checker-tls/checker/collect.go

package checker

import (
	"context"
	"crypto/sha256"
	"encoding/hex"
	"fmt"
	"log"
	"sort"
	"strings"
	"sync"
	"time"

	sdk "git.happydns.org/checker-sdk-go/checker"
	"git.happydns.org/checker-tls/contract"
)

func (p *tlsProvider) Collect(ctx context.Context, opts sdk.CheckerOptions) (any, error) {
	raw, ok := sdk.GetOption[[]sdk.DiscoveryEntry](opts, OptionEndpoints)
	if !ok {
		return nil, fmt.Errorf("no discovery entries in options: did the host wire AutoFillDiscoveryEntries?")
	}

	timeoutMs := sdk.GetIntOption(opts, OptionProbeTimeoutMs, DefaultProbeTimeoutMs)
	if timeoutMs <= 0 {
		timeoutMs = DefaultProbeTimeoutMs
	}
	timeout := time.Duration(timeoutMs) * time.Millisecond
	enumerate := sdk.GetBoolOption(opts, OptionEnumerateCiphers, false)

	entries, warnings := contract.ParseEntries(raw)
	for _, w := range warnings {
		log.Printf("checker-tls: discarding malformed entry: %v", w)
	}
	// An empty entry set is not an error: it is the steady state on any
	// target where no producer has published yet, and the first run after
	// a fresh publication when the producer hasn't finished its own cycle.
	// The rule surfaces this as StatusUnknown rather than StatusError so a
	// freshly-enrolled domain doesn't flap red.
	if len(entries) == 0 {
		return &TLSData{Probes: map[string]TLSProbe{}, CollectedAt: time.Now()}, nil
	}

	probes := make(map[string]TLSProbe, len(entries))
	var mu sync.Mutex
	var wg sync.WaitGroup
	sem := make(chan struct{}, MaxConcurrentProbes)
dispatch:
	for _, e := range entries {
		select {
		case sem <- struct{}{}:
		case <-ctx.Done():
			break dispatch
		}
		wg.Add(1)
		go func() {
			defer wg.Done()
			defer func() { <-sem }()
			pr := probe(ctx, e.Endpoint, timeout)
			log.Printf("checker-tls: %s %s:%d → tls=%s handshake_ok=%t elapsed=%dms err=%q",
				pr.Type, pr.Host, pr.Port, pr.TLSVersion, pr.TLSHandshakeOK, pr.ElapsedMS, pr.Error)
			if enumerate && pr.TLSHandshakeOK {
				enumRes, skipReason := enumerateEndpoint(ctx, e.Endpoint, enumerationBudget)
				switch {
				case enumRes != nil && enumRes.Skipped != "":
					pr.Enum = enumRes
					log.Printf("checker-tls: enum %s:%d → error: %s (duration=%dms)",
						pr.Host, pr.Port, enumRes.Skipped, enumRes.DurationMS)
				case enumRes != nil:
					pr.Enum = enumRes
					log.Printf("checker-tls: enum %s:%d → versions=%d duration=%dms",
						pr.Host, pr.Port, len(enumRes.Versions), enumRes.DurationMS)
				case skipReason != "":
					log.Printf("checker-tls: enum %s:%d → skipped: %s",
						pr.Host, pr.Port, skipReason)
				}
			}
			mu.Lock()
			probes[e.Ref] = pr
			mu.Unlock()
		}()
	}
	wg.Wait()

	return &TLSData{
		Probes:      probes,
		CollectedAt: time.Now(),
	}, nil
}

// ShareKey implements sdk.ObservationSharer. A TLS probe result depends only on
// the set of endpoints actually dialed and the probe knobs, never on which
// domain or service published them: the observation is a map keyed by each
// endpoint's contract Ref (host|port|effective SNI|STARTTLS|require), so two
// targets that resolve to the same endpoint set produce identical probes. This
// lets the host dial a host:port once and serve every target that points at it
// instead of re-handshaking per record — the highest-value case here, since
// dane, xmpp, srv, dav, … all funnel endpoints into this single checker.
//
// The probe timeout and cipher-enumeration flag are folded in because they
// change what is collected (a tighter timeout can fail a slow handshake;
// enumeration adds the Enum block). Inputs that yield no probable endpoint
// return "" so the host falls back to the default per-target caching.
func (p *tlsProvider) ShareKey(opts sdk.CheckerOptions) (string, error) {
	raw, ok := sdk.GetOption[[]sdk.DiscoveryEntry](opts, OptionEndpoints)
	if !ok {
		return "", nil
	}

	entries, _ := contract.ParseEntries(raw)
	if len(entries) == 0 {
		return "", nil
	}

	refs := make([]string, 0, len(entries))
	for _, e := range entries {
		refs = append(refs, e.Ref)
	}
	sort.Strings(refs)

	timeoutMs := sdk.GetIntOption(opts, OptionProbeTimeoutMs, DefaultProbeTimeoutMs)
	if timeoutMs <= 0 {
		timeoutMs = DefaultProbeTimeoutMs
	}
	enumerate := sdk.GetBoolOption(opts, OptionEnumerateCiphers, false)

	h := sha256.Sum256(fmt.Appendf(nil, "%d|%t|%s", timeoutMs, enumerate, strings.Join(refs, ",")))
	return "tls:" + hex.EncodeToString(h[:8]), nil
}