steward/cmd/stewardd/main.go

// Command stewardd is Steward's daemon: the DHCP server, its REST API and
// the web UI. It also has the commands that work on the database directly
// (init, user, backup, restore, check).
package main

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

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

	"steward/internal/auth"
	"steward/internal/certs"
	"steward/internal/config"
	"steward/internal/dhcpd"
	"steward/internal/engine"
	"steward/internal/jarl"
	"steward/internal/model"
	"steward/internal/server"
	"steward/internal/store"
	"steward/internal/validate"
	"steward/internal/version"
	"steward/internal/web"
)

var configPath string

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

func rootCmd() *cobra.Command {
	root := &cobra.Command{
		Use:           "stewardd",
		Short:         "Steward's DHCP server, API and web UI",
		SilenceUsage:  true,
		SilenceErrors: true,
		Version:       version.String(),
	}
	root.PersistentFlags().StringVarP(&configPath, "config", "c", config.DefaultPath, "configuration file")
	root.AddCommand(serveCmd(), initCmd(), userCmd(), checkCmd(), backupCmd(), restoreCmd(), configCmd(), versionCmd())
	return root
}

func newLogger(cfg *config.Config, w io.Writer) *slog.Logger {
	var level slog.Level
	switch cfg.Log.Level {
	case "debug":
		level = slog.LevelDebug
	case "warn":
		level = slog.LevelWarn
	case "error":
		level = slog.LevelError
	}
	opts := &slog.HandlerOptions{Level: level}
	if cfg.Log.Format == "json" {
		return slog.New(slog.NewJSONHandler(w, opts))
	}
	// The journal adds its own 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))
}

type app struct {
	cfg   *config.Config
	log   *slog.Logger
	store *store.Store
}

func openApp(allowMissing bool) (*app, error) {
	cfg, err := config.Load(configPath, allowMissing)
	if err != nil {
		return nil, fmt.Errorf("loading the configuration: %w", err)
	}
	log := newLogger(cfg, os.Stderr)
	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 the database %s: %w", cfg.Database, err)
	}
	return &app{cfg: cfg, log: log, store: st}, nil
}

// hostEnv describes this machine's interfaces for validation.
func hostEnv(h *dhcpd.HostInterfaces) func() *validate.Env {
	return func() *validate.Env {
		all, err := h.All()
		if err != nil {
			return nil
		}
		return &validate.Env{Interfaces: all}
	}
}

func serveCmd() *cobra.Command {
	return &cobra.Command{
		Use:   "serve",
		Short: "Serve DHCP, the API and the web UI",
		Args:  cobra.NoArgs,
		RunE: func(cmd *cobra.Command, args []string) error {
			a, err := openApp(true)
			if err != nil {
				return err
			}
			defer a.store.Close()
			release, err := lockDatabase(a.cfg.Database)
			if err != nil {
				return err
			}
			defer release()
			ctx, stop := signal.NotifyContext(context.Background(), os.Interrupt, syscall.SIGTERM)
			defer stop()
			a.log.Info("starting", "version", version.String(), "database", a.cfg.Database)
			if users, err := a.store.ListUsers(ctx); err == nil && len(users) == 0 {
				a.log.Warn("there are no users yet; create an administrator with 'stewardd init' (root can use 'steward' on the local socket meanwhile)")
			}
			eng, err := engine.New(ctx, a.store, a.log)
			if err != nil {
				return err
			}
			host := &dhcpd.HostInterfaces{}
			eng.Env = hostEnv(host)
			pinger := &dhcpd.ICMPPinger{Log: a.log}
			d := dhcpd.New(a.log, a.store, host, pinger)
			if err := d.LoadLeases(ctx); err != nil {
				return fmt.Errorf("reading the leases: %w", err)
			}
			d.SetConfig(eng.State())
			eng.OnChange = d.SetConfig
			for _, p := range eng.Problems() {
				a.log.Warn("configuration problem: "+p.Message, "object", p.Object, "code", p.Code)
			}
			var acme *certs.Manager
			if a.cfg.TLS.ACME.Domain != "" {
				tok, err := a.cfg.Jarl.ReadToken()
				if err != nil {
					return err
				}
				jc, err := jarl.New(a.cfg.Jarl.Server, tok, a.cfg.Jarl.CACert, a.cfg.Jarl.Insecure)
				if err != nil {
					return err
				}
				acme = certs.New(ctx, certs.Config{ACME: a.cfg.TLS.ACME, Version: version.Short()}, jc, a.store, a.log)
				go acme.Run(ctx)
			}
			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, Socket: a.cfg.Socket, PingProblem: pinger.Problem, ACME: acme,
				ExcludeInterfaces: a.cfg.Interfaces.Exclude,
			}, eng, d, host, a.log, web.FS())
			go d.Run(ctx)
			hup := make(chan os.Signal, 1)
			signal.Notify(hup, syscall.SIGHUP)
			go func() {
				for range hup {
					sdNotify("RELOADING=1")
					srv.ReloadTLS()
					sdNotify("READY=1")
				}
			}()
			err = srv.Serve(ctx, func() { sdNotify("READY=1") })
			sdNotify("STOPPING=1")
			a.log.Info("stopped")
			return err
		},
	}
}

