Set device to maintenance mode if TPM error is detected

We have seen errors like "point is not on the required curve" in the
past that prevents the device from upgrading because TPM fails to perform
an ECDH key agreement to generate the KEK and decrypt the vault key.
To catch these errors early, check if TPM is not functioning properly on
a critical path and set the device to maintenance mode.

Signed-off-by: Shahriyar Jalayeri <shahriyar@zededa.com>

pkg/pillar/cmd/vaultmgr/vaultmgr.go

@@ -512,13 +512,21 @@ func publishVaultKey(ctx *vaultMgrContext, vaultName string) error {
 		}
 		keyBytes, err := etpm.FetchSealedVaultKey(log)
 		if err != nil {
-			return fmt.Errorf("Failed to retrieve key from TPM %w", err)
+			return fmt.Errorf("failed to retrieve key from TPM %w", err)
 		}
 
 		encryptedKey, err := etpm.EncryptDecryptUsingTpm(keyBytes, true)
 		if err != nil {
-			// XXX should we still send with no key?
-			return fmt.Errorf("Failed to encrypt vault key %w", err)
+			// if we can't encrypt the vault key using TPM, possibly something
+			// is horribly wrong with TPM, this can bite us later for example
+			// when we try to upgrade the device, so better to fail early.
+			key := types.EncryptedVaultKeyFromDevice{
+				Name:     vaultName,
+				TpmError: err}
+			ctx.pubVaultKeyFromDevice.Publish(key.Key(), key)
+
+			// still return error, this will get logged
+			return fmt.Errorf("failed to encrypt vault key %w", err)
 		}
 
 		hash := sha256.New()
@@ -531,7 +539,7 @@ func publishVaultKey(ctx *vaultMgrContext, vaultName string) error {
 		}
 		encryptedVaultKey, err = proto.Marshal(keyData)
 		if err != nil {
-			return fmt.Errorf("Failed to Marshal keyData %w", err)
+			return fmt.Errorf("failed to Marshal keyData %w", err)
 		}
 	}
 

pkg/pillar/cmd/zedagent/attesttask.go

@@ -791,6 +791,13 @@ func handleEncryptedKeyFromDeviceImpl(ctxArg interface{}, key string,
 	log.Noticef("handleEncryptedKeyFromDeviceImpl len %d",
 		len(vaultKey.EncryptedVaultKey))
 
+	if vaultKey.TpmError != nil {
+		ctx.maintenanceMode = true
+		ctx.maintModeReason = types.MaintenanceModeReasonTpmFailure
+		log.Errorf("setting maintenance mode due to TPM failurec: %v", vaultKey.TpmError)
+		return
+	}
+
 	if ctx.attestCtx == nil {
 		log.Fatalf("[ATTEST] Uninitialized access to attestCtx")
 	}

pkg/pillar/types/vaultmgrtypes.go

@@ -86,6 +86,7 @@ type EncryptedVaultKeyFromDevice struct {
 	Name              string
 	EncryptedVaultKey []byte // empty if no TPM enabled
 	IsTpmEnabled      bool
+	TpmError          error // only set if there is unrecoverable error, for example TPM is kaputt.
 }
 
 // Key returns name of the vault corresponding to this object

pkg/pillar/types/zedagenttypes.go

@@ -395,6 +395,7 @@ const (
 	MaintenanceModeReasonUserRequested
 	MaintenanceModeReasonVaultLockedUp
 	MaintenanceModeReasonNoDiskSpace
+	MaintenanceModeReasonTpmFailure
 )
 
 // String returns the verbose equivalent of MaintenanceModeReason code
@@ -408,6 +409,8 @@ func (mmr MaintenanceModeReason) String() string {
 		return "MaintenanceModeReasonVaultLockedUp"
 	case MaintenanceModeReasonNoDiskSpace:
 		return "MaintenanceModeReasonNoDiskSpace"
+	case MaintenanceModeReasonTpmFailure:
+		return "MaintenanceModeReasonTpmFailure"
 	default:
 		return fmt.Sprintf("Unknown MaintenanceModeReason %d", mmr)
 	}