jarl/cmd/jarld/main.go

// Command jarld is the Jarl daemon: it serves the web UI and the REST API
// and writes BIND's configuration. It also has administrative subcommands
// for setting up and migrating a server.
package main

import (
	"bufio"
	"context"
	"errors"
	"fmt"
	"io"
	"log/slog"
	"net"
	"net/netip"
	"os"
	"os/signal"
	"path/filepath"
	"strings"
	"syscall"
	"time"

	"github.com/spf13/cobra"
	"golang.org/x/term"

	"jarl/internal/auth"
	"jarl/internal/bind"
	"jarl/internal/config"
	"jarl/internal/engine"
	"jarl/internal/model"
	"jarl/internal/render"
	"jarl/internal/server"
	"jarl/internal/stats"
	"jarl/internal/store"
	"jarl/internal/version"
	"jarl/internal/web"
)

func main() {
	if err := rootCmd().Execute(); err != nil {
		fmt.Fprintln(os.Stderr, "error:", err)
		os.Exit(1)
	}
}

var configPath string

func rootCmd() *cobra.Command {
	root := &cobra.Command{
		Use:           "jarld",
		Short:         "Jarl: manage BIND (named) through a web UI, a REST API and a CLI",
		SilenceUsage:  true,
		SilenceErrors: true,
		Version:       version.Version,
	}
	def := os.Getenv("JARL_CONFIG")
	if def == "" {
		def = config.DefaultPath
	}
	root.PersistentFlags().StringVarP(&configPath, "config", "c", def, "configuration file")
	root.AddCommand(serveCmd(), initCmd(), userCmd(), importCmd(), adoptCmd(), checkCmd(), renderCmd(), resyncCmd(), backupCmd(), configCmd(), versionCmd())
	return root
}

// app holds what most commands need.
type app struct {
	cfg   *config.Config
	log   *slog.Logger
	store *store.Store
	eng   *engine.Engine
	owner *dirOwner
}

func newLogger(cfg *config.Config, w io.Writer) *slog.Logger {
	var level slog.Level
	switch strings.ToLower(cfg.Log.Level) {
	case "debug":
		level = slog.LevelDebug
	case "warn", "warning":
		level = slog.LevelWarn
	case "error":
		level = slog.LevelError
	default:
		level = slog.LevelInfo
	}
	opts := &slog.HandlerOptions{Level: level}
	if cfg.Log.Format == "json" {
		return slog.New(slog.NewJSONHandler(w, opts))
	}
	// systemd's journal adds timestamps.
	if os.Getenv("JOURNAL_STREAM") != "" {
		opts.ReplaceAttr = func(groups []string, a slog.Attr) slog.Attr {
			if a.Key == slog.TimeKey && len(groups) == 0 {
				return slog.Attr{}
			}
			return a
		}
	}
	return slog.New(slog.NewTextHandler(w, opts))
}

func openApp(allowMissingConfig bool) (*app, error) {
	cfg, err := config.Load(configPath, allowMissingConfig)
	if err != nil {
		return nil, fmt.Errorf("loading configuration: %w", err)
	}
	log := newLogger(cfg, os.Stderr)
	owner := newDirOwner(filepath.Dir(cfg.Database))
	if err := os.MkdirAll(filepath.Dir(cfg.Database), 0o750); err != nil {
		return nil, err
	}
	st, err := store.Open(cfg.Database)
	if err != nil {
		return nil, fmt.Errorf("opening database %s: %w", cfg.Database, err)
	}
	tools := &bind.Tools{NamedCheckConf: cfg.BIND.NamedCheckConf, NamedCheckZone: cfg.BIND.NamedCheckZone, RNDC: cfg.BIND.RNDC, RNDCArgs: cfg.BIND.RNDCArgs}
	eng := engine.New(st, tools, engine.Config{
		Paths:         render.Paths{ConfigDir: cfg.BIND.ConfigDir, ZoneDir: cfg.BIND.ZoneDir, CacheDir: cfg.BIND.CacheDir},
		MainConfig:    cfg.BIND.MainConfig,
		StagingDir:    cfg.StagingDir,
		FileGroup:     cfg.BIND.FileGroup,
		ReloadTimeout: time.Duration(cfg.BIND.ReloadTimeout),
		NoReload:      cfg.BIND.NoReload,
		Extras:        renderExtras(cfg),
	}, log)
	return &app{cfg: cfg, log: log, store: st, eng: eng, owner: owner}, nil
}

