wild-central/internal/api/v1/helpers.go

package v1

import (
	"fmt"
	"log/slog"
	"os"
	"path/filepath"
	"strings"
	"time"

	"github.com/wild-cloud/wild-central/internal/certbot"
	"github.com/wild-cloud/wild-central/internal/config"
	"github.com/wild-cloud/wild-central/internal/dnsmasq"
	"github.com/wild-cloud/wild-central/internal/domains"
	"github.com/wild-cloud/wild-central/internal/haproxy"
	"github.com/wild-cloud/wild-central/internal/network"
	"github.com/wild-cloud/wild-central/internal/sse"
)

// EnsureCentralDomain registers (or updates) Central's own domain as a domain
// so it participates in DNS, HAProxy routing, and TLS cert tracking like any
// other domain. Called on startup and when global config changes.
func (api *API) EnsureCentralDomain() {
	centralDomain := api.getCentralDomain()
	port := api.getRunningPort()
	backend := fmt.Sprintf("127.0.0.1:%d", port)

	// Check if the current registration already matches — skip if so.
	if centralDomain != "" {
		if existing, _ := api.domains.Get(centralDomain); existing != nil &&
			existing.Source == "central" &&
			existing.Backend.Address == backend {
			return
		}
	}

	// Clean up stale Central registrations (e.g. domain changed).
	doms, _ := api.domains.List()
	for _, dom := range doms {
		if dom.Source == "central" && dom.DomainName != centralDomain {
			_ = api.domains.Deregister(dom.DomainName)
		}
	}

	if centralDomain == "" {
		return
	}

	_ = api.domains.Register(domains.Domain{
		DomainName: centralDomain,
		Source:     "central",
		Backend: domains.Backend{
			Address: backend,
			Type:    domains.BackendHTTP,
		},
		TLS: domains.TLSTerminate,
	})
}

// Reconcile runs networking reconciliation immediately. Called on startup
// and automatically whenever a domain is registered/updated/deregistered.
func (api *API) Reconcile() {
	api.reconcileNetworking()
}

// reconcileNetworking reads all registered domains and regenerates
// dnsmasq DNS entries and HAProxy routes to match.
func (api *API) reconcileNetworking() {
	doms, err := api.domains.List()
	if err != nil {
		slog.Error("reconcile: failed to list domains", "error", err)
		return
	}

	globalCfg, err := config.LoadState(api.statePath())
	if err != nil {
		slog.Warn("reconcile: failed to load state, using empty", "error", err)
		globalCfg = &config.State{}
	}

	// Build HAProxy routes from registered domains
	var instanceRoutes []haproxy.InstanceRoute
	var httpRoutes []haproxy.HTTPRoute

	for _, dom := range doms {
		switch dom.EffectiveBackendType() {
		case domains.BackendTCPPassthrough:
			instanceRoutes = append(instanceRoutes, haproxy.InstanceRoute{
				Name:       dom.DomainName,
				Domain:     dom.DomainName,
				BackendIP:  extractHost(dom.EffectiveBackendAddress()),
				Subdomains: dom.Subdomains,
			})
		case domains.BackendDNSOnly:
			// dns-only: no gateway routing, just DNS resolution
		case domains.BackendHTTP:
			var routeBackends []haproxy.HTTPRouteBackend
			for _, r := range dom.EffectiveRoutes() {
				routeBackends = append(routeBackends, haproxy.HTTPRouteBackend{
					Paths:      r.Paths,
					Backend:    r.Backend.Address,
					HealthPath: r.Backend.Health,
					Headers:    r.Headers,
					IPAllow:    r.IPAllow,
				})
			}
			httpRoutes = append(httpRoutes, haproxy.HTTPRoute{
				Name:       dom.DomainName,
				Domain:     dom.DomainName,
				Subdomains: dom.Subdomains,
				Routes:     routeBackends,
			})
		}
	}

	// Clean up any 0-byte cert files that would poison HAProxy validation.
	// A 0-byte .pem in the certs directory causes the global `bind ssl crt`
	// to fail, taking down ALL L7 routes — not just the broken domain.
	certsDir := "/etc/haproxy/certs/"
	if entries, err := os.ReadDir(certsDir); err == nil {
		for _, e := range entries {
			if strings.HasSuffix(e.Name(), ".pem") {
				if info, err := e.Info(); err == nil && info.Size() == 0 {
					slog.Warn("reconcile: removing 0-byte cert file", "file", e.Name())
					_ = os.Remove(filepath.Join(certsDir, e.Name()))
				}
			}
		}
	}

