not happen, to overcome config daemon stuck, please remove the PCI file on the host under the operator configuration path",
+				"path", consts.PfAppliedConfig, "pciAddress", iface.PciAddress)
+			return false, err
 		}
 		if !exist {
 			continue
 		}
 		if pfStatus.ExternallyManaged {
 			log.Log.V(2).Info("PF is extenally managed, skip FW TotalVfs reset")
-			return true
+			return true, nil
 		}
 	}
-	return false
+	return false, nil
 }
 
-// nicNotConfiguredByOperator returns true if one of the ports(interface) of the NIC is not configured by operator
-func (p *MellanoxPlugin) nicNotConfiguredByOperator(nicPortsMap map[string]sriovnetworkv1.InterfaceExt) bool {
+// nicConfiguredByOperator returns true if one of the ports(interface) of the NIC is configured by operator
+func (p *MellanoxPlugin) nicConfiguredByOperator(nicPortsMap map[string]sriovnetworkv1.InterfaceExt) (bool, error) {
 	for _, iface := range nicPortsMap {
 		_, exist, err := p.helpers.LoadPfsStatus(iface.PciAddress)
 		if err != nil {
 			log.Log.Error(err, "failed to load PF status from disk", "address", iface.PciAddress)
-			continue
+			return false, err
 		}
 		if exist {
 			log.Log.V(2).Info("PF configured by the operator", "interface", iface)
-			return true
+			return true, nil
 		}
 	}
-	return false
+
+	return false, nil
 }