// openStats opens the statistics database (next to, not inside, Jarl's
// configuration database).
func openStats(a *app) (*stats.Store, error) {
	path := a.cfg.Stats.Database
	if err := os.MkdirAll(filepath.Dir(path), 0o750); err != nil {
		return nil, err
	}
	ss, err := stats.OpenStore(path)
	if err != nil {
		return nil, err
	}
	return ss, nil
}

func statsOptions(cfg *config.Config) stats.Options {
	sc := cfg.Stats
	o := stats.Options{QueryLog: sc.QueryLog, LogFile: sc.LogFile, StatisticsChannel: sc.StatisticsChannel, TopN: sc.TopN,
		RetainMinute: time.Duration(sc.Retention.Minute), RetainHour: time.Duration(sc.Retention.Hour), RetainDay: time.Duration(sc.Retention.Day),
		Anonymize: sc.AnonymizeClients}
	for _, p := range sc.ExcludeClients {
		if pf, err := config.ParsePrefix(p); err == nil {
			o.Exclude = append(o.Exclude, pf.Masked())
		}
	}
	return o
}

// zoneIndex lists the zones named answers authoritatively, per view, so
// statistics can be broken down by zone.
func zoneIndex(a *app) stats.ZoneIndex {
	return func() map[string][]string {
		st, err := a.store.LoadState(context.Background())
		if err != nil {
			return nil
		}
		out := map[string][]string{}
		for _, z := range st.Zones {
			if z.Type != model.ZonePrimary && z.Type != model.ZoneSecondary && z.Type != model.ZoneMirror {
				continue
			}
			v := st.ViewName(z)
			out[v] = append(out[v], strings.TrimSuffix(z.Name, "."))
		}
		return out
	}
}

// renderExtras turns jarl.yaml settings that change named's configuration
// into render extras.
func renderExtras(cfg *config.Config) render.Extras {
	var x render.Extras
	if cfg.Stats.Enabled {
		h := &render.StatsHooks{Channel: netip.MustParseAddrPort(cfg.Stats.StatisticsChannel)}
		if cfg.Stats.QueryLog {
			h.LogFile, h.LogVersions, h.LogSize = cfg.Stats.LogFile, cfg.Stats.LogVersions, cfg.Stats.LogSize
		}
		x.Stats = h
	}
	return x
}

func (a *app) close() {
	a.store.Close()
	// Commands run as root must not leave root-owned files where the
	// daemon (running as its own user) needs to write.
	a.owner.fix(filepath.Dir(a.cfg.Database), a.cfg.StagingDir, a.cfg.BIND.ConfigDir, a.cfg.BIND.ZoneDir, a.cfg.BIND.CacheDir)
}

// localPrincipal is used for administrative commands run on the server.
func (a *app) localPrincipal(ctx context.Context, username string) (*auth.Principal, error) {
	if username == "" {
		username = os.Getenv("SUDO_USER")
	}
	if username != "" {
		if u, err := a.store.GetUserByName(ctx, username); err == nil {
			if !u.Admin {
				return nil, fmt.Errorf("user %s is not an administrator", username)
			}
			return &auth.Principal{User: u, Method: "local"}, nil
		}
	}
	// Fall back to the first administrator, or a synthetic one.
	users, err := a.store.ListUsers(ctx)
	if err != nil {
		return nil, err
	}
	for _, u := range users {
		if u.Admin && !u.Disabled {
			return &auth.Principal{User: &model.User{ID: u.ID, Username: u.Username + " (local)", Admin: true}, Method: "local"}, nil
		}
	}
	return &auth.Principal{User: &model.User{Username: "local", Admin: true}, Method: "local"}, nil
}

