fix(config): lock Manager, harden defaults handling and path/IP helpers

Guard viper with an RWMutex, make the singleton race-free and stop
NewManager replacing the global (add SetConfigManager), write saved
configs 0600, search CWD last, add ConfigFileUsed and Config.Validate,
nil-safe PathsConfig.Set, confine PathsConfig.Join, bound GetIPs DNS
lookup, and drop dead Unmarshal in SetConfig. Mark audit status.

audit/pkg/config.audit.md

@@ -24,7 +24,7 @@ validators. Two serious issues:
 |---|----------|------|---------|
 | 1 | **Critical** | Locking | `Manager.Set`/`Get` over a lock-free `*viper.Viper` → concurrent map write → process-fatal |
 | 2 | **High** | Locking | `GetConfigManager()` is an unsynchronised lazy singleton; `NewManager()` also clobbers the global as a side effect |
-| 3 | **High** | Security | Insecure defaults: `cors.allowed_origins: ["*"]`, `allowed_headers: ["*"]`, `sslmode: disable`, `user: postgres` + blank password |
+| 3 | **High** | Security | **OPEN (deferred)** Insecure defaults: `cors.allowed_origins: ["*"]`, `allowed_headers: ["*"]`, `sslmode: disable`, `user: postgres` + blank password |
 | 4 | **High** | Security | `SaveConfig` writes all secrets in plaintext at mode `0644` (viper default, never overridden) |
 | 5 | Medium | Security | `AddConfigPath(".")` is searched first — CWD config injection |
 | 6 | Medium | Observability | `Load()` swallows `ConfigFileNotFoundError` with no log at all |
@@ -37,6 +37,24 @@ validators. Two serious issues:
 | 13 | Low | Correctness | `ServersConfig.GetDefault()` returns a pointer to a copy of a map value |
 | 14 | Low | Security | `PathsConfig.Join` does not confine the result to the base path |
 
+## Resolution status (2026-09-30)
+
+- **#1** — Fixed: `sync.RWMutex` guards every viper access, options included
+- **#2** — Fixed: mutex-guarded singleton; `NewManager` no longer touches the global (new `SetConfigManager` publishes explicitly)
+- **#4** — Fixed: `SetConfigPermissions(0o600)` plus `chmod 0600` after write (secrets are not stripped)
+- **#5** — Fixed: search order is `/etc/resolvespec`, `$HOME/.resolvespec`, `./config`, `.` (CWD last, not dropped)
+- **#6** — Partly fixed: `ConfigFileUsed()` added; no log line because `pkg/config` cannot import `logger` (import cycle)
+- **#7** — Fixed: `Set` has a pointer receiver and allocates
+- **#8** — Not fixed: still a bare map; `Set` documented as not concurrency-safe
+- **#9** — Fixed: `LookupIPAddr` with a 2s timeout, fallback normalised to bare IPs and populates the slice
+- **#10** — Fixed: dead `Unmarshal` removed, `SetConfig` is atomic
+- **#11** — Fixed: recover removed (nothing in the function can panic)
+- **#12** — Partly fixed: `Config.Validate()` added, but it is not called from `GetConfig()`. The `*` CORS+credentials check is not implemented
+- **#13** — Documented only: `GetDefault` returns a pointer to a copy
+- **#14** — Fixed: `Join` errors if the result escapes the base
+- **#3** — Open: default flips deferred by decision (breaking change).
+- Tests: `pkg/config/hardening_test.go`.
+
 ---
 
 ## Findings

cmd/testserver/main.go

@@ -24,6 +24,7 @@ import (
 func main() {
 	// Load configuration
 	cfgMgr := config.NewManager()
+	config.SetConfigManager(cfgMgr)
 	if err := cfgMgr.Load(); err != nil {
 		log.Fatalf("Failed to load configuration: %v", err)
 	}

pkg/config/config.go

@@ -1,6 +1,9 @@
 package config
 
-import "time"
+import (
+	"fmt"
+	"time"
+)
 
