felhom-controller/controller/internal/stacks/setup_gate.go

package stacks

import (
	"context"
	"encoding/json"
	"fmt"
	"io"
	"net/http"
	"os"
	"path/filepath"
	"regexp"
	"sort"
	"strconv"
	"strings"
	"time"

	"gopkg.in/yaml.v3"
)

// ── The setup gate (v0.280.0, `09` §3 decision 46) ─────────────────────────────────────────────────────
//
// 34 catalog apps let the FIRST VISITOR create the admin, and every app is on the internet from its first
// minute. So an app whose template says `setup_gate: true` is installed CLOSED: traefik sends each request to
// the controller first (forwardAuth), and the controller lets it through only for the household — a browser
// that holds a valid dashboard session (the handshake lives in internal/web/setup_gate.go). The gate OPENS
// when the app's own status says the first admin exists (`setup_done_probe:`), or when the household presses
// "Done, I set it up". Opening REMOVES the gate's traefik file: the app's own docker-label router is then the
// only one, exactly as if it had never been gated.
//
//   setup_gate: true
//   setup_done_probe:                                  # optional; without it, the household's button opens it
//     url: http://n8n:5678/rest/settings               # read on the docker network, never through traefik
//     field: data.userManagement.showSetupOnFirstLoad  # dotted JSON path
//     done: "false"                                    # the field's value once the setup is done, as text
//
// Order is load-bearing (spike F2, audits/login-gate-2026-09-29/B): the gate file is written BEFORE the app's
// first start. traefik holds a file router whose service does not exist yet and enables it the moment the
// app's service appears — measured: 0 app answers to a stranger across ~530 polls during two installs.
// A gate that cannot be written refuses the install; a gated app is never published open.
//
// The record (`setup_gate:` in app.yaml) is a life record: it survives a controller restart (the file on disk
// is reconciled from it) and a restore (carryLifeRecords). An install that LOADS kept data never gates — the
// data comes back with its admin.
// Pinned by internal/stacks/setup_gate_test.go.

// SetupDoneProbe is `.felhom.yml`'s `setup_done_probe:`.
type SetupDoneProbe struct {
	URL   string `yaml:"url" json:"url"`
	Field string `yaml:"field" json:"field"`
	Done  string `yaml:"done" json:"done"`
	// DoneStatus (v0.282.0, R-715): a non-200 HTTP status that itself means "set up" (gramps-web answers 405).
	DoneStatus int `yaml:"done_status,omitempty" json:"done_status,omitempty"`
}

// Setup gate states.
const (
	SetupGateClosed = "closed"
	SetupGateOpen   = "open"
	// Who opened it.
	SetupGateByProbe     = "probe"
	SetupGateByHousehold = "household"
)

// SetupGateRecord is app.yaml's `setup_gate:`.
type SetupGateRecord struct {
	State    string   `yaml:"state" json:"state"`
	Since    string   `yaml:"since" json:"since"`
	Hosts    []string `yaml:"hosts,omitempty" json:"hosts,omitempty"`
	OpenedAt string   `yaml:"opened_at,omitempty" json:"opened_at,omitempty"`
	OpenedBy string   `yaml:"opened_by,omitempty" json:"opened_by,omitempty"`
	// SignupOpenUntil (v0.281.0, decision 47): the household opened sign-up until this time (signup_block.go).
	SignupOpenUntil string `yaml:"signup_open_until,omitempty" json:"signup_open_until,omitempty"`
	// NativeLock (v0.282.0): the app's own sign-up switch — "applied", "lifted" (the window), or "" (never set).
	NativeLock string `yaml:"native_lock,omitempty" json:"native_lock,omitempty"`
}

// Closed reports whether the gate stands.
func (r *SetupGateRecord) Closed() bool { return r != nil && r.State == SetupGateClosed }

// setupGateAuthURL is where traefik asks. The controller sits on traefik-public as felhom-controller (wireController).
const setupGateAuthURL = "http://felhom-controller:8080/__felhom_gate/auth"

// setupGatePriority beats every docker-label router (traefik's default priority is the rule's length). The rule's
// length is ADDED so an app's path-scoped router (adventurelog's backend) still wins over its host-only one, as
// it does without the gate.
const setupGatePriority = 100000

// gateRouter is one traefik router an app publishes, as its compose labels define it.
type gateRouter struct {
	Name         string
	Rule         string
	Service      string
	CertResolver string
}

var hostInRule = regexp.MustCompile("Host\\(`([^`]+)`\\)")