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("the passwords don't match")
	}
	return string(b), nil
}

// readLine reads one line a byte at a time, so nothing after it is
// buffered away from the password prompt that follows.
func readLine(r io.Reader) string {
	var b []byte
	one := make([]byte, 1)
	for {
		n, err := r.Read(one)
		if n == 1 {
			if one[0] == '\n' {
				break
			}
			b = append(b, one[0])
		}
		if err != nil {
			break
		}
	}
	return strings.TrimRight(string(b), "\r")
}

func localPrincipal(u *model.User) *auth.Principal {
	return &auth.Principal{User: u, Method: "local", IP: "local"}
}

func audit(ctx context.Context, st *store.Store, p *auth.Principal, action, object string) {
	ev := &model.AuditEvent{Action: action, Object: object, Result: "success", Auth: p.Method, Username: p.Name()}
	if p.User != nil {
		ev.UserID = p.User.ID
	}
	st.AddAudit(ctx, ev)
}

func createToken(ctx context.Context, st *store.Store, u *model.User, name string) (string, error) {
	secret, pid, hash := auth.NewToken()
	if err := st.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
	cmd := &cobra.Command{
		Use:   "init",
		Short: "Create the database and the first administrator",
		Long: `Creates Steward's database with a first administrator.

Steward starts with no subnets: add them with 'steward subnet add', in the
web UI, or import a Kea configuration with 'steward import kea'.`,
		Args: cobra.NoArgs,
		RunE: func(cmd *cobra.Command, args []string) error {
			a, err := openApp(true)
			if err != nil {
				return err
			}
			defer a.store.Close()
			ctx := context.Background()
			users, err := a.store.ListUsers(ctx)
			if err != nil {
				return err
			}
			if len(users) > 0 {
				return fmt.Errorf("Steward is already set up (%d users); add users with 'stewardd user add' or 'steward user add'", len(users))
			}
			// Asked for when not given, on a terminal.
			if !cmd.Flags().Changed("user") && !stdin && term.IsTerminal(int(os.Stdin.Fd())) {
				fmt.Fprint(os.Stderr, "The first administrator's user name [admin]: ")
				if name := strings.TrimSpace(readLine(os.Stdin)); name != "" {
					username = name
				}
			}
			if err := server.ValidUsername(username); err != nil {
				return err
			}
			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 stewardd init"}
			if err := a.store.CreateUser(ctx, u); err != nil {
				return err
			}
			audit(ctx, a.store, localPrincipal(u), "system.init", "user "+u.Username)
			fmt.Printf("Created the administrator %q in %s.\n", u.Username, a.cfg.Database)
			if withToken {
				tok, err := createToken(ctx, a.store, u, "cli")
				if err != nil {
					return err
				}
				fmt.Printf("\nAPI token for the CLI (shown only once):\n\n  %s\n", tok)
			}
			fmt.Println("\nNext steps:")
			fmt.Println("  1. Start the daemon:          sudo systemctl enable --now steward")
			fmt.Println("  2. Add a subnet:              sudo steward subnet add 10.0.10.0/24 --interface eth0 --pool 10.0.10.100-10.0.10.199 --option routers=10.0.10.1")
			fmt.Println("     or import Kea's setup:     sudo steward import kea /etc/kea/kea-dhcp4.conf --leases /var/lib/kea/kea-leases4.csv")
			fmt.Printf("  3. Open the web UI:           http://%s/ (see listen in %s)\n", a.cfg.Listen[0], configPath)
			return nil
		},
	}
	cmd.Flags().StringVarP(&username, "user", "u", "admin", "the administrator's name (best your own: each person should have their own account); asked for on a terminal")
	cmd.Flags().BoolVar(&stdin, "password-stdin", false, "read the password from standard input")
	cmd.Flags().BoolVar(&withToken, "token", false, "also make an API token for the CLI")
	return cmd
}

