checker-alias/checker/collect.go

package checker

import (
	"context"
	"encoding/json"
	"fmt"
	"strings"
	"sync"

	"github.com/miekg/dns"

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

// Collect runs the alias testsuite and returns an *AliasData populated with
// findings, a resolution chain, and optional coexistence / DNSSEC observations.
func (p *aliasProvider) Collect(ctx context.Context, opts sdk.CheckerOptions) (any, error) {
	owner, err := resolveOwner(opts)
	if err != nil {
		return nil, err
	}

	maxChain := sdk.GetIntOption(opts, "maxChainLength", 8)
	minTTL := uint32(sdk.GetIntOption(opts, "minTargetTTL", 60))
	requireTarget := sdk.GetBoolOption(opts, "requireResolvableTarget", true)
	allowApexCNAME := sdk.GetBoolOption(opts, "allowApexCNAME", false)
	recognizeApex := sdk.GetBoolOption(opts, "recognizeApexFlattening", true)

	data := &AliasData{Owner: owner}

	resolver := systemResolver()

	// 1. Find apex and authoritative servers.
	apex, servers, err := findApex(ctx, owner, resolver)
	if err != nil {
		data.Findings = append(data.Findings, AliasFinding{
			Code:     "alias_no_apex",
			Severity: SeverityCrit,
			Message:  fmt.Sprintf("could not locate zone apex of %s: %v", owner, err),
			Subject:  owner,
			Hint:     "Check that the parent delegation exists and that the zone is published.",
		})
		return data, nil
	}
	data.Apex = apex
	data.AuthServers = servers
	data.OwnerIsApex = lowerFQDN(owner) == lowerFQDN(apex)

	// 2. Detect DNAME substitutions from owner up to apex (exclusive of apex).
	data.DNAMESubstitutions = collectDNAMEs(ctx, servers, owner, apex)

	// 3. Walk the CNAME/DNAME chain.
	chainCtx := &chainCtx{
		data:         data,
		maxLen:       maxChain,
		minTTL:       minTTL,
		servers:      servers,
		apex:         apex,
		seenOwners:   map[string]bool{},
		recFallback:  resolver,
		followTarget: requireTarget,
	}
	chainCtx.walk(ctx, owner)

	// 4. Apex checks (flattening, CNAME-at-apex coexistence).
	if data.OwnerIsApex {
		checkApex(ctx, data, servers, apex, allowApexCNAME, recognizeApex)
	}

	// 5. Coexistence at owner (applies at any level, not just apex).
	checkCoexistence(ctx, data, servers, owner, allowApexCNAME, recognizeApex)

	// 6. DNSSEC checks.
	checkDNSSEC(ctx, data, servers, apex, owner)

	// 7. Chain-level validations (loops, length, TTL, target resolvability).
	validateChain(data, requireTarget)

	return data, nil
}

// resolveOwner derives the FQDN to check from the auto-filled options. The
// "service" option takes precedence (it carries a dns.CNAME whose owner is
// authoritative); otherwise we fall back to subdomain + domain_name.
func resolveOwner(opts sdk.CheckerOptions) (string, error) {
	if svcMsg, ok := sdk.GetOption[serviceMessage](opts, "service"); ok && len(svcMsg.Service) > 0 {
		var c cnameService
		if err := json.Unmarshal(svcMsg.Service, &c); err == nil && c.Record != nil && c.Record.Hdr.Name != "" {
			return lowerFQDN(c.Record.Hdr.Name), nil
		}
	}

	parent, _ := sdk.GetOption[string](opts, "domain_name")
	sub, _ := sdk.GetOption[string](opts, "subdomain")
	if parent == "" {
		return "", fmt.Errorf("missing 'domain_name' option")
	}
	parent = strings.TrimSuffix(parent, ".")
	if sub == "" || sub == "@" {
		return lowerFQDN(parent), nil
	}
	sub = strings.TrimSuffix(sub, ".")
	return lowerFQDN(sub + "." + parent), nil
}

// chainCtx carries the mutable state of a chain walk.
type chainCtx struct {
	data         *AliasData
	maxLen       int
	minTTL       uint32
	servers      []string
	apex         string
	seenOwners   map[string]bool
	recFallback  string
	followTarget bool
}

// walk follows CNAME/DNAME hops starting from name. It writes hops into
// data.Chain and may add findings.
func (c *chainCtx) walk(ctx context.Context, name string) {
	current := lowerFQDN(name)
	currentServers := c.servers

	for i := 0; i <= c.maxLen+1; i++ {
		if c.seenOwners[current] {
			c.data.Findings = append(c.data.Findings, AliasFinding{
				Code:     "alias_loop",
				Severity: SeverityCrit,
				Message:  fmt.Sprintf("