all: implement self-upgrade flag from API

So upgrading don't have to be initiated manually, helping large
deployments to upgrade to latest ctrld version easily.

cmd/cli/cli.go

@@ -25,7 +25,7 @@ import (
 	"sync/atomic"
 	"time"
 
-	"github.com/Masterminds/semver"
+	"github.com/Masterminds/semver/v3"
 	"github.com/cuonglm/osinfo"
 	"github.com/go-playground/validator/v10"
 	"github.com/kardianos/service"

cmd/cli/prog.go

@@ -11,6 +11,7 @@ import (
 	"net/netip"
 	"net/url"
 	"os"
+	"os/exec"
 	"runtime"
 	"slices"
 	"sort"
@@ -21,6 +22,7 @@ import (
 	"syscall"
 	"time"
 
+	"github.com/Masterminds/semver/v3"
 	"github.com/kardianos/service"
 	"github.com/rs/zerolog"
 	"github.com/spf13/viper"
@@ -304,6 +306,16 @@ func (p *prog) apiConfigReload() {
 	logger := mainLog.Load().With().Str("mode", "api-reload").Logger()
 	logger.Debug().Msg("starting custom config reload timer")
 	lastUpdated := time.Now().Unix()
+	curVerStr := curVersion()
+	curVer, err := semver.NewVersion(curVerStr)
+	isStable := curVer != nil && curVer.Prerelease() == ""
+	if err != nil || !isStable {
+		l := mainLog.Load().Warn()
+		if err != nil {
+			l = l.Err(err)
+		}
+		l.Msgf("current version is not stable, skipping self-upgrade: %s", curVerStr)
+	}
 
 	doReloadApiConfig := func(forced bool, logger zerolog.Logger) {
 		resolverConfig, err := controld.FetchResolverConfig(cdUID, rootCmd.Version, cdDev)
@@ -313,6 +325,11 @@ func (p *prog) apiConfigReload() {
 			return
 		}
 
+		// Performing self-upgrade check for production version.
+		if isStable {
+			selfUpgradeCheck(resolverConfig.Ctrld.VersionTarget, curVer, &logger)
+		}
+
 		if resolverConfig.DeactivationPin != nil {
 			newDeactivationPin := *resolverConfig.DeactivationPin
 			curDeactivationPin := cdDeactivationPin.Load()
@@ -1422,6 +1439,46 @@ func selfUninstallCheck(uninstallErr error, p *prog, logger zerolog.Logger) {
 	}
 }
 
+// selfUpgradeCheck checks if the version target vt is greater
+// than the current one cv, perform self-upgrade then.
+//
+// The callers must ensure curVer and logger are non-nil.
+func selfUpgradeCheck(vt string, cv *semver.Version, logger *zerolog.Logger) {
+	if vt == "" {
+		logger.Debug().Msg("no version target set, skipped checking self-upgrade")
+		return
+	}
+	vts := vt
+	if !strings.HasPrefix(vts, "v") {
+		vts = "v" + vts
+	}
+	targetVer, err := semver.NewVersion(vts)
+	if err != nil {
+		logger.Warn().Err(err).Msgf("invalid target version, skipped self-upgrade: %s", vt)
+		return
+	}
+	if !targetVer.GreaterThan(cv) {
+		logger.Debug().
+			Str("target", vt).
+			Str("current", cv.String()).
+			Msgf("target version is not greater than current one, skipped self-upgrade")
+		return
+	}
+
+	exe, err := os.Executable()
+	if err != nil {
+		mainLog.Load().Error().Err(err).Msg("failed to get executable path, skipped self-upgrade")
+		return
+	}
+	cmd := exec.Command(exe, "upgrade", "prod", "-vv")
+	cmd.SysProcAttr = sysProcAttrForSelfUpgrade()
+	if err := cmd.Start(); err != nil {
+		mainLog.Load().Error().Err(err).Msg("failed to start self-upgrade")
+		return
+	}
+	mainLog.Load().Debug().Msgf("self-upgrade triggered, version target: %s", vts)
+}
+
 // leakOnUpstreamFailure reports whether ctrld should initiate a recovery flow
 // when upstream failures occur.
 func (p *prog) leakOnUpstreamFailure() bool {

cmd/cli/self_upgrade_others.go

@@ -0,0 +1,12 @@
+//go:build !windows
+
+package cli
+
+import (
+	"syscall"
+)
+
+// sysProcAttrForSelfUpgrade returns *syscall.SysProcAttr instance for running self-upgrade command.
+func sysProcAttrForSelfUpgrade() *syscall.SysProcAttr {
+	return &syscall.SysProcAttr{Setsid: true}
+}