func userCmd() *cobra.Command {
	cmd := &cobra.Command{Use: "user", Short: "Manage users directly in the database (when the API can't be used)"}
	var admin, stdin bool
	add := &cobra.Command{
		Use: "add NAME", Short: "Add 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.store.Close()
			if err := server.ValidUsername(args[0]); err != nil {
				return err
			}
			pw, err := readPassword("Password for "+args[0]+": ", stdin)
			if err != nil {
				return err
			}
			hash, err := auth.HashPassword(pw)
			if err != nil {
				return err
			}
			u := &model.User{Username: args[0], PasswordHash: hash, Admin: admin}
			if err := a.store.CreateUser(context.Background(), u); err != nil {
				if errors.Is(err, store.ErrConflict) {
					return fmt.Errorf("user %s already exists", args[0])
				}
				return err
			}
			audit(context.Background(), a.store, localPrincipal(&model.User{Username: "root"}), "user.create", "user "+u.Username)
			fmt.Printf("added user %s\n", u.Username)
			return nil
		},
	}
	add.Flags().BoolVar(&admin, "admin", false, "an administrator")
	add.Flags().BoolVar(&stdin, "password-stdin", false, "read the password from standard input")
	cmd.AddCommand(add)
	pw := &cobra.Command{
		Use: "password NAME", Short: "Set a user's password (and log them out)", Args: cobra.ExactArgs(1),
		RunE: func(cmd *cobra.Command, args []string) error {
			a, err := openApp(true)
			if err != nil {
				return err
			}
			defer a.store.Close()
			ctx := context.Background()
			u, err := a.store.GetUserByName(ctx, args[0])
			if err != nil {
				return fmt.Errorf("no user %s", args[0])
			}
			p, err := readPassword("New password for "+u.Username+": ", stdin)
			if err != nil {
				return err
			}
			if u.PasswordHash, err = auth.HashPassword(p); err != nil {
				return err
			}
			u.Disabled = false
			if err := a.store.UpdateUser(ctx, u); err != nil {
				return err
			}
			a.store.DeleteUserSessions(ctx, u.ID, "")
			audit(ctx, a.store, localPrincipal(&model.User{Username: "root"}), "user.password", "user "+u.Username)
			fmt.Printf("set the password of %s\n", u.Username)
			return nil
		},
	}
	pw.Flags().BoolVar(&stdin, "password-stdin", false, "read the password from standard input")
	cmd.AddCommand(pw)
	cmd.AddCommand(&cobra.Command{
		Use: "token NAME TOKEN-NAME", Short: "Make an API token for a user (printed once)", Args: cobra.ExactArgs(2),
		RunE: func(cmd *cobra.Command, args []string) error {
			a, err := openApp(true)
			if err != nil {
				return err
			}
			defer a.store.Close()
			ctx := context.Background()
			u, err := a.store.GetUserByName(ctx, args[0])
			if err != nil {
				return fmt.Errorf("no user %s", args[0])
			}
			tok, err := createToken(ctx, a.store, u, args[1])
			if err != nil {
				return err
			}
			audit(ctx, a.store, localPrincipal(&model.User{Username: "root"}), "token.create", "token "+u.Username+"/"+args[1])
			fmt.Println(tok)
			return nil
		},
	})
	return cmd
}