// expandComposeVars fills ${NAME} and ${NAME:-default} from env, as compose does for a label value.
func expandComposeVars(s string, env map[string]string) string {
	return os.Expand(s, func(v string) string {
		if name, def, ok := strings.Cut(v, ":-"); ok {
			if val := env[name]; val != "" {
				return val
			}
			return def
		}
		return env[v]
	})
}

// gateRoutersFromCompose reads the routers an app publishes from its compose labels, filled with the app's
// env. A router with no `service` label takes its container's only service; two services and no label is
// refused (the gate would not know where to send the household).
func gateRoutersFromCompose(composePath string, env map[string]string) ([]gateRouter, error) {
	data, err := os.ReadFile(composePath)
	if err != nil {
		return nil, err
	}
	var doc struct {
		Services map[string]struct {
			Labels interface{} `yaml:"labels"`
		} `yaml:"services"`
	}
	if err := yaml.Unmarshal(data, &doc); err != nil {
		return nil, fmt.Errorf("compose: %w", err)
	}
	var out []gateRouter
	names := make([]string, 0, len(doc.Services))
	for n := range doc.Services {
		names = append(names, n)
	}
	sort.Strings(names)
	for _, svc := range names {
		labels := map[string]string{}
		switch l := doc.Services[svc].Labels.(type) {
		case []interface{}:
			for _, e := range l {
				k, v, _ := strings.Cut(fmt.Sprint(e), "=")
				labels[strings.TrimSpace(k)] = strings.TrimSpace(v)
			}
		case map[string]interface{}:
			for k, v := range l {
				labels[k] = fmt.Sprint(v)
			}
		}
		if strings.ToLower(labels["traefik.enable"]) != "true" {
			continue
		}
		var services []string
		routers := map[string]*gateRouter{}
		for k, v := range labels {
			parts := strings.Split(k, ".")
			if len(parts) < 5 || parts[0] != "traefik" || parts[1] != "http" {
				continue
			}
			switch parts[2] {
			case "services":
				if !containsStr(services, parts[3]) {
					services = append(services, parts[3])
				}
			case "routers":
				r := routers[parts[3]]
				if r == nil {
					r = &gateRouter{Name: parts[3]}
					routers[parts[3]] = r
				}
				switch strings.Join(parts[4:], ".") {
				case "rule":
					r.Rule = expandComposeVars(v, env)
				case "service":
					r.Service = expandComposeVars(v, env)
				case "tls.certresolver":
					r.CertResolver = v
				}
			}
		}
		rnames := make([]string, 0, len(routers))
		for n := range routers {
			rnames = append(rnames, n)
		}
		sort.Strings(rnames)
		for _, n := range rnames {
			r := routers[n]
			if r.Rule == "" {
				continue
			}
			if r.Service == "" {
				if len(services) != 1 {
					return nil, fmt.Errorf("router %s (service %s) names no traefik service and its container has %d", n, svc, len(services))
				}
				r.Service = services[0]
			}
			out = append(out, *r)
		}
	}
	if len(out) == 0 {
		return nil, fmt.Errorf("the compose file publishes no traefik router")
	}
	return out, nil
}

// gateHosts is every host the routers match.
func gateHosts(rs []gateRouter) []string {
	var hosts []string
	for _, r := range rs {
		for _, m := range hostInRule.FindAllStringSubmatch(r.Rule, -1) {
			h := strings.ToLower(m[1])
			if !containsStr(hosts, h) {
				hosts = append(hosts, h)
			}
		}
	}
	sort.Strings(hosts)
	return hosts
}