// ---- serve

func serveCmd() *cobra.Command {
	return &cobra.Command{
		Use:   "serve",
		Short: "Run the daemon (web UI and API)",
		RunE: func(cmd *cobra.Command, args []string) error {
			a, err := openApp(true)
			if err != nil {
				return err
			}
			defer a.close()
			ctx, stop := signal.NotifyContext(context.Background(), os.Interrupt, syscall.SIGTERM)
			defer stop()
			if users, err := a.store.ListUsers(ctx); err == nil && len(users) == 0 {
				a.log.Warn("there are no users yet; create an administrator with 'jarld init' or 'jarld user add --admin NAME'")
			}
			adopted, aerr := a.eng.Adopted()
			if aerr == nil && !adopted {
				a.log.Warn("named does not load Jarl's configuration yet; changes are written but named is not reloaded until 'jarld adopt' runs", "main_config", a.cfg.BIND.MainConfig)
			}
			if adopted {
				// Settings in jarl.yaml (such as stats) can change named's
				// configuration; bring it up to date.
				if wrote, res, err := a.eng.SyncConfig(ctx); err != nil {
					a.log.Warn("could not update named's configuration for the settings in jarl.yaml; fix the problem and restart jarl, or run 'jarl resync'", "err", err)
				} else if wrote {
					a.log.Info("updated named's configuration for the settings in jarl.yaml", "files", res.Files)
				}
				// Response logging is switched at runtime (the option only
				// counts when named starts). Turn it off again when
				// statistics no longer use it, unless the configuration
				// asks for it.
				if !a.cfg.Stats.Enabled || !a.cfg.Stats.QueryLog {
					if st, err := a.store.LoadState(ctx); err == nil && st.FindOption(model.ScopeOptions, 0, "responselog") == nil {
						if changed, err := a.eng.SetResponseLogging(ctx, false); err == nil && changed {
							a.log.Info("turned off named's response logging, which query statistics no longer use")
						}
					}
				}
			}
			var collector *stats.Collector
			collectorDone := make(chan struct{})
			if a.cfg.Stats.Enabled {
				ss, err := openStats(a)
				if err != nil {
					return err
				}
				defer ss.Close()
				opts := statsOptions(a.cfg)
				if adopted {
					opts.EnsureLogging = func(ctx context.Context) (bool, error) { return a.eng.SetResponseLogging(ctx, true) }
				}
				collector = stats.NewCollector(ss, opts, zoneIndex(a), a.log)
				cctx, cancel := context.WithCancel(ctx)
				go func() {
					collector.Run(cctx)
					close(collectorDone)
				}()
				// Stop the collector (it writes what it has) before the
				// database closes, however serving ends.
				defer func() { cancel(); <-collectorDone }()
			}
			srv := server.New(server.Config{
				Listen: a.cfg.Listen, TLSCertFile: a.cfg.TLS.CertFile, TLSKeyFile: a.cfg.TLS.KeyFile,
				TrustedProxies: a.cfg.Proxies(), SessionLifetime: time.Duration(a.cfg.Session.Lifetime),
				SessionIdle: time.Duration(a.cfg.Session.IdleTimeout), CookieSecure: a.cfg.Session.CookieSecure,
				Version: version.Version, Stats: collector, Socket: a.cfg.Socket,
				HTTPReq: server.HTTPReq{
					Enabled:           a.cfg.ACME.HTTPReq.Enabled,
					AllowPlainHTTP:    !a.cfg.ACME.HTTPReq.RequireHTTPS,
					AllowLoopbackHTTP: a.cfg.ACME.HTTPReq.AllowLoopbackHTTP,
				},
			}, a.eng, a.log, web.FS())
			hup := make(chan os.Signal, 1)
			signal.Notify(hup, syscall.SIGHUP)
			go func() {
				for range hup {
					sdNotify("RELOADING=1")
					srv.ReloadTLS()
					sdNotify("READY=1")
				}
			}()
			go auditRetention(ctx, a)
			a.log.Info("jarld starting", "version", version.Version, "database", a.cfg.Database, "bind_config", a.cfg.BIND.ConfigDir)
			err = srv.Serve(ctx, func() { sdNotify("READY=1\nSTATUS=serving") })
			sdNotify("STOPPING=1")
			a.log.Info("jarld stopped")
			return err
		},
	}
}

func auditRetention(ctx context.Context, a *app) {
	if a.cfg.Audit.RetentionDays <= 0 {
		return
	}
	t := time.NewTicker(6 * time.Hour)
	defer t.Stop()
	for {
		before := time.Now().Add(-time.Duration(a.cfg.Audit.RetentionDays) * 24 * time.Hour)
		if n, err := a.store.PruneAudit(ctx, before); err != nil {
			a.log.Warn("pruning audit log failed", "err", err)
		} else if n > 0 {
			a.log.Info("pruned audit log", "events", n)
		}
		select {
		case <-ctx.Done():
			return
		case <-t.C:
		}
	}
}

// sdNotify tells systemd about state changes (Type=notify).
func sdNotify(state string) {
	sock := os.Getenv("NOTIFY_SOCKET")
	if sock == "" {
		return
	}
	addr := &net.UnixAddr{Name: sock, Net: "unixgram"}
	if strings.HasPrefix(sock, "@") {
		addr.Name = "\x00" + sock[1:]
	}
	c, err := net.DialUnix("unixgram", nil, addr)
	if err != nil {
		return
	}
	defer c.Close()
	c.Write([]byte(state))
}

