fix(modelregistry): address audit findings

Replace try-lock/sleep scheme with blocking locks, add ErrModelNotFound/
ErrModelExists/ErrInvalidModel sentinels, make RegisterModelWithRules
atomic, snapshot in IterateModels, guard defaultRegistry access, cap the
pointer-unwrap depth, and recover panics in callbacks and reflection.
Security hooks now allow-by-default only on ErrModelNotFound. Add tests.

pkg/modelregistry/model_registry.go

@@ -1,10 +1,12 @@
 package modelregistry
 
 import (
+	"errors"
 	"fmt"
 	"reflect"
 	"sync"
-	"time"
+
+	"github.com/bitechdev/ResolveSpec/pkg/logger"
 )
 
 // ModelRules defines the permissions and security settings for a model
@@ -52,6 +54,21 @@ var defaultRegistry = &DefaultModelRegistry{
 var registries = []*DefaultModelRegistry{defaultRegistry}
 var registriesMutex sync.RWMutex
 
+// Sentinel errors so callers (notably the security layer) can distinguish
+// "not registered" from every other failure with errors.Is.
+var (
+	ErrModelNotFound = errors.New("model not found")
+	ErrModelExists   = errors.New("model already registered")
+	ErrInvalidModel  = errors.New("invalid model")
+)
+
+// maxUnwrapDepth bounds pointer/slice/array unwrapping so a recursive type
+// (type T *T) cannot spin forever.
+const maxUnwrapDepth = 16
+
+// Lock ordering: registriesMutex is always taken before a registry's mutex,
+// never the reverse. No caller-supplied code ever runs while a lock is held.
+
 // NewModelRegistry creates a new model registry
 func NewModelRegistry() *DefaultModelRegistry {
 	return &DefaultModelRegistry{
@@ -60,44 +77,19 @@ func NewModelRegistry() *DefaultModelRegistry {
 	}
 }
 
-// lockRetryAttempts/lockRetryDelay bound how long the try-lock helpers below
-// will spin before giving up, so a contended registriesMutex can never hang
-// a caller of GetDefaultRegistry/SetDefaultRegistry.
-const (
-	lockRetryAttempts = 20
-	lockRetryDelay    = 1 * time.Millisecond
-)
-
-// GetDefaultRegistry returns the current default registry. It uses a
-// bounded TryRLock instead of a blocking RLock so it can never hang;
-// if the lock can't be acquired in time it falls back to the last known
-// value without synchronization.
+// GetDefaultRegistry returns the current default registry.
 func GetDefaultRegistry() *DefaultModelRegistry {
-	for i := 0; i < lockRetryAttempts; i++ {
-		if registriesMutex.TryRLock() {
-			defer registriesMutex.RUnlock()
-			return defaultRegistry
-		}
-		time.Sleep(lockRetryDelay)
-	}
+	registriesMutex.RLock()
+	defer registriesMutex.RUnlock()
 	return defaultRegistry
 }
 
-// SetDefaultRegistry replaces the default registry. It uses a bounded
-// TryLock instead of a blocking Lock so it can never hang; if the lock
-// can't be acquired in time the call is a no-op.
+// SetDefaultRegistry replaces the default registry. A nil registry is ignored.
 func SetDefaultRegistry(registry *DefaultModelRegistry) {
-	acquired := false
-	for i := 0; i < lockRetryAttempts; i++ {
-		if registriesMutex.TryLock() {
-			acquired = true
-			break
-		}
-		time.Sleep(lockRetryDelay)
-	}
-	if !acquired {
+	if registry == nil {
 		return
 	}
+	registriesMutex.Lock()
 	defer registriesMutex.Unlock()
 
 	foundAt := -1
@@ -123,99 +115,96 @@ func AddRegistry(registry *DefaultModelRegistry) {
 	registries = append(registries, registry)
 }
 
-// tryLock attempts to acquire the registry's write lock, retrying briefly.
-// Returns false if it could not be acquired within the bound.
-func (r *DefaultModelRegistry) tryLock() bool {
-	for i := 0; i < lockRetryAttempts; i++ {
-		if r.mutex.TryLock() {
-			return true
-		}
-		time.Sleep(lockRetryDelay)
-	}
-	return false
+// registriesSnapshot returns a copy of the registry list so callers can
+// iterate without holding registriesMutex.
+func registriesSnapshot() []*DefaultModelRegistry {
+	registriesMutex.RLock()
+	defer registriesMutex.RUnlock()
+	return append([]*DefaultModelRegistry(nil), registries...)
 }
 