 // Config represents the complete application configuration
 type Config struct {
@@ -198,3 +201,23 @@ type EventBrokerRetryPolicyConfig struct {
 // This is a map of path name to file system path
 // Example: "data_dir": "/var/lib/myapp/data"
 type PathsConfig map[string]string
+
+// Validate checks every configuration section for invalid or unsafe values.
+func (c *Config) Validate() error {
+	if err := c.Servers.Validate(); err != nil {
+		return fmt.Errorf("servers: %w", err)
+	}
+	if c.Middleware.RateLimitRPS < 0 || c.Middleware.RateLimitBurst < 0 {
+		return fmt.Errorf("middleware: rate_limit_rps and rate_limit_burst must not be negative")
+	}
+	if c.Middleware.MaxRequestSize <= 0 {
+		return fmt.Errorf("middleware: max_request_size must be greater than 0")
+	}
+	if c.EventBroker.Enabled && c.EventBroker.WorkerCount <= 0 {
+		return fmt.Errorf("event_broker: worker_count must be greater than 0")
+	}
+	if c.DBManager.MaxOpenConns < 0 || c.DBManager.MaxIdleConns < 0 || c.DBManager.RetryAttempts < 0 {
+		return fmt.Errorf("dbmanager: max_open_conns, max_idle_conns and retry_attempts must not be negative")
+	}
+	return nil
+}

pkg/config/hardening_test.go

@@ -0,0 +1,79 @@
+package config
+
+import (
+	"os"
+	"path/filepath"
+	"sync"
+	"testing"
+)
+
+func TestManagerConcurrentSetGet(t *testing.T) {
+	m := NewManager()
+	var wg sync.WaitGroup
+	for i := 0; i < 8; i++ {
+		wg.Add(2)
+		go func() {
+			defer wg.Done()
+			for j := 0; j < 200; j++ {
+				m.Set("x.y", j)
+			}
+		}()
+		go func() {
+			defer wg.Done()
+			for j := 0; j < 200; j++ {
+				_ = m.Get("x.y")
+				_, _ = m.GetConfig()
+			}
+		}()
+	}
+	wg.Wait()
+}
+
+func TestNewManagerDoesNotReplaceGlobal(t *testing.T) {
+	g := GetConfigManager()
+	_ = NewManager()
+	if GetConfigManager() != g {
+		t.Fatal("NewManager replaced the global manager")
+	}
+}
+
+func TestSaveConfigPermissions(t *testing.T) {
+	path := filepath.Join(t.TempDir(), "out.yaml")
+	if err := os.WriteFile(path, nil, 0o644); err != nil {
+		t.Fatal(err)
+	}
+	if err := NewManager().SaveConfig(path); err != nil {
+		t.Fatal(err)
+	}
+	if fi, _ := os.Stat(path); fi.Mode().Perm() != 0o600 {
+		t.Fatalf("mode = %v, want 0600", fi.Mode().Perm())
+	}
+}
+
+func TestPathsSetNilAndJoinConfined(t *testing.T) {
+	var pc PathsConfig
+	pc.Set("data", "data")
+	if got, _ := pc.Get("data"); got != "data" {
+		t.Fatalf("got %q", got)
+	}
+	if _, err := pc.Join("data", "../../etc/passwd"); err == nil {
+		t.Fatal("expected traversal error")
+	}
+	if p, err := pc.Join("data", "a", "b"); err != nil || p != filepath.Join("data", "a", "b") {
+		t.Fatalf("got %q, %v", p, err)
+	}
+}
+
+func TestConfigValidate(t *testing.T) {
+	cfg, err := NewManager().GetConfig()
+	if err != nil {
+		t.Fatal(err)
+	}
+	if err := cfg.Validate(); err != nil {
+		t.Fatalf("defaults should validate: %v", err)
+	}
+	cfg.EventBroker.Enabled, cfg.EventBroker.WorkerCount = true, 0
+	if cfg.Validate() == nil {
+		t.Fatal("expected worker_count error")
+	}
+}

pkg/config/manager.go

@@ -2,37 +2,60 @@ package config
 
 import (
 	"fmt"
+	"os"
 	"strings"
+	"sync"
 
 	"github.com/spf13/viper"
 )
 
-// Manager handles configuration loading from multiple sources
+// Manager handles configuration loading from multiple sources.
+// viper.Viper is not safe for concurrent use, so every access to it is guarded by mu.
 type Manager struct {
-	v *viper.Viper
+	mu sync.RWMutex
+	v  *viper.Viper
 }
 