// renderSetupGate is the traefik file-provider config that puts the gate in front of every router the app
// publishes: same rule, higher priority, forwardAuth, then the app's own docker service.
func renderSetupGate(name string, rs []gateRouter) string {
	var b strings.Builder
	mw := "felhom-setup-gate-" + name
	fmt.Fprintf(&b, "# Setup gate for %s — managed by felhom-controller (`09` §3 decision 46).\n", name)
	b.WriteString("# The app is closed to everyone but the household until its first setup is done; then this file is removed.\n")
	b.WriteString("http:\n  middlewares:\n")
	fmt.Fprintf(&b, "    %s:\n      forwardAuth:\n        address: %q\n", mw, setupGateAuthURL)
	b.WriteString("  routers:\n")
	for _, r := range rs {
		fmt.Fprintf(&b, "    %s-%s:\n", mw, r.Name)
		fmt.Fprintf(&b, "      rule: %q\n", r.Rule)
		fmt.Fprintf(&b, "      priority: %d\n", setupGatePriority+len(r.Rule))
		b.WriteString("      entryPoints:\n        - websecure\n")
		if r.CertResolver != "" {
			fmt.Fprintf(&b, "      tls:\n        certResolver: %s\n", r.CertResolver)
		} else {
			b.WriteString("      tls: {}\n")
		}
		fmt.Fprintf(&b, "      middlewares:\n        - %s@file\n", mw)
		fmt.Fprintf(&b, "      service: %q\n", r.Service+"@docker")
	}
	return b.String()
}

func (m *Manager) setupGateDir() string {
	return filepath.Join(m.cfg.Paths.StacksDir, "traefik", "dynamic")
}

func (m *Manager) setupGatePath(name string) string {
	return filepath.Join(m.setupGateDir(), "setup-gate-"+name+".yml")
}

// writeSetupGate writes (or refreshes) the app's gate file. Returns the hosts it covers.
func (m *Manager) writeSetupGate(name, composePath string, env map[string]string) ([]string, error) {
	rs, err := gateRoutersFromCompose(composePath, env)
	if err != nil {
		return nil, err
	}
	if err := os.MkdirAll(m.setupGateDir(), 0o755); err != nil {
		return nil, err
	}
	want := renderSetupGate(name, rs)
	p := m.setupGatePath(name)
	if cur, err := os.ReadFile(p); err == nil && string(cur) == want {
		return gateHosts(rs), nil
	}
	tmp := p + ".tmp"
	if err := os.WriteFile(tmp, []byte(want), 0o644); err != nil {
		return nil, err
	}
	if err := os.Rename(tmp, p); err != nil {
		return nil, err
	}
	return gateHosts(rs), nil
}

func (m *Manager) removeSetupGateFile(name string) error {
	err := os.Remove(m.setupGatePath(name))
	if err != nil && !os.IsNotExist(err) {
		return err
	}
	return nil
}

// prepareSetupGate is DeployStack's step for a `setup_gate: true` template on a FRESH install: the file first,
// then the record the caller saves with the app.
func (m *Manager) prepareSetupGate(name, composePath string, env map[string]string) (*SetupGateRecord, error) {
	hosts, err := m.writeSetupGate(name, composePath, env)
	if err != nil {
		return nil, err
	}
	m.logger.Printf("[INFO] [stacks] %s: setup gate CLOSED before the first start — only the household reaches %v until the first setup is done", name, hosts)
	return &SetupGateRecord{State: SetupGateClosed, Since: m.now().UTC().Format(time.RFC3339), Hosts: hosts}, nil
}

// OpenSetupGate opens an app's gate: the record first, then the file (a failed removal is retried by the
// reconcile; the reverse order could re-gate an opened app). by = SetupGateByProbe | SetupGateByHousehold.
func (m *Manager) OpenSetupGate(name, by string) error {
	st, ok := m.GetStack(name)
	if !ok {
		return fmt.Errorf("stack %q not found", name)
	}
	if st.AppConfig == nil || !st.AppConfig.SetupGate.Closed() {
		return ErrSetupGateNotClosed
	}
	dir := filepath.Dir(st.ComposePath)
	now := m.now().UTC().Format(time.RFC3339)
	// Decision 47: the sign-up block goes up BEFORE the gate comes down, so there is no moment where a stranger
	// can sign up. Cannot write it → the gate stays closed (the loop or the next press tries again).
	block := strings.TrimSpace(st.Meta.SignupBlock)
	if block != "" {
		if err := m.writeSignupBlock(name, st.AppConfig.SetupGate.Hosts, block); err != nil {
			return fmt.Errorf("setup gate %s: the sign-up block could not be written, so the gate stays closed: %w", name, err)
		}
	}
	opened := false
	m.mutateAppConfig(name, dir, "setup_gate", func(cfg *AppConfig) bool {
		if !cfg.SetupGate.Closed() {
			return false
		}
		cfg.SetupGate.State, cfg.SetupGate.OpenedAt, cfg.SetupGate.OpenedBy = SetupGateOpen, now, by
		opened = true
		return true
	})
	if !opened {
		if block != "" {
			_ = m.removeSignupBlockFile(name) // still gated: the household may still need the sign-up address
		}
		return fmt.Errorf("setup gate %s: the record could not be written", name)
	}
	if err := m.removeSetupGateFile(name); err != nil {
		m.logger.Printf("[ERROR] [stacks] %s: setup gate opened but its traefik file could not be removed (%v) — the reconcile retries", name, err)
	}
	m.logger.Printf("[INFO] [stacks] %s: setup gate OPENED by %s — the app is reached as without a gate", name, by)
	// v0.282.0 (decision 47): the app's own sign-up switch, after the block is up and the gate is down. One restart.
	if st.Meta.AfterSetup != nil {
		m.goNativeLock(name, true, "the gate opened ("+by+")")
	}
	return nil
}

