jarl/internal/config/config.go

// Package config loads jarld's configuration file.
package config

import (
	"bytes"
	"fmt"
	"net/netip"
	"os"
	"path/filepath"
	"regexp"
	"strings"
	"time"

	"go.yaml.in/yaml/v3"
)

// DefaultPath is where jarld looks for its configuration.
const DefaultPath = "/etc/jarl/jarl.yaml"

// Config is the daemon configuration.
type Config struct {
	Listen         []string `yaml:"listen"`
	TLS            TLS      `yaml:"tls"`
	TrustedProxies []string `yaml:"trusted_proxies"`
	Database       string   `yaml:"database"`
	StagingDir     string   `yaml:"staging_dir"`
	Log            Log      `yaml:"log"`
	Session        Session  `yaml:"session"`
	BIND           BIND     `yaml:"bind"`
	Audit          Audit    `yaml:"audit"`
	Stats          Stats    `yaml:"stats"`
}

// TLS configures HTTPS.
type TLS struct {
	CertFile string `yaml:"cert_file"`
	KeyFile  string `yaml:"key_file"`
}

// Log configures logging.
type Log struct {
	Level  string `yaml:"level"`  // debug, info, warn, error
	Format string `yaml:"format"` // text, json
}

// Session configures web sessions.
type Session struct {
	Lifetime     Duration `yaml:"lifetime"`
	IdleTimeout  Duration `yaml:"idle_timeout"`
	CookieSecure string   `yaml:"cookie_secure"` // auto, always, never
}

// BIND says where BIND's files and tools are.
type BIND struct {
	MainConfig     string   `yaml:"main_config"`
	ConfigDir      string   `yaml:"config_dir"`
	ZoneDir        string   `yaml:"zone_dir"`
	CacheDir       string   `yaml:"cache_dir"`
	FileGroup      string   `yaml:"file_group"`
	NamedCheckConf string   `yaml:"named_checkconf"`
	NamedCheckZone string   `yaml:"named_checkzone"`
	RNDC           string   `yaml:"rndc"`
	RNDCArgs       []string `yaml:"rndc_args"`
	ReloadTimeout  Duration `yaml:"reload_timeout"`
	NoReload       bool     `yaml:"no_reload"`
}

// Audit configures the audit log.
type Audit struct {
	// RetentionDays deletes older audit events (0 keeps everything).
	RetentionDays int `yaml:"retention_days"`
}

// Stats configures query statistics. When enabled, Jarl makes named publish
// its counters on a local statistics channel and log every response to a
// file, and jarld aggregates both into a separate database.
type Stats struct {
	Enabled bool `yaml:"enabled"`
	// QueryLog enables per-query data (top names and clients, per-view and
	// per-zone counts) from named's response log. Without it only named's
	// own counters are collected.
	QueryLog bool `yaml:"query_log"`
	// Database is the statistics database, kept apart from Jarl's
	// configuration database.
	Database string `yaml:"database"`
	// StatisticsChannel is the address named serves its counters on
	// (only reachable from this host).
	StatisticsChannel string `yaml:"statistics_channel"`
	// LogFile is where named logs responses. named's AppArmor profile
	// allows writing under /var/log/named.
	LogFile     string         `yaml:"log_file"`
	LogSize     string         `yaml:"log_size"`
	LogVersions int            `yaml:"log_versions"`
	Retention   StatsRetention `yaml:"retention"`
	// TopN is how many names and clients are kept per hour and per day.
	TopN int `yaml:"top_n"`
	// AnonymizeClients stores client networks (IPv4 /24, IPv6 /48)
	// instead of addresses.
	AnonymizeClients bool `yaml:"anonymize_clients"`
	// ExcludeClients are addresses or prefixes whose queries are not
	// counted per name or client (e.g. monitoring probes).
	ExcludeClients []string `yaml:"exclude_clients"`
}

// StatsRetention says how long each resolution of the statistics is kept.
type StatsRetention struct {
	Minute Duration `yaml:"minute"`
	Hour   Duration `yaml:"hour"`
	Day    Duration `yaml:"day"`
}

// Duration is a time.Duration written as "15s" or "12h" in YAML. Days
// ("30d") are accepted too.
type Duration time.Duration

// UnmarshalYAML implements yaml.Unmarshaler.
func (d *Duration) UnmarshalYAML(n *yaml.Node) error {
	var s string
	if err := n.Decode(&s); err != nil {
		return err
	}
	v, err := ParseDuration(s)
	if err != nil {
		return err
	}
	*d = Duration(v)
	return nil
}

// MarshalYAML implements yaml.Marshaler.
func (d Duration) MarshalYAML() (any, error) { return time.Duration(d).String(), nil }