-// tryRLock attempts to acquire the registry's read lock, retrying briefly.
-// Returns false if it could not be acquired within the bound.
-func (r *DefaultModelRegistry) tryRLock() bool {
-	for i := 0; i < lockRetryAttempts; i++ {
-		if r.mutex.TryRLock() {
-			return true
+// validateModel checks the model is a struct (or pointer/slice/array of one)
+// and returns the normalised non-pointer struct value. It takes no locks.
+func validateModel(model interface{}) (result interface{}, err error) {
+	// Reflection on a pathological type must fail the registration, not crash the process.
+	defer func() {
+		if r := recover(); r != nil {
+			result = nil
+			err = fmt.Errorf("%w: %v", ErrInvalidModel, logger.HandlePanic("modelregistry.validateModel", r))
 		}
-		time.Sleep(lockRetryDelay)
-	}
-	return false
-}
+	}()
 
-func (r *DefaultModelRegistry) RegisterModel(name string, model interface{}) error {
-	if !r.tryLock() {
-		return fmt.Errorf("failed to register model %s: registry locked", name)
-	}
-	defer r.mutex.Unlock()
-
-	if _, exists := r.models[name]; exists {
-		return fmt.Errorf("model %s already registered", name)
-	}
-
-	// Validate that model is a non-pointer struct
 	modelType := reflect.TypeOf(model)
 	if modelType == nil {
-		return fmt.Errorf("model cannot be nil")
+		return nil, fmt.Errorf("%w: model cannot be nil", ErrInvalidModel)
 	}
 
 	originalType := modelType
 
 	// Unwrap pointers, slices, and arrays to check the underlying type
-	for modelType.Kind() == reflect.Pointer || modelType.Kind() == reflect.Slice || modelType.Kind() == reflect.Array {
+	for depth := 0; modelType.Kind() == reflect.Pointer || modelType.Kind() == reflect.Slice || modelType.Kind() == reflect.Array; depth++ {
+		if depth >= maxUnwrapDepth {
+			return nil, fmt.Errorf("%w: type %s nests deeper than %d levels", ErrInvalidModel, originalType.String(), maxUnwrapDepth)
+		}
 		modelType = modelType.Elem()
 	}
 
-	// Validate that the underlying type is a struct
 	if modelType.Kind() != reflect.Struct {
-		return fmt.Errorf("model must be a struct or pointer to struct, got %s", originalType.String())
+		return nil, fmt.Errorf("%w: model must be a struct or pointer to struct, got %s", ErrInvalidModel, originalType.String())
 	}
 
 	// If a pointer/slice/array was passed, unwrap to the base struct
 	if originalType != modelType {
-		// Create a zero value of the struct type
 		model = reflect.New(modelType).Elem().Interface()
 	}
 
-	// Additional check: ensure model is not a pointer
-	finalType := reflect.TypeOf(model)
-	if finalType.Kind() == reflect.Pointer {
-		return fmt.Errorf("model must be a non-pointer struct, got pointer to %s. Use MyModel{} instead of &MyModel{}", finalType.Elem().Name())
+	if finalType := reflect.TypeOf(model); finalType.Kind() == reflect.Pointer {
+		return nil, fmt.Errorf("%w: model must be a non-pointer struct, got pointer to %s. Use MyModel{} instead of &MyModel{}", ErrInvalidModel, finalType.Elem().Name())
+	}
+	return model, nil
+}
+
+// registerLocked validates the model outside the lock, then writes the model
+// and its rules under a single lock acquisition so no reader can observe the
+// model without its final rules.
+func (r *DefaultModelRegistry) registerLocked(name string, model interface{}, rules ModelRules) error {
+	model, err := validateModel(model)
+	if err != nil {
+		return err
 	}
 
-	r.models[name] = model
-	// Initialize with default rules if not already set
-	if _, exists := r.rules[name]; !exists {
-		r.rules[name] = DefaultModelRules()
+	r.mutex.Lock()
+	defer r.mutex.Unlock()
+
+	if _, exists := r.models[name]; exists {
+		return fmt.Errorf("%w: %s", ErrModelExists, name)
 	}
+	r.models[name] = model
+	r.rules[name] = rules
 	return nil
 }
 