// ---- init and users

func readPassword(prompt string, fromStdin bool) (string, error) {
	if fromStdin || !term.IsTerminal(int(os.Stdin.Fd())) {
		line, err := bufio.NewReader(os.Stdin).ReadString('\n')
		if err != nil && !errors.Is(err, io.EOF) {
			return "", err
		}
		return strings.TrimRight(line, "\r\n"), nil
	}
	fmt.Fprint(os.Stderr, prompt)
	b, err := term.ReadPassword(int(os.Stdin.Fd()))
	fmt.Fprintln(os.Stderr)
	if err != nil {
		return "", err
	}
	fmt.Fprint(os.Stderr, "Repeat: ")
	b2, err := term.ReadPassword(int(os.Stdin.Fd()))
	fmt.Fprintln(os.Stderr)
	if err != nil {
		return "", err
	}
	if string(b) != string(b2) {
		return "", errors.New("passwords don't match")
	}
	return string(b), nil
}

func createToken(ctx context.Context, a *app, u *model.User, name string) (string, error) {
	secret, pid, hash := auth.NewToken()
	if err := a.store.CreateToken(ctx, &model.Token{UserID: u.ID, Name: name, PublicID: pid, SecretHash: hash}); err != nil {
		return "", err
	}
	return secret, nil
}

func initCmd() *cobra.Command {
	var username string
	var stdin, withToken bool
	var nameservers []string
	var rname string
	cmd := &cobra.Command{
		Use:   "init",
		Short: "Create the database, the first administrator and default server options",
		Long: `Creates Jarl's database with a first administrator account.

Jarl starts empty. Either import your existing BIND configuration with
'jarld import', or build one from scratch in the web UI.`,
		RunE: func(cmd *cobra.Command, args []string) error {
			a, err := openApp(true)
			if err != nil {
				return err
			}
			defer a.close()
			ctx := context.Background()
			users, err := a.store.ListUsers(ctx)
			if err != nil {
				return err
			}
			if len(users) > 0 {
				return fmt.Errorf("Jarl is already initialized (%d users); use 'jarld user add' to add users", len(users))
			}
			pw, err := readPassword("Password for "+username+": ", stdin)
			if err != nil {
				return err
			}
			hash, err := auth.HashPassword(pw)
			if err != nil {
				return err
			}
			u := &model.User{Username: username, PasswordHash: hash, Admin: true, Comment: "created by jarld init"}
			if err := a.store.CreateUser(ctx, u); err != nil {
				return err
			}
			p := &auth.Principal{User: u, Method: "local"}
			set := engine.SettingsInput{}
			if len(nameservers) > 0 {
				set.Nameservers = &nameservers
			}
			if rname != "" {
				set.RName = &rname
			}
			if set.Nameservers != nil || set.RName != nil {
				if _, err := a.eng.Submit(ctx, []engine.Request{engine.NewRequest(engine.OpSettingsUpdate, 0, set)}, p, "cli", "initial settings", engine.ApplyOptions{NoReload: true}); err != nil {
					return err
				}
			}
			a.eng.Audit(ctx, p, "system.init", "user", u.ID, u.Username, "success", "")
			fmt.Printf("Created administrator %q in %s.\n", u.Username, a.cfg.Database)
			if withToken {
				tok, err := createToken(ctx, a, u, "cli")
				if err != nil {
					return err
				}
				fmt.Printf("\nAPI token for the CLI (shown only once):\n\n  %s\n\n", tok)
			}
			fmt.Println("\nNext steps:")
			fmt.Println("  1. Import your current BIND configuration:  sudo -u jarl jarld import")
			fmt.Println("  2. Switch named over to Jarl:              sudo jarld adopt")
			fmt.Println("  3. Start the daemon:                        sudo systemctl enable --now jarl")
			return nil
		},
	}
	cmd.Flags().StringVarP(&username, "user", "u", "admin", "administrator username")
	cmd.Flags().BoolVar(&stdin, "password-stdin", false, "read the password from standard input")
	cmd.Flags().BoolVar(&withToken, "token", false, "also create an API token for the CLI")
	cmd.Flags().StringSliceVar(&nameservers, "nameservers", nil, "default name servers for new zones (e.g. ns1.example.com.)")
	cmd.Flags().StringVar(&rname, "contact", "", "default SOA contact for new zones (e.g. hostmaster@example.com)")
	return cmd
}