// ErrSetupGateNotClosed: the app has no closed gate (never gated, or already open).
var ErrSetupGateNotClosed = fmt.Errorf("the app has no closed setup gate")

// SetupGateHost maps a host to the app that owns it and whether that app's gate is closed.
func (m *Manager) SetupGateHost(host string) (name string, closed bool, found bool) {
	host = strings.ToLower(host)
	m.mu.RLock()
	defer m.mu.RUnlock()
	// R-741: a closed install hold answers first (its routers outrank the gate's); an app with both is closed while
	// either is.
	for n, st := range m.stacks {
		if st.AppConfig == nil || !st.AppConfig.InstallHold.Closed() {
			continue
		}
		if containsStr(st.AppConfig.InstallHold.Hosts, host) {
			return n, true, true
		}
	}
	for n, st := range m.stacks {
		if st.AppConfig == nil || st.AppConfig.SetupGate == nil {
			continue
		}
		if containsStr(st.AppConfig.SetupGate.Hosts, host) {
			return n, st.AppConfig.SetupGate.Closed(), true
		}
	}
	return "", false, false
}

// setupGateProbeFetch reads a probe URL: the HTTP status and the body (a seam: tests never reach a network).
var setupGateProbeFetch = func(url string) (int, []byte, error) {
	c := &http.Client{Timeout: 5 * time.Second}
	resp, err := c.Get(url)
	if err != nil {
		return 0, nil, err
	}
	defer resp.Body.Close()
	b, err := io.ReadAll(io.LimitReader(resp.Body, 1<<20))
	return resp.StatusCode, b, err
}

// probeOnce asks the app's own status once. v0.282.0 (R-715): `done_status:` — an app that answers a fixed non-200
// status once it is set up (gramps-web: 405 "Users already exist") — counts as done on that status alone; any other
// non-200 is "cannot read" (fail closed). A 200 is read as JSON: `field` is a dotted path whose parts may be list
// indexes (`setup.0.status` — ghost; `0.done` — home-assistant), compared as text with `done`.
func probeOnce(p *SetupDoneProbe) (done bool, got string, err error) {
	status, body, err := setupGateProbeFetch(p.URL)
	if err != nil {
		return false, "", err
	}
	if p.DoneStatus != 0 && status == p.DoneStatus {
		return true, fmt.Sprintf("HTTP %d", status), nil
	}
	if status != http.StatusOK {
		return false, fmt.Sprintf("HTTP %d", status), fmt.Errorf("HTTP %d", status)
	}
	if p.Field == "" {
		return false, "no field", fmt.Errorf("the probe names no field")
	}
	done, got = probeSaysDone(body, p.Field, p.Done)
	return done, got, nil
}

// probeSaysDone reads the field at the dotted path (a numeric part indexes a list) and compares its text form with
// done. Anything it cannot read is "not done" — the gate stays closed (fail closed).
func probeSaysDone(body []byte, field, done string) (bool, string) {
	var v interface{}
	if err := json.Unmarshal(body, &v); err != nil {
		return false, "not JSON"
	}
	for _, k := range strings.Split(field, ".") {
		switch cur := v.(type) {
		case map[string]interface{}:
			nv, ok := cur[k]
			if !ok {
				return false, "no field " + field
			}
			v = nv
		case []interface{}:
			n, err := strconv.Atoi(k)
			if err != nil || n < 0 || n >= len(cur) {
				return false, "no field " + field
			}
			v = cur[n]
		default:
			return false, "no field " + field
		}
	}
	got := fmt.Sprint(v)
	return got == done, got
}