+func (r *DefaultModelRegistry) RegisterModel(name string, model interface{}) error {
+	return r.registerLocked(name, model, DefaultModelRules())
+}
+
 func (r *DefaultModelRegistry) GetModel(name string) (interface{}, error) {
-	if !r.tryRLock() {
-		return nil, fmt.Errorf("failed to get model %s: registry locked", name)
-	}
+	r.mutex.RLock()
 	defer r.mutex.RUnlock()
 
 	model, exists := r.models[name]
 	if !exists {
-		return nil, fmt.Errorf("model %s not found", name)
+		return nil, fmt.Errorf("%w: %s", ErrModelNotFound, name)
 	}
 
 	return model, nil
 }
 
 func (r *DefaultModelRegistry) GetAllModels() map[string]interface{} {
-	if !r.tryRLock() {
-		return make(map[string]interface{})
-	}
+	r.mutex.RLock()
 	defer r.mutex.RUnlock()
 
-	result := make(map[string]interface{})
+	result := make(map[string]interface{}, len(r.models))
 	for k, v := range r.models {
 		result[k] = v
 	}
@@ -225,9 +214,13 @@ func (r *DefaultModelRegistry) GetAllModels() map[string]interface{} {
 func (r *DefaultModelRegistry) GetModelByEntity(schema, entity string) (interface{}, error) {
 	// Try full name first
 	fullName := fmt.Sprintf("%s.%s", schema, entity)
-	if model, err := r.GetModel(fullName); err == nil {
+	model, err := r.GetModel(fullName)
+	if err == nil {
 		return model, nil
 	}
+	if !errors.Is(err, ErrModelNotFound) {
+		return nil, err
+	}
 
 	// Fallback to entity name only
 	return r.GetModel(entity)
@@ -238,9 +231,8 @@ func (r *DefaultModelRegistry) SetModelRules(name string, rules ModelRules) erro
 	r.mutex.Lock()
 	defer r.mutex.Unlock()
 
-	// Check if model exists
 	if _, exists := r.models[name]; !exists {
-		return fmt.Errorf("model %s not found", name)
+		return fmt.Errorf("%w: %s", ErrModelNotFound, name)
 	}
 
 	r.rules[name] = rules
@@ -253,12 +245,10 @@ func (r *DefaultModelRegistry) GetModelRules(name string) (ModelRules, error) {
 	r.mutex.RLock()
 	defer r.mutex.RUnlock()
 
-	// Check if model exists
 	if _, exists := r.models[name]; !exists {
-		return ModelRules{}, fmt.Errorf("model %s not found", name)
+		return ModelRules{}, fmt.Errorf("%w: %s", ErrModelNotFound, name)
 	}
 
-	// Return rules if set, otherwise return default rules
 	if rules, exists := r.rules[name]; exists {
 		return rules, nil
 	}
@@ -266,84 +256,62 @@ func (r *DefaultModelRegistry) GetModelRules(name string) (ModelRules, error) {
 	return DefaultModelRules(), nil
 }
 
-// RegisterModelWithRules registers a model with specific rules
+// RegisterModelWithRules registers a model with specific rules atomically
 func (r *DefaultModelRegistry) RegisterModelWithRules(name string, model interface{}, rules ModelRules) error {
-	// First register the model
-	if err := r.RegisterModel(name, model); err != nil {
-		return err
-	}
-
-	// Then set the rules (we need to lock again for rules)
-	r.mutex.Lock()
-	defer r.mutex.Unlock()
-	r.rules[name] = rules
-
-	return nil
+	return r.registerLocked(name, model, rules)
 }
 
 // Global convenience functions using the default registry
 
 // RegisterModel registers a model with the default global registry
 func RegisterModel(model interface{}, name string) error {
-	return defaultRegistry.RegisterModel(name, model)
+	return GetDefaultRegistry().RegisterModel(name, model)
 }
 
 // GetModelByName retrieves a model by searching through all registries in order
 // Returns the first match found
 func GetModelByName(name string) (interface{}, error) {
-	registriesMutex.RLock()
-	defer registriesMutex.RUnlock()
-
-	for _, registry := range registries {
+	for _, registry := range registriesSnapshot() {
 		if model, err := registry.GetModel(name); err == nil {
 			return model, nil
 		}
 	}
 
