Merge pull request #109 from almaslennikov/bf3-reset

fix: consider config reset successful after reboot

Author: almaslennikov

internal/controller/nicdevice_controller.go

@@ -533,6 +533,14 @@ func (r *NicDeviceReconciler) handleConfigurationSpecValidation(ctx context.Cont
 				return nil
 			}
 
+			// If configuration reset was applied (nvConfigUpdateRequired == false) and reboot has happened (uptime > sinceStatusUpdate),
+			// there might still be discrepancies in the mlxconfig query that might lead to a boot loop.
+			// In this case, consider the reset successful
+			if status.device.Spec.Configuration.ResetToDefault {
+				status.rebootRequired = false
+				return nil
+			}
+
 			log.Log.Info("nv config failed to update after reboot for device", "device", status.device.Name)
 			r.EventRecorder.Event(status.device, v1.EventTypeWarning, consts.FirmwareError, consts.FwConfigNotAppliedAfterRebootErrorMsg)
 			err = r.updateConfigInProgressStatusCondition(ctx, status.device, consts.FirmwareError, metav1.ConditionFalse, consts.FwConfigNotAppliedAfterRebootErrorMsg)

pkg/host/configvalidation.go

@@ -219,6 +219,10 @@ func (v *configValidationImpl) ValidateResetToDefault(nvConfig types.NvConfigQue
 	// ResetToDefault requires us to set ADVANCED_PCI_SETTINGS=true, which is not a default value
 	// Deleting this key from maps so that it doesn't interfere with comparisons
 	delete(nvConfig.DefaultConfig, consts.AdvancedPCISettingsParam)
+	// We want to retain the BF3 operation mode after reset, so not taking it into consideration
+	delete(nvConfig.DefaultConfig, consts.BF3OperationModeParam)
+	delete(nvConfig.CurrentConfig, consts.BF3OperationModeParam)
+	delete(nvConfig.NextBootConfig, consts.BF3OperationModeParam)
 
 	alreadyResetInCurrent := false
 	willResetInNextBoot := false