// ParseDuration parses a Go duration, also accepting a "d" suffix.
func ParseDuration(s string) (time.Duration, error) {
	s = strings.TrimSpace(s)
	if n, ok := strings.CutSuffix(s, "d"); ok {
		var days int
		if _, err := fmt.Sscanf(n, "%d", &days); err != nil || fmt.Sprint(days) != n {
			return 0, fmt.Errorf("invalid duration %q", s)
		}
		return time.Duration(days) * 24 * time.Hour, nil
	}
	return time.ParseDuration(s)
}

// Default returns the defaults for a Debian/Ubuntu BIND 9.20 server.
func Default() *Config {
	return &Config{
		Listen:     []string{"127.0.0.1:8053"},
		Database:   "/var/lib/jarl/jarl.db",
		StagingDir: "/var/lib/jarl/staging",
		Log:        Log{Level: "info", Format: "text"},
		Session:    Session{Lifetime: Duration(12 * time.Hour), IdleTimeout: Duration(2 * time.Hour), CookieSecure: "auto"},
		BIND: BIND{
			MainConfig:     "/etc/bind/named.conf",
			ConfigDir:      "/etc/bind/jarl",
			ZoneDir:        "/var/lib/bind/jarl",
			CacheDir:       "/var/cache/bind/jarl",
			FileGroup:      "bind",
			NamedCheckConf: "/usr/bin/named-checkconf",
			NamedCheckZone: "/usr/bin/named-checkzone",
			RNDC:           "/usr/sbin/rndc",
			ReloadTimeout:  Duration(15 * time.Second),
		},
		Stats: Stats{
			QueryLog:          true,
			Database:          "/var/lib/jarl/stats.db",
			StatisticsChannel: "127.0.0.1:8058",
			LogFile:           "/var/log/named/jarl-responses.log",
			LogSize:           "20m",
			LogVersions:       2,
			Retention:         StatsRetention{Minute: Duration(7 * 24 * time.Hour), Hour: Duration(90 * 24 * time.Hour), Day: Duration(730 * 24 * time.Hour)},
			TopN:              100,
		},
	}
}

// Load reads the configuration file at path over the defaults. A missing
// file is not an error when allowMissing is set.
func Load(path string, allowMissing bool) (*Config, error) {
	c := Default()
	b, err := os.ReadFile(path)
	if err != nil {
		if os.IsNotExist(err) && allowMissing {
			return c, c.check()
		}
		return nil, err
	}
	dec := yaml.NewDecoder(bytes.NewReader(b))
	dec.KnownFields(true)
	if err := dec.Decode(c); err != nil && err.Error() != "EOF" {
		return nil, fmt.Errorf("%s: %w", path, err)
	}
	return c, c.check()
}

func (c *Config) check() error {
	if len(c.Listen) == 0 {
		return fmt.Errorf("listen: at least one address is needed")
	}
	if (c.TLS.CertFile == "") != (c.TLS.KeyFile == "") {
		return fmt.Errorf("tls: set both cert_file and key_file")
	}
	for _, p := range c.TrustedProxies {
		if _, err := ParsePrefix(p); err != nil {
			return fmt.Errorf("trusted_proxies: %v", err)
		}
	}
	for name, p := range map[string]string{"database": c.Database, "bind.config_dir": c.BIND.ConfigDir, "bind.zone_dir": c.BIND.ZoneDir, "bind.cache_dir": c.BIND.CacheDir} {
		if p == "" || !filepath.IsAbs(p) {
			return fmt.Errorf("%s must be an absolute path", name)
		}
	}
	switch c.Session.CookieSecure {
	case "auto", "always", "never":
	default:
		return fmt.Errorf("session.cookie_secure must be auto, always or never")
	}
	if err := c.Stats.check(); err != nil {
		return fmt.Errorf("stats: %w", err)
	}
	switch c.Log.Format {
	case "text", "json":
	default:
		return fmt.Errorf("log.format must be text or json")
	}
	return nil
}

func (s *Stats) check() error {
	if !s.Enabled {
		return nil
	}
	if s.Database == "" || !filepath.IsAbs(s.Database) {
		return fmt.Errorf("database must be an absolute path")
	}
	ap, err := netip.ParseAddrPort(s.StatisticsChannel)
	if err != nil {
		return fmt.Errorf("statistics_channel %q must be ADDRESS:PORT, e.g. 127.0.0.1:8058", s.StatisticsChannel)
	}
	if !ap.Addr().IsLoopback() {
		return fmt.Errorf("statistics_channel must be a loopback address (named's counters must not be reachable from the network)")
	}
	if s.QueryLog {
		if !filepath.IsAbs(s.LogFile) {
			return fmt.Errorf("log_file must be an absolute path")
		}
		if strings.ContainsAny(s.LogFile, "\"\n") {
			return fmt.Errorf("log_file contains invalid characters")
		}
		if !logSizeRe.MatchString(s.LogSize) {
			return fmt.Errorf("log_size %q must be a size such as 20m or 1g", s.LogSize)
		}
		if s.LogVersions < 1 || s.LogVersions > 100 {
			return fmt.Errorf("log_versions must be between 1 and 100")
		}
	}
	r := s.Retention
	if r.Minute < Duration(time.Hour) || r.Hour < Duration(24*time.Hour) || r.Day < Duration(24*time.Hour) {
		return fmt.Errorf("retention must keep minutes for at least 1h, hours and days for at least 1d")
	}
	if s.TopN < 10 || s.TopN > 10000 {
		return fmt.Errorf("top_n must be between 10 and 10000")
	}
	for _, p := range s.ExcludeClients {
		if _, err := ParsePrefix(p); err != nil {
			return fmt.Errorf("exclude_clients: %v", err)
		}
	}
	return nil
}