-var configInstance *Manager
+var (
+	configInstance *Manager
+	configMu       sync.Mutex
+)
 
 // GetConfigManager returns a singleton configuration manager instance
 func GetConfigManager() *Manager {
+	configMu.Lock()
+	defer configMu.Unlock()
 	if configInstance == nil {
 		configInstance = NewManager()
 	}
 	return configInstance
 }
 
-// NewManager creates a new configuration manager with defaults
+// SetConfigManager publishes m as the global manager returned by GetConfigManager.
+// NewManager no longer does this implicitly.
+func SetConfigManager(m *Manager) {
+	configMu.Lock()
+	defer configMu.Unlock()
+	configInstance = m
+}
+
+// NewManager creates a new, isolated configuration manager with defaults.
+// It does not replace the global manager; use SetConfigManager for that.
 func NewManager() *Manager {
 	v := viper.New()
 
 	// Set configuration file settings
 	v.SetConfigName("config")
 	v.SetConfigType("yaml")
-	v.AddConfigPath(".")
+	// Most trusted location first; the working directory is the least trustworthy
-	v.AddConfigPath("./config")
+	// and is searched last (viper takes the first match).
 	v.AddConfigPath("/etc/resolvespec")
 	v.AddConfigPath("$HOME/.resolvespec")
+	v.AddConfigPath("./config")
+	v.AddConfigPath(".")
+
+	// Saved configs may contain secrets; never write them world-readable
+	v.SetConfigPermissions(0o600)
 
 	// Enable environment variable support
 	v.SetEnvPrefix("RESOLVESPEC")
@@ -42,8 +65,7 @@ func NewManager() *Manager {
 	// Set default values
 	setDefaults(v)
 
-	configInstance = &Manager{v: v}
+	return &Manager{v: v}
-	return configInstance
 }
 
 // NewManagerWithOptions creates a new configuration manager with custom options
@@ -61,6 +83,8 @@ type Option func(*Manager)
 // WithConfigFile sets a specific config file path
 func WithConfigFile(path string) Option {
 	return func(m *Manager) {
+		m.mu.Lock()
+		defer m.mu.Unlock()
 		m.v.SetConfigFile(path)
 	}
 }
@@ -68,6 +92,8 @@ func WithConfigFile(path string) Option {
 // WithConfigName sets the config file name (without extension)
 func WithConfigName(name string) Option {
 	return func(m *Manager) {
+		m.mu.Lock()
+		defer m.mu.Unlock()
 		m.v.SetConfigName(name)
 	}
 }
@@ -75,6 +101,8 @@ func WithConfigName(name string) Option {
 // WithConfigPath adds a path to search for config files
 func WithConfigPath(path string) Option {
 	return func(m *Manager) {
+		m.mu.Lock()
+		defer m.mu.Unlock()
 		m.v.AddConfigPath(path)
 	}
 }
@@ -82,13 +110,19 @@ func WithConfigPath(path string) Option {
 // WithEnvPrefix sets the environment variable prefix
 func WithEnvPrefix(prefix string) Option {
 	return func(m *Manager) {
+		m.mu.Lock()
+		defer m.mu.Unlock()
 		m.v.SetEnvPrefix(prefix)
 	}
 }
 
-// Load attempts to load configuration from file and environment
+// Load attempts to load configuration from file and environment.
+// A missing config file is not an error (defaults and env vars are used); check
+// ConfigFileUsed after Load to see whether a file was actually read.
 func (m *Manager) Load() error {
-	// Try to read config file (not an error if it doesn't exist)
+	m.mu.Lock()
+	defer m.mu.Unlock()
+
 	if err := m.v.ReadInConfig(); err != nil {
 		if _, ok := err.(viper.ConfigFileNotFoundError); !ok {
 			return fmt.Errorf("error reading config file: %w", err)
@@ -99,8 +133,19 @@ func (m *Manager) Load() error {
 	return nil
 }
 