cmd/cli/self_upgrade_windows.go

@@ -0,0 +1,18 @@
+package cli
+
+import (
+	"syscall"
+)
+
+// From: https://learn.microsoft.com/en-us/windows/win32/procthread/process-creation-flags?redirectedfrom=MSDN
+
+// SYSCALL_CREATE_NO_WINDOW set flag to run process without a console window.
+const SYSCALL_CREATE_NO_WINDOW = 0x08000000
+
+// sysProcAttrForSelfUpgrade returns *syscall.SysProcAttr instance for running self-upgrade command.
+func sysProcAttrForSelfUpgrade() *syscall.SysProcAttr {
+	return &syscall.SysProcAttr{
+		CreationFlags: syscall.CREATE_NEW_PROCESS_GROUP | SYSCALL_CREATE_NO_WINDOW,
+		HideWindow:    true,
+	}
+}

cmd/cli/service.go

@@ -4,10 +4,12 @@ import (
 	"bytes"
 	"errors"
 	"fmt"
+	"io"
 	"os"
 	"os/exec"
 	"runtime"
 
+	"github.com/coreos/go-systemd/v22/unit"
 	"github.com/kardianos/service"
 
 	"github.com/Control-D-Inc/ctrld/internal/router"
@@ -132,6 +134,59 @@ func (s *systemd) Status() (service.Status, error) {
 	return s.Service.Status()
 }
 
+func (s *systemd) Start() error {
+	const systemdUnitFile = "/etc/systemd/system/ctrld.service"
+	f, err := os.Open(systemdUnitFile)
+	if err != nil {
+		return err
+	}
+	defer f.Close()
+	if opts, change := ensureSystemdKillMode(f); change {
+		mode := os.FileMode(0644)
+		buf, err := io.ReadAll(unit.Serialize(opts))
+		if err != nil {
+			return err
+		}
+		if err := os.WriteFile(systemdUnitFile, buf, mode); err != nil {
+			return err
+		}
+		if out, err := exec.Command("systemctl", "daemon-reload").CombinedOutput(); err != nil {
+			return fmt.Errorf("systemctl daemon-reload failed: %w\n%s", err, string(out))
+		}
+		mainLog.Load().Debug().Msg("set KillMode=process successfully")
+	}
+	return s.Service.Start()
+}
+
+// ensureSystemdKillMode ensure systemd unit file is configured with KillMode=process.
+// This is necessary for running self-upgrade flow.
+func ensureSystemdKillMode(r io.Reader) (opts []*unit.UnitOption, change bool) {
+	opts, err := unit.DeserializeOptions(r)
+	if err != nil {
+		mainLog.Load().Error().Err(err).Msg("failed to deserialize options")
+		return
+	}
+	change = true
+	needKillModeOpt := true
+	killModeOpt := unit.NewUnitOption("Service", "KillMode", "process")
+	for _, opt := range opts {
+		if opt.Match(killModeOpt) {
+			needKillModeOpt = false
+			change = false
+			break
+		}
+		if opt.Section == killModeOpt.Section && opt.Name == killModeOpt.Name {
+			opt.Value = killModeOpt.Value
+			needKillModeOpt = false
+			break
+		}
+	}
+	if needKillModeOpt {
+		opts = append(opts, killModeOpt)
+	}
+	return opts, change
+}
+
 func newLaunchd(s service.Service) *launchd {
 	return &launchd{
 		Service:      s,

cmd/cli/service_test.go

@@ -0,0 +1,28 @@
+package cli
+
+import (
+	"strings"
+	"testing"
+)
+
+func Test_ensureSystemdKillMode(t *testing.T) {
+	tests := []struct {
+		name       string
+		unitFile   string
+		wantChange bool
+	}{
+		{"no KillMode", "[Service]\nExecStart=/bin/sleep 1", true},
+		{"not KillMode=process", "[Service]\nExecStart=/bin/sleep 1\nKillMode=mixed", true},
+		{"KillMode=process", "[Service]\nExecStart=/bin/sleep 1\nKillMode=process", false},
+		{"invalid unit file", "[Service\nExecStart=/bin/sleep 1\nKillMode=process", false},
+	}
+	for _, tc := range tests {
+		tc := tc
+		t.Run(tc.name, func(t *testing.T) {
+			t.Parallel()
+			if _, change := ensureSystemdKillMode(strings.NewReader(tc.unitFile)); tc.wantChange != change {
+				t.Errorf("ensureSystemdKillMode(%q) = %v, want %v", tc.unitFile, change, tc.wantChange)
+			}
+		})
+	}
+}