	// Only include L7 HTTP routes for domains that have a valid cert.
	var activeHTTPRoutes []haproxy.HTTPRoute
	for _, route := range httpRoutes {
		if hasCertForDomain(certsDir, route.Domain) {
			activeHTTPRoutes = append(activeHTTPRoutes, route)
		} else {
			slog.Warn("reconcile: skipping L7 route (no cert)", "domain", route.Domain)
		}
	}

	// Generate and write HAProxy config
	genOpts := haproxy.GenerateOpts{
		HTTPRoutes: activeHTTPRoutes,
		CertsDir:   certsDir,
	}
	haproxyCfg := api.haproxy.GenerateWithOpts(instanceRoutes, nil, genOpts)

	if err := api.haproxy.WriteConfig(haproxyCfg); err != nil {
		// Validation failed — identify and exclude broken domains, then retry
		brokenDomains := haproxy.FindBrokenServices(haproxyCfg, haproxy.ParseValidationErrors(err.Error()))
		if len(brokenDomains) > 0 {
			slog.Error("reconcile: excluding broken domains and retrying",
				"broken", brokenDomains, "error", err)

			exclude := map[string]bool{}
			for _, d := range brokenDomains {
				exclude[d] = true
			}

			var filteredInstances []haproxy.InstanceRoute
			for _, r := range instanceRoutes {
				if !exclude[r.Domain] {
					filteredInstances = append(filteredInstances, r)
				}
			}
			var filteredHTTP []haproxy.HTTPRoute
			for _, r := range activeHTTPRoutes {
				if !exclude[r.Domain] {
					filteredHTTP = append(filteredHTTP, r)
				}
			}

			genOpts.HTTPRoutes = filteredHTTP
			haproxyCfg = api.haproxy.GenerateWithOpts(filteredInstances, nil, genOpts)
			if err := api.haproxy.WriteConfig(haproxyCfg); err != nil {
				slog.Error("reconcile: retry also failed", "error", err)
			} else {
				if err := api.haproxy.ReloadService(); err != nil {
					slog.Warn("reconcile: failed to reload HAProxy", "error", err)
				} else {
					api.broadcastHaproxyEvent("haproxy:config", "HAProxy config regenerated (excluded broken domains)")
				}
			}
		} else {
			slog.Error("reconcile: failed to write HAProxy config (no broken domains identified)", "error", err)
		}
	} else {
		if err := api.haproxy.ReloadService(); err != nil {
			slog.Warn("reconcile: failed to reload HAProxy", "error", err)
		} else {
			api.broadcastHaproxyEvent("haproxy:config", "HAProxy config regenerated from registered domains")
		}
	}

	// Build dnsmasq DNS entries from registered domains.
	// DNS target IP depends on backend type:
	//   tcp-passthrough/dns-only → backend IP (LAN traffic goes direct)
	//   http → Central IP (HAProxy terminates TLS)
	// Use the configured dnsmasq IP (eth0) as Central's IP, not auto-detected
	// (auto-detect can pick wlan0 if that's the default route).
	centralIP := globalCfg.Cloud.Dnsmasq.IP
	if centralIP == "" {
		centralIP, _ = network.GetWildCentralIP()
	}
	if centralIP == "" {
		centralIP = "127.0.0.1"
	}

	var dnsEntries []dnsmasq.DNSEntry
	for _, dom := range doms {
		if dom.DomainName == "" {
			continue
		}

		dnsIP := centralIP
		bt := dom.EffectiveBackendType()
		if bt == domains.BackendTCPPassthrough || bt == domains.BackendDNSOnly {
			dnsIP = extractHost(dom.EffectiveBackendAddress())
		}

		dnsEntries = append(dnsEntries, dnsmasq.DNSEntry{
			Domain: dom.DomainName,
			IP:     dnsIP,
		})
	}

	if err := api.dnsmasq.UpdateConfig(globalCfg, dnsEntries, true); err != nil {
		slog.Error("reconcile: failed to update dnsmasq", "error", err)
	} else {
		api.broadcastDnsmasqEvent("dnsmasq:config", "DNS config regenerated from registered domains")
	}

	// Ensure TLS certs exist for HTTP domains (where Central terminates TLS).
	// For domains under the gateway domain, a wildcard cert covers them all.
	// For others, provision individual certs.
	if len(httpRoutes) > 0 {
		api.ensureTLSCerts(globalCfg, doms)
	}