-	return nil, fmt.Errorf("model %s not found in any registry", name)
+	return nil, fmt.Errorf("%w: %s (in any registry)", ErrModelNotFound, name)
 }
 
-// IterateModels iterates over all models in the default global registry
+// IterateModels iterates over all models in the default global registry.
+// It iterates over a snapshot, so fn may safely call back into the registry.
+// A panic in fn is recovered and logged with the model name, and iteration
+// continues with the remaining models.
 func IterateModels(fn func(name string, model interface{})) {
-	defaultRegistry.mutex.RLock()
-	defer defaultRegistry.mutex.RUnlock()
-
-	for name, model := range defaultRegistry.models {
-		fn(name, model)
+	for name, model := range GetDefaultRegistry().GetAllModels() {
+		callIsolated(name, model, fn)
 	}
 }
 
-// GetModels returns a list of all models from all registries
-// Models are collected in registry order, with duplicates included
-func GetModels() []interface{} {
-	acquired := false
-	for i := 0; i < lockRetryAttempts; i++ {
-		if registriesMutex.TryRLock() {
-			acquired = true
-			break
+func callIsolated(name string, model interface{}, fn func(name string, model interface{})) {
+	defer func() {
+		if r := recover(); r != nil {
+			_ = logger.HandlePanic("modelregistry.IterateModels", r, "model", name)
 		}
-		time.Sleep(lockRetryDelay)
-	}
-	if !acquired {
-		return nil
-	}
-	defer registriesMutex.RUnlock()
+	}()
+	fn(name, model)
+}
 
+// GetModels returns a list of all models from all registries.
+// Only the first occurrence of each model name is included.
+func GetModels() []interface{} {
 	var models []interface{}
 	seen := make(map[string]bool)
 
-	for _, registry := range registries {
-		if !registry.tryRLock() {
-			continue
-		}
-		for name, model := range registry.models {
-			// Only add the first occurrence of each model name
+	for _, registry := range registriesSnapshot() {
+		for name, model := range registry.GetAllModels() {
 			if !seen[name] {
 				models = append(models, model)
 				seen[name] = true
 			}
 		}
-		registry.mutex.RUnlock()
 	}
 
 	return models
@@ -351,31 +319,31 @@ func GetModels() []interface{} {
 
 // SetModelRules sets the rules for a specific model in the default registry
 func SetModelRules(name string, rules ModelRules) error {
-	return defaultRegistry.SetModelRules(name, rules)
+	return GetDefaultRegistry().SetModelRules(name, rules)
 }
 
 // GetModelRules retrieves the rules for a specific model from the default registry
 func GetModelRules(name string) (ModelRules, error) {
-	return defaultRegistry.GetModelRules(name)
+	return GetDefaultRegistry().GetModelRules(name)
 }
 
 // GetModelRulesByName retrieves the rules for a model by searching through all registries in order
-// Returns the first match found
+// Returns the first match found. The error wraps ErrModelNotFound when no registry has the model.
 func GetModelRulesByName(name string) (ModelRules, error) {
-	registriesMutex.RLock()
-	defer registriesMutex.RUnlock()
-
-	for _, registry := range registries {
-		if _, err := registry.GetModel(name); err == nil {
-			// Model found in this registry, get its rules
-			return registry.GetModelRules(name)
+	for _, registry := range registriesSnapshot() {
+		rules, err := registry.GetModelRules(name)
+		if err == nil {
+			return rules, nil
+		}
+		if !errors.Is(err, ErrModelNotFound) {
+			return ModelRules{}, err
 		}
 	}
 
-	return ModelRules{}, fmt.Errorf("model %s not found in any registry", name)
+	return ModelRules{}, fmt.Errorf("%w: %s (in any registry)", ErrModelNotFound, name)
 }
 
 // RegisterModelWithRules registers a model with specific rules in the default registry
 func RegisterModelWithRules(model interface{}, name string, rules ModelRules) error {
-	return defaultRegistry.RegisterModelWithRules(name, model, rules)
+	return GetDefaultRegistry().RegisterModelWithRules(name, model, rules)
 }