+// ConfigFileUsed returns the config file that was read by Load, or "" if none was
+// found (i.e. the manager is running on defaults and environment variables only).
+func (m *Manager) ConfigFileUsed() string {
+	m.mu.RLock()
+	defer m.mu.RUnlock()
+	return m.v.ConfigFileUsed()
+}
+
 // GetConfig returns the complete configuration
 func (m *Manager) GetConfig() (*Config, error) {
+	m.mu.RLock()
+	defer m.mu.RUnlock()
+
 	var cfg Config
 	if err := m.v.Unmarshal(&cfg); err != nil {
 		return nil, fmt.Errorf("failed to unmarshal config: %w", err)
@@ -108,16 +153,11 @@ func (m *Manager) GetConfig() (*Config, error) {
 	return &cfg, nil
 }
 
-// SetConfig sets the complete configuration
+// SetConfig sets the complete configuration atomically
 func (m *Manager) SetConfig(cfg *Config) error {
-	configMap := make(map[string]interface{})
+	m.mu.Lock()
+	defer m.mu.Unlock()
 
-	// Marshal the config to a map structure that viper can use
-	if err := m.v.Unmarshal(&configMap); err != nil {
-		return fmt.Errorf("failed to prepare config map: %w", err)
-	}
-
-	// Use viper's merge to apply the config
 	m.v.Set("servers", cfg.Servers)
 	m.v.Set("tracing", cfg.Tracing)
 	m.v.Set("cache", cfg.Cache)
@@ -135,34 +175,54 @@ func (m *Manager) SetConfig(cfg *Config) error {
 
 // Get returns a configuration value by key
 func (m *Manager) Get(key string) interface{} {
+	m.mu.RLock()
+	defer m.mu.RUnlock()
 	return m.v.Get(key)
 }
 
 // GetString returns a string configuration value
 func (m *Manager) GetString(key string) string {
+	m.mu.RLock()
+	defer m.mu.RUnlock()
 	return m.v.GetString(key)
 }
 
 // GetInt returns an int configuration value
 func (m *Manager) GetInt(key string) int {
+	m.mu.RLock()
+	defer m.mu.RUnlock()
 	return m.v.GetInt(key)
 }
 
 // GetBool returns a bool configuration value
 func (m *Manager) GetBool(key string) bool {
+	m.mu.RLock()
+	defer m.mu.RUnlock()
 	return m.v.GetBool(key)
 }
 
 // Set sets a configuration value
 func (m *Manager) Set(key string, value interface{}) {
+	m.mu.Lock()
+	defer m.mu.Unlock()
 	m.v.Set(key, value)
 }
 
-// SaveConfig writes the current configuration to the specified path
+// SaveConfig writes the current configuration to the specified path.
+// The file contains the entire merged configuration, including secrets
+// (database/redis passwords, error-tracking DSN), so it is written with mode 0600.
+// Prefer supplying secrets via RESOLVESPEC_* environment variables.
 func (m *Manager) SaveConfig(path string) error {
+	m.mu.RLock()
+	defer m.mu.RUnlock()
+
 	if err := m.v.WriteConfigAs(path); err != nil {
 		return fmt.Errorf("failed to save config to %s: %w", path, err)
 	}
+	// viper only applies its permissions when creating the file; tighten a pre-existing one too
+	if err := os.Chmod(path, 0o600); err != nil {
+		return fmt.Errorf("failed to restrict permissions on %s: %w", path, err)
+	}
 	return nil
 }
 

pkg/config/paths.go

@@ -4,6 +4,7 @@ import (
 	"fmt"
 	"os"
 	"path/filepath"
+	"strings"
 )
 
 // Get retrieves a path by name
@@ -34,9 +35,13 @@ func (pc PathsConfig) GetOrDefault(name, defaultPath string) string {
 	return path
 }
 
-// Set sets a path by name
+// Set sets a path by name. It takes a pointer so a nil map can be allocated.
-func (pc PathsConfig) Set(name, path string) {
+// PathsConfig is not safe for concurrent mutation; populate it before sharing.
-	pc[name] = path
+func (pc *PathsConfig) Set(name, path string) {
+	if *pc == nil {
+		*pc = make(PathsConfig)
+	}
+	(*pc)[name] = path
 }
 
 // Has checks if a path exists by name
@@ -92,7 +97,8 @@ func (pc PathsConfig) AbsPath(name string) (string, error) {
 	return absPath, nil
 }
 