func userCmd() *cobra.Command {
	cmd := &cobra.Command{Use: "user", Short: "Manage users locally (for bootstrapping and recovery)"}
	var admin, stdin, withToken, service bool
	var grants []string
	add := &cobra.Command{
		Use:   "add NAME",
		Short: "Create a user",
		Args:  cobra.ExactArgs(1),
		RunE: func(cmd *cobra.Command, args []string) error {
			a, err := openApp(true)
			if err != nil {
				return err
			}
			defer a.close()
			ctx := context.Background()
			u := &model.User{Username: args[0], Admin: admin}
			for _, g := range grants {
				gr, err := auth.ParseGrant(g)
				if err != nil {
					return fmt.Errorf("grant %q: %w", g, err)
				}
				u.Grants = append(u.Grants, gr)
			}
			if !service {
				pw, err := readPassword("Password for "+u.Username+": ", stdin)
				if err != nil {
					return err
				}
				if u.PasswordHash, err = auth.HashPassword(pw); err != nil {
					return err
				}
			}
			if err := a.store.CreateUser(ctx, u); err != nil {
				return err
			}
			a.eng.Audit(ctx, &auth.Principal{User: &model.User{Username: "local"}, Method: "local"}, "user.create", "user", u.ID, u.Username, "success", "jarld user add")
			fmt.Printf("Created user %s.\n", u.Username)
			if withToken {
				tok, err := createToken(ctx, a, u, "cli")
				if err != nil {
					return err
				}
				fmt.Printf("API token (shown only once): %s\n", tok)
			}
			return nil
		},
	}
	add.Flags().BoolVar(&admin, "admin", false, "make the user an administrator")
	add.Flags().BoolVar(&stdin, "password-stdin", false, "read the password from standard input")
	add.Flags().BoolVar(&withToken, "token", false, "also create an API token")
	add.Flags().BoolVar(&service, "service", false, "service account without password (API tokens only)")
	add.Flags().StringArrayVar(&grants, "grant", nil, `permission, e.g. "records:write zones=example.com" (repeatable)`)

	passwd := &cobra.Command{
		Use:   "passwd NAME",
		Short: "Set a user's password (and enable the account)",
		Args:  cobra.ExactArgs(1),
		RunE: func(cmd *cobra.Command, args []string) error {
			a, err := openApp(true)
			if err != nil {
				return err
			}
			defer a.close()
			ctx := context.Background()
			u, err := a.store.GetUserByName(ctx, args[0])
			if err != nil {
				return fmt.Errorf("user %s: %w", args[0], err)
			}
			pw, err := readPassword("New password for "+u.Username+": ", stdin)
			if err != nil {
				return err
			}
			if u.PasswordHash, err = auth.HashPassword(pw); err != nil {
				return err
			}
			u.Disabled = false
			if err := a.store.UpdateUser(ctx, u); err != nil {
				return err
			}
			a.store.DeleteUserSessions(ctx, u.ID, "")
			a.eng.Audit(ctx, &auth.Principal{User: &model.User{Username: "local"}, Method: "local"}, "user.password", "user", u.ID, u.Username, "success", "jarld user passwd")
			fmt.Printf("Password of %s changed.\n", u.Username)
			return nil
		},
	}
	passwd.Flags().BoolVar(&stdin, "password-stdin", false, "read the password from standard input")

	token := &cobra.Command{
		Use:   "token NAME [TOKEN-NAME]",
		Short: "Create an API token for a user",
		Args:  cobra.RangeArgs(1, 2),
		RunE: func(cmd *cobra.Command, args []string) error {
			a, err := openApp(true)
			if err != nil {
				return err
			}
			defer a.close()
			ctx := context.Background()
			u, err := a.store.GetUserByName(ctx, args[0])
			if err != nil {
				return fmt.Errorf("user %s: %w", args[0], err)
			}
			name := "cli"
			if len(args) > 1 {
				name = args[1]
			}
			tok, err := createToken(ctx, a, u, name)
			if err != nil {
				return err
			}
			fmt.Println(tok)
			return nil
		},
	}
	list := &cobra.Command{
		Use:   "list",
		Short: "List users",
		RunE: func(cmd *cobra.Command, args []string) error {
			a, err := openApp(true)
			if err != nil {
				return err
			}
			defer a.close()
			users, err := a.store.ListUsers(context.Background())
			if err != nil {
				return err
			}
			for _, u := range users {
				flags := []string{}
				if u.Admin {
					flags = append(flags, "admin")
				}
				if u.Disabled {
					flags = append(flags, "disabled")
				}
				if !u.HasPassword {
					flags = append(flags, "no password")
				}
				var gs []string
				for _, g := range u.Grants {
					gs = append(gs, auth.FormatGrant(g))
				}
				fmt.Printf("%-20s %-25s %s\n", u.Username, strings.Join(flags, ","), strings.Join(gs, "; "))
			}
			return nil
		},
	}
	cmd.AddCommand(add, passwd, token, list)
	return cmd
}