fix: don't validate machine.install if installed

As Talos doesn't consume `.machine.install` if already installed, there
is no point in validating it once already installed.

This fixes a problem users often run into: after a reboot/upgrade the
system disk blockdevice name changes, due to the kernel upgrade, or just
unpredictable behavior of device discovery. Talos fails to boot as it
can't validate the machine config, while it's already installed, so
actual blockdevice name doesn't matter.

Signed-off-by: Andrey Smirnov <andrey.smirnov@siderolabs.com>

internal/app/machined/internal/server/v1alpha1/v1alpha1_server.go

@@ -152,6 +152,16 @@ func (s *Server) Register(obj *grpc.Server) {
 	timeapi.RegisterTimeServiceServer(obj, &TimeServer{ConfigProvider: s.Controller.Runtime()})
 }
 
+// modeWrapper overrides RequiresInstall() based on actual installed status.
+type modeWrapper struct {
+	runtime.Mode
+	installed bool
+}
+
+func (m modeWrapper) RequiresInstall() bool {
+	return m.Mode.RequiresInstall() && !m.installed
+}
+
 // ApplyConfiguration implements machine.MachineService.
 //
 //nolint:gocyclo,cyclop
@@ -169,7 +179,12 @@ func (s *Server) ApplyConfiguration(ctx context.Context, in *machine.ApplyConfig
 		return nil, status.Error(codes.InvalidArgument, err.Error())
 	}
 
-	warnings, err := cfgProvider.Validate(s.Controller.Runtime().State().Platform().Mode())
+	warnings, err := cfgProvider.Validate(
+		modeWrapper{
+			Mode:      s.Controller.Runtime().State().Platform().Mode(),
+			installed: s.Controller.Runtime().State().Machine().Installed(),
+		},
+	)
 	if err != nil {
 		return nil, status.Error(codes.InvalidArgument, err.Error())
 	}

internal/app/machined/pkg/controllers/config/acquire.go

@@ -206,6 +206,19 @@ func (ctrl *AcquireController) stateDisk(ctx context.Context, r controller.Runti
 	}
 }
 
+// validationModeDiskConfig is a "fake" validation mode for config loaded from disk.
+type validationModeDiskConfig struct{}
+
+// RequiresInstall implements validation.RuntimeMode interface.
+func (validationModeDiskConfig) RequiresInstall() bool {
+	return false
+}
+
+// String implements validation.RuntimeMode interface.
+func (validationModeDiskConfig) String() string {
+	return "diskConfig"
+}
+
 // loadFromDisk is a helper function for stateDisk.
 func (ctrl *AcquireController) loadFromDisk(logger *zap.Logger) (config.Provider, error) {
 	logger.Debug("loading config from STATE", zap.String("path", ctrl.ConfigPath))
@@ -232,7 +245,8 @@ func (ctrl *AcquireController) loadFromDisk(logger *zap.Logger) (config.Provider
 		return nil, nil
 	}
 
-	warnings, err := cfg.Validate(ctrl.ValidationMode)
+	// if the STATE partition is present & contains machine config, Talos is already installed
+	warnings, err := cfg.Validate(validationModeDiskConfig{})
 	if err != nil {
 		return nil, fmt.Errorf("failed to validate on-disk config: %w", err)
 	}

internal/integration/api/apply-config.go

@@ -383,7 +383,7 @@ func (suite *ApplyConfigSuite) TestApplyNoReboot() {
 	nodeCtx := client.WithNodes(suite.ctx, node)
 
 	provider, err := suite.ReadConfigFromNode(nodeCtx)
-	suite.Assert().Nilf(err, "failed to read existing config from node %q: %w", node, err)
+	suite.Require().Nilf(err, "failed to read existing config from node %q: %s", node, err)
 
 	cfg := provider.RawV1Alpha1()
 	suite.Require().NotNil(cfg)
@@ -395,7 +395,7 @@ func (suite *ApplyConfigSuite) TestApplyNoReboot() {
 	suite.Require().NoError(err)
 
 	cfgDataOut, err := provider.Bytes()
-	suite.Assert().Nilf(err, "failed to marshal updated machine config data (node %q): %w", node, err)
+	suite.Require().Nilf(err, "failed to marshal updated machine config data (node %q): %s", node, err)
 
 	_, err = suite.Client.ApplyConfiguration(
 		nodeCtx, &machineapi.ApplyConfigurationRequest{
@@ -429,7 +429,7 @@ func (suite *ApplyConfigSuite) TestApplyDryRun() {
 	nodeCtx := client.WithNodes(suite.ctx, node)
 
 	provider, err := suite.ReadConfigFromNode(nodeCtx)
-	suite.Assert().Nilf(err, "failed to read existing config from node %q: %w", node, err)
+	suite.Require().Nilf(err, "failed to read existing config from node %q: %s", node, err)
 
 	cfg := provider.RawV1Alpha1()
 	suite.Require().NotNil(cfg)
@@ -441,7 +441,7 @@ func (suite *ApplyConfigSuite) TestApplyDryRun() {
 	suite.Require().NoError(err)
 
 	cfgDataOut, err := provider.Bytes()
-	suite.Assert().Nilf(err, "failed to marshal updated machine config data (node %q): %w", node, err)
+	suite.Require().Nilf(err, "failed to marshal updated machine config data (node %q): %s", node, err)
 
 	reply, err := suite.Client.ApplyConfiguration(
 		nodeCtx, &machineapi.ApplyConfigurationRequest{
@@ -451,8 +451,8 @@ func (suite *ApplyConfigSuite) TestApplyDryRun() {
 		},
 	)
 
-	suite.Assert().Nilf(err, "failed to apply configuration (node %q): %w", node, err)
-	suite.Require().Contains(reply.Messages[0].ModeDetails, "Dry run summary")
+	suite.Require().Nilf(err, "failed to apply configuration (node %q): %s", node, err)
+	suite.Assert().Contains(reply.Messages[0].ModeDetails, "Dry run summary")
 }
 
 // TestApplyTry applies the config in try mode with a short timeout.
@@ -496,7 +496,7 @@ func (suite *ApplyConfigSuite) TestApplyTry() {
 	)
 
 	cfgDataOut, err := container.NewV1Alpha1(cfg).Bytes()
-	suite.Assert().Nilf(err, "failed to marshal updated machine config data (node %q): %s", node, err)
+	suite.Require().Nilf(err, "failed to marshal updated machine config data (node %q): %s", node, err)
 
 	_, err = suite.Client.ApplyConfiguration(
 		nodeCtx, &machineapi.ApplyConfigurationRequest{
@@ -508,7 +508,7 @@ func (suite *ApplyConfigSuite) TestApplyTry() {
 	suite.Assert().Nilf(err, "failed to apply configuration (node %q): %s", node, err)
 
 	provider, err = getMachineConfig(nodeCtx)
-	suite.Assert().Nilf(err, "failed to read existing config from node %q: %w", node, err)
+	suite.Require().Nilf(err, "failed to read existing config from node %q: %w", node, err)
 
 	suite.Assert().NotNil(provider.Config().Machine().Network())
 	suite.Assert().NotNil(provider.Config().Machine().Network().Devices())