var logSizeRe = regexp.MustCompile(`^[0-9]+[kKmMgG]?$`)

// ParsePrefix parses an address or prefix.
func ParsePrefix(s string) (netip.Prefix, error) {
	if strings.Contains(s, "/") {
		return netip.ParsePrefix(s)
	}
	a, err := netip.ParseAddr(s)
	if err != nil {
		return netip.Prefix{}, err
	}
	return netip.PrefixFrom(a, a.BitLen()), nil
}

// Proxies returns the trusted proxy prefixes.
func (c *Config) Proxies() []netip.Prefix {
	var out []netip.Prefix
	for _, p := range c.TrustedProxies {
		if pf, err := ParsePrefix(p); err == nil {
			out = append(out, pf.Masked())
		}
	}
	return out
}

// Example is a commented example configuration.
const Example = `# Jarl daemon configuration (/etc/jarl/jarl.yaml).
# Every setting is optional; the values shown are the defaults for
# Debian/Ubuntu with BIND 9.20.

# Addresses the web UI and API listen on. Keep this off the internet:
# bind to localhost or a management network, and use TLS (below) or a
# TLS-terminating reverse proxy.
listen:
  - "127.0.0.1:8053"

# HTTPS. The certificate is reread when the file changes (or on SIGHUP),
# so renewed Let's Encrypt certificates are picked up automatically.
tls:
  cert_file: ""
  key_file: ""

# Reverse proxies whose X-Forwarded-For / X-Forwarded-Proto headers are
# trusted (addresses or prefixes).
trusted_proxies: []

database: /var/lib/jarl/jarl.db
staging_dir: /var/lib/jarl/staging

log:
  level: info      # debug, info, warn or error
  format: text     # text or json (both go to the journal under systemd)

session:
  lifetime: 12h      # web sessions end after this long...
  idle_timeout: 2h   # ...or after this long without activity
  cookie_secure: auto  # auto, always or never

bind:
  # The file named loads. 'jarld adopt' makes it include Jarl's config.
  main_config: /etc/bind/named.conf
  # Where Jarl writes named.conf/keys.conf, zone files, and where named
  # keeps files of secondary zones. named (AppArmor) may read /etc/bind
  # and write /var/lib/bind and /var/cache/bind.
  config_dir: /etc/bind/jarl
  zone_dir: /var/lib/bind/jarl
  cache_dir: /var/cache/bind/jarl
  # Group of generated files; named must be able to read them.
  file_group: bind
  named_checkconf: /usr/bin/named-checkconf
  named_checkzone: /usr/bin/named-checkzone
  rndc: /usr/sbin/rndc
  # Extra rndc arguments, e.g. ["-k", "/etc/bind/rndc.key", "-s", "127.0.0.1"].
  rndc_args: []
  # How long to wait for named to load changed zones before rolling back.
  reload_timeout: 15s
  # Write files but never run rndc (no named on this host).
  no_reload: false

audit:
  # Delete audit events older than this many days (0 keeps them forever).
  retention_days: 0

# Query statistics (docs/stats.md). Off by default. When enabled, Jarl makes
# named publish its counters on a local statistics channel and log every
# response, and jarld aggregates both into a separate database. Changing
# this section takes effect when jarld restarts (it rewrites named.conf).
stats:
  enabled: false
  # Per-query data (top names and clients, per view and zone). false
  # collects only named's own counters, with no logging overhead.
  query_log: true
  database: /var/lib/jarl/stats.db
  statistics_channel: 127.0.0.1:8058
  log_file: /var/log/named/jarl-responses.log
  log_size: 20m
  log_versions: 2
  retention:
    minute: 7d     # per-minute counters
    hour: 90d      # hourly counters and top lists
    day: 730d      # daily counters and top lists
  top_n: 100       # names and clients kept per hour and per day
  anonymize_clients: false   # store 10.10.0.0/24 instead of 10.10.0.37
  exclude_clients: []        # e.g. [127.0.0.1] to ignore local monitoring
`