func checkCmd() *cobra.Command {
	return &cobra.Command{
		Use:   "check",
		Short: "Check the configuration in the database against this machine",
		Args:  cobra.NoArgs,
		RunE: func(cmd *cobra.Command, args []string) error {
			a, err := openApp(false)
			if err != nil {
				return err
			}
			defer a.store.Close()
			eng, err := engine.New(context.Background(), a.store, a.log)
			if err != nil {
				return err
			}
			eng.Env = hostEnv(&dhcpd.HostInterfaces{})
			ps := eng.Problems()
			for _, p := range ps {
				fmt.Printf("%s: %s: %s [%s]\n", p.Severity, p.Object, p.Message, p.Code)
			}
			if validate.HasErrors(ps) {
				return errors.New("the configuration has errors")
			}
			if len(ps) == 0 {
				fmt.Println("ok: no problems")
			}
			return nil
		},
	}
}

// ---- backups

// backupName is a backup's file name; the time in it orders and prunes
// backups.
const backupPrefix, backupSuffix = "steward-", ".db"

func backupCmd() *cobra.Command {
	return &cobra.Command{
		Use:   "backup [FILE]",
		Short: "Write a backup of the database (and delete old ones)",
		Long: `Writes a consistent copy of the database, also while stewardd runs. Without
FILE it goes to backup.dir as steward-<time>.db, and backups older than
backup.retention are deleted (keeping at least backup.keep_at_least).`,
		Args: cobra.MaximumNArgs(1),
		RunE: func(cmd *cobra.Command, args []string) error {
			cfg, err := config.Load(configPath, true)
			if err != nil {
				return err
			}
			if _, err := os.Stat(cfg.Database); err != nil {
				return fmt.Errorf("no database at %s yet (run 'stewardd init')", cfg.Database)
			}
			st, err := store.OpenExisting(cfg.Database)
			if err != nil {
				return err
			}
			defer st.Close()
			dest := ""
			if len(args) == 1 {
				dest = args[0]
			} else {
				if err := os.MkdirAll(cfg.Backup.Dir, 0o700); err != nil {
					return err
				}
				dest = filepath.Join(cfg.Backup.Dir, backupPrefix+time.Now().UTC().Format("20060102T150405Z")+backupSuffix)
			}
			tmp := dest + ".tmp"
			os.Remove(tmp)
			if err := st.Backup(context.Background(), tmp); err != nil {
				return fmt.Errorf("writing the backup: %w", err)
			}
			if err := os.Rename(tmp, dest); err != nil {
				return err
			}
			fmt.Println(dest)
			if len(args) == 0 {
				return prune(cfg)
			}
			return nil
		},
	}
}

func prune(cfg *config.Config) error {
	entries, err := os.ReadDir(cfg.Backup.Dir)
	if err != nil {
		return err
	}
	type bk struct {
		name string
		t    time.Time
	}
	var list []bk
	for _, e := range entries {
		n := e.Name()
		if !strings.HasPrefix(n, backupPrefix) || !strings.HasSuffix(n, backupSuffix) {
			continue
		}
		t, err := time.Parse("20060102T150405Z", strings.TrimSuffix(strings.TrimPrefix(n, backupPrefix), backupSuffix))
		if err != nil {
			continue
		}
		list = append(list, bk{n, t})
	}
	sort.Slice(list, func(i, j int) bool { return list[i].t.After(list[j].t) })
	cut := time.Now().Add(-time.Duration(cfg.Backup.Retention))
	for i, b := range list {
		if i < cfg.Backup.KeepAtLeast || b.t.After(cut) {
			continue
		}
		os.Remove(filepath.Join(cfg.Backup.Dir, b.name))
	}
	return nil
}