	// Nudge DDNS to pick up any domain changes (public toggle, new domains).
	// The runner reads fresh params on each check via its callback.
	api.ddns.Trigger()

	slog.Info("reconcile: networking updated",
		"domains", len(doms),
		"l4Routes", len(instanceRoutes),
		"l7Routes", len(httpRoutes),
	)
}

// ensureTLSCerts logs which certificates are missing for registered domains.
// Does NOT auto-provision — cert provisioning is user-initiated via the
// Certificates page. This just logs warnings so the operator knows what's needed.
func (api *API) ensureTLSCerts(globalCfg *config.State, doms []domains.Domain) {
	gatewayDomain := ""
	if parts := strings.SplitN(globalCfg.Cloud.Central.Domain, ".", 2); len(parts) == 2 {
		gatewayDomain = parts[1]
	}

	for _, dom := range doms {
		if dom.TLS != domains.TLSTerminate || dom.DomainName == "" {
			continue
		}

		// Check if covered by wildcard
		if gatewayDomain != "" && strings.HasSuffix(dom.DomainName, "."+gatewayDomain) {
			wildcardCert := certbot.HAProxyCertPath(gatewayDomain)
			if _, err := os.Stat(wildcardCert); err != nil {
				slog.Warn("reconcile/tls: no wildcard cert for gateway domain — provision via Certificates page",
					"domain", dom.DomainName, "needed", "*."+gatewayDomain)
			}
			continue
		}

		// Check individual cert
		if _, err := os.Stat(certbot.HAProxyCertPath(dom.DomainName)); err != nil {
			slog.Warn("reconcile/tls: no cert for domain — provision via Certificates page",
				"domain", dom.DomainName)
		}
	}
}

// hasCertForDomain checks if a valid (non-empty) cert exists for a domain —
// either an individual cert (<domain>.pem) or a wildcard cert that covers it.
func hasCertForDomain(_, domain string) bool {
	// Check individual cert
	if isValidCertFile(certbot.HAProxyCertPath(domain)) {
		return true
	}
	// Check wildcard certs — a *.example.com cert covers foo.example.com
	parts := strings.SplitN(domain, ".", 2)
	if len(parts) == 2 {
		wildcardBase := parts[1]
		if isValidCertFile(certbot.HAProxyCertPath(wildcardBase)) {
			return true
		}
	}
	return false
}

// isValidCertFile checks that a cert file exists and is non-empty.
func isValidCertFile(path string) bool {
	info, err := os.Stat(path)
	return err == nil && info.Size() > 0
}

// extractHost gets the host part from a host:port string
func extractHost(addr string) string {
	for i := len(addr) - 1; i >= 0; i-- {
		if addr[i] == ':' {
			return addr[:i]
		}
	}
	return addr
}

// broadcastDnsmasqEvent broadcasts SSE events for dnsmasq status changes
func (api *API) broadcastDnsmasqEvent(eventType string, message string) {
	if api.sseManager == nil {
		return
	}

	// Get current dnsmasq status
	status, err := api.dnsmasq.GetStatus()
	if err != nil {
		status = nil
	}

	event := &sse.Event{
		ID:           fmt.Sprintf("dnsmasq-%d", time.Now().UnixNano()),
		Type:         eventType,
		InstanceName: "global",
		Timestamp:    time.Now(),
		Data: map[string]any{
			"message": message,
			"status":  status,
		},
	}

	api.sseManager.Broadcast(event)
}

// broadcastHaproxyEvent broadcasts SSE events for HAProxy status changes
func (api *API) broadcastHaproxyEvent(eventType string, message string) {
	if api.sseManager == nil {
		return
	}

	event := &sse.Event{
		ID:           fmt.Sprintf("haproxy-%d", time.Now().UnixNano()),
		Type:         eventType,
		InstanceName: "global",
		Timestamp:    time.Now(),
		Data: map[string]any{
			"message": message,
		},
	}

	api.sseManager.Broadcast(event)
}

// broadcastCentralStatusEvent broadcasts SSE events for central status changes
func (api *API) broadcastCentralStatusEvent(startTime time.Time) {
	if api.sseManager == nil {
		return
	}

	uptime := time.Since(startTime)

	event := &sse.Event{
		ID:           fmt.Sprintf("central-%d", time.Now().UnixNano()),
		Type:         "central:status",
		InstanceName: "global",
		Timestamp:    time.Now(),
		Data: map[string]any{
			"status":        "running",
			"version":       api.version,
			"uptime":        uptime.String(),
			"uptimeSeconds": int(uptime.Seconds()),
		},
	}

	api.sseManager.Broadcast(event)
}