-// Join joins path segments with a named base path
+// Join joins path segments with a named base path.
+// It returns an error if the result would escape the base path.
 func (pc PathsConfig) Join(name string, elem ...string) (string, error) {
 	base, err := pc.Get(name)
 	if err != nil {
@@ -100,7 +106,14 @@ func (pc PathsConfig) Join(name string, elem ...string) (string, error) {
 	}
 
 	parts := append([]string{base}, elem...)
-	return filepath.Join(parts...), nil
+	joined := filepath.Join(parts...)
+
+	// filepath.Join resolves ".." rather than rejecting it; ensure the result stays under base
+	cleanBase := filepath.Clean(base)
+	if joined != cleanBase && !strings.HasPrefix(joined, cleanBase+string(os.PathSeparator)) {
+		return "", fmt.Errorf("path %q escapes base path '%s'", joined, name)
+	}
+	return joined, nil
 }
 
 // List returns all configured path names

pkg/config/server.go

@@ -1,10 +1,12 @@
 package config
 
 import (
+	"context"
 	"fmt"
 	"net"
 	"os"
 	"strings"
+	"time"
 )
 
 // ApplyGlobalDefaults applies global server defaults to this instance
@@ -95,7 +97,8 @@ func (sc *ServersConfig) Validate() error {
 	return nil
 }
 
-// GetDefault returns the default server instance configuration
+// GetDefault returns the default server instance configuration.
+// The returned pointer refers to a copy: mutating it does not modify sc.Instances.
 func (sc *ServersConfig) GetDefault() (*ServerInstanceConfig, error) {
 	if sc.DefaultServer == "" {
 		return nil, fmt.Errorf("no default server configured")
@@ -109,41 +112,37 @@ func (sc *ServersConfig) GetDefault() (*ServerInstanceConfig, error) {
 	return &instance, nil
 }
 
-// GetIPs - GetIP for pc
+// GetIPs returns the hostname, a comma-separated list of non-loopback IPs and the
+// same IPs as []net.IP. The lookup is bounded by a short timeout.
 func GetIPs() (hostname string, ipList string, ipNetList []net.IP) {
-	defer func() {
-		if err := recover(); err != nil {
-			fmt.Println("Recovered in GetIPs", err)
-		}
-	}()
 	hostname, _ = os.Hostname()
-	ipaddrlist := make([]net.IP, 0)
+	ipNetList = make([]net.IP, 0)
-	iplist := ""
-	addrs, err := net.LookupIP(hostname)
-	if err != nil {
-		return hostname, iplist, ipaddrlist
-	}
 
-	for _, a := range addrs {
+	ctx, cancel := context.WithTimeout(context.Background(), 2*time.Second)
-		// cfg.LogInfo("\nFound IP Host Address: %s", a)
+	defer cancel()
-		if strings.Contains(a.String(), "127.0.0.1") {
+
-			continue
+	var ips []string
-		}
+	if addrs, err := net.DefaultResolver.LookupIPAddr(ctx, hostname); err == nil {
-		iplist = fmt.Sprintf("%s,%s", iplist, a)
+		for _, a := range addrs {
-		ipaddrlist = append(ipaddrlist, a)
+			if a.IP.IsLoopback() {
-	}
-	if iplist == "" {
-		iff, _ := net.InterfaceAddrs()
-		for _, a := range iff {
-			// cfg.LogInfo("\nFound IP Address: %s", a)
-			if strings.Contains(a.String(), "127.0.0.1") {
 				continue
 			}
-			iplist = fmt.Sprintf("%s,%s", iplist, a)
+			ips = append(ips, a.IP.String())
-
+			ipNetList = append(ipNetList, a.IP)
 		}
-
 	}
-	iplist = strings.TrimLeft(iplist, ",")
+
-	return hostname, iplist, ipaddrlist
+	if len(ips) == 0 {
+		ifaceAddrs, _ := net.InterfaceAddrs()
+		for _, a := range ifaceAddrs {
+			ipn, ok := a.(*net.IPNet)
+			if !ok || ipn.IP.IsLoopback() {
+				continue
+			}
+			ips = append(ips, ipn.IP.String())
+			ipNetList = append(ipNetList, ipn.IP)
+		}
+	}
+
+	return hostname, strings.Join(ips, ","), ipNetList
 }