func restoreCmd() *cobra.Command {
	var yes bool
	cmd := &cobra.Command{
		Use:   "restore FILE",
		Short: "Replace the database with a backup (stewardd must be stopped)",
		Args:  cobra.ExactArgs(1),
		RunE: func(cmd *cobra.Command, args []string) error {
			cfg, err := config.Load(configPath, true)
			if err != nil {
				return err
			}
			release, err := lockDatabase(cfg.Database)
			if err != nil {
				return fmt.Errorf("%w; stop it first: systemctl stop steward", err)
			}
			defer release()
			src, err := store.OpenExisting(args[0])
			if err != nil {
				return fmt.Errorf("%s isn't a database: %w", args[0], err)
			}
			v, err := src.Version(context.Background())
			if err != nil {
				src.Close()
				return fmt.Errorf("%s isn't a Steward database: %w", args[0], err)
			}
			var ok string
			if err := src.DB().QueryRow(`PRAGMA integrity_check`).Scan(&ok); err != nil || ok != "ok" {
				src.Close()
				return fmt.Errorf("%s fails SQLite's integrity check: %s %v", args[0], ok, err)
			}
			src.Close()
			if v > store.SchemaVersion() {
				return fmt.Errorf("the backup is from a newer Steward (schema %d, this one has %d)", v, store.SchemaVersion())
			}
			if !yes {
				fmt.Fprintf(os.Stderr, "Replace %s with %s? [y/N] ", cfg.Database, args[0])
				var a string
				fmt.Scanln(&a)
				if a != "y" && a != "yes" {
					return errors.New("not restored")
				}
			}
			stamp := time.Now().UTC().Format("20060102T150405Z")
			for _, ext := range []string{"", "-wal", "-shm"} {
				if _, err := os.Stat(cfg.Database + ext); err == nil {
					if err := os.Rename(cfg.Database+ext, cfg.Database+".before-restore-"+stamp+ext); err != nil {
						return err
					}
				}
			}
			b, err := os.ReadFile(args[0])
			if err != nil {
				return err
			}
			if err := os.WriteFile(cfg.Database, b, 0o640); err != nil {
				return err
			}
			st, err := store.Open(cfg.Database)
			if err != nil {
				return err
			}
			defer st.Close()
			audit(context.Background(), st, localPrincipal(&model.User{Username: "root"}), "system.restore", args[0])
			fmt.Printf("restored %s; the old database is %s.before-restore-%s\n", args[0], cfg.Database, stamp)
			return nil
		},
	}
	cmd.Flags().BoolVarP(&yes, "yes", "y", false, "don't ask")
	return cmd
}

func configCmd() *cobra.Command {
	var example bool
	cmd := &cobra.Command{
		Use: "config", Short: "Show the configuration in use (--example: a commented example)", Args: cobra.NoArgs,
		RunE: func(cmd *cobra.Command, args []string) error {
			if example {
				fmt.Print(config.Example)
				return nil
			}
			cfg, err := config.Load(configPath, true)
			if err != nil {
				return err
			}
			fmt.Printf("listen:   %s\nsocket:   %s\ndatabase: %s\nbackups:  %s\n", strings.Join(cfg.Listen, ", "), cfg.Socket, cfg.Database, cfg.Backup.Dir)
			return nil
		},
	}
	cmd.Flags().BoolVar(&example, "example", false, "print a commented example configuration")
	return cmd
}

func versionCmd() *cobra.Command {
	return &cobra.Command{
		Use: "version", Short: "Print the version", Args: cobra.NoArgs,
		Run: func(cmd *cobra.Command, args []string) { fmt.Println("stewardd", version.String()) },
	}
}