// SetupGateTick is one pass of the gate's loop: every closed gate's file exists; every app whose gate is not
// closed has none (a removed app, an opened gate whose removal failed); a closed gate whose app is running and
// whose probe says done opens.
func (m *Manager) SetupGateTick() {
	type item struct {
		name, dir, compose string
		rec                *SetupGateRecord
		probe              *SetupDoneProbe
		running            bool
	}
	var items []item
	keep := map[string]bool{}
	m.mu.RLock()
	for n, st := range m.stacks {
		if !st.Deployed || st.AppConfig == nil || !st.AppConfig.SetupGate.Closed() {
			continue
		}
		rec := *st.AppConfig.SetupGate
		items = append(items, item{name: n, dir: filepath.Dir(st.ComposePath), compose: st.ComposePath, rec: &rec,
			probe: st.Meta.SetupDoneProbe, running: st.State == StateRunning || st.State == StateUnhealthy})
		keep[n] = true
	}
	m.mu.RUnlock()

	// Stale files: a gate file whose app is not closed-gated any more.
	if ents, err := os.ReadDir(m.setupGateDir()); err == nil {
		for _, e := range ents {
			n := e.Name()
			if !strings.HasPrefix(n, "setup-gate-") || !strings.HasSuffix(n, ".yml") {
				continue
			}
			app := strings.TrimSuffix(strings.TrimPrefix(n, "setup-gate-"), ".yml")
			if !keep[app] {
				if err := m.removeSetupGateFile(app); err == nil {
					m.logger.Printf("[INFO] [stacks] %s: removed a setup-gate file for an app whose gate is not closed", app)
				}
			}
		}
	}

	for _, it := range items {
		cfg := LoadAppConfigDecrypted(it.dir, m.encKey)
		if cfg != nil {
			if _, err := m.writeSetupGate(it.name, it.compose, cfg.Env); err != nil {
				m.logger.Printf("[ERROR] [stacks] %s: the setup gate's traefik file could not be (re)written: %v", it.name, err)
			}
		}
		if it.probe == nil || it.probe.URL == "" || !it.running {
			continue
		}
		done, got, err := probeOnce(it.probe)
		if err != nil {
			if m.isDebug() {
				m.logger.Printf("[DEBUG] [stacks] %s: setup probe unreadable (%v) — gate stays closed", it.name, err)
			}
			continue
		}
		if m.isDebug() {
			m.logger.Printf("[DEBUG] [stacks] %s: setup probe %s = %q (done when %q)", it.name, it.probe.Field, got, it.probe.Done)
		}
		if done {
			if err := m.OpenSetupGate(it.name, SetupGateByProbe); err != nil {
				m.logger.Printf("[ERROR] [stacks] %s: the probe says the setup is done but the gate did not open: %v", it.name, err)
			}
		}
	}
	m.reconcileSignupBlocks()
	m.installHoldTick()
}

// RunSetupGateLoop runs SetupGateTick every interval until ctx ends.
func (m *Manager) RunSetupGateLoop(ctx context.Context, interval time.Duration) {
	t := time.NewTicker(interval)
	defer t.Stop()
	m.SetupGateTick()
	for {
		select {
		case <-ctx.Done():
			return
		case <-t.C:
			m.SetupGateTick()
		}
	}
}

// MarkDefaultLoginChanged records the household's word that it changed the template's known default login by
// hand (R-710). The page then stops naming the default. It changes nothing in the app.
func (m *Manager) MarkDefaultLoginChanged(name, by string) error {
	st, ok := m.GetStack(name)
	if !ok || !st.Deployed {
		return fmt.Errorf("stack %q is not installed", name)
	}
	rec := &DefaultLoginRecord{ChangedAt: m.now().UTC().Format(time.RFC3339), By: by}
	done := false
	m.mutateAppConfig(name, filepath.Dir(st.ComposePath), "default_login", func(cfg *AppConfig) bool {
		cfg.DefaultLogin = rec
		done = true
		return true
	})
	if !done {
		return fmt.Errorf("%s: app.yaml could not be read", name)
	}
	m.logger.Printf("[INFO] [stacks] %s: the household says it changed the default login by hand — the page stops naming it", name)
	if err := m.OpenInstallHold(name, InstallHoldByHousehold); err != nil { // R-741
		m.logger.Printf("[ERROR] [stacks] %s: %v — the loop retries", name, err)
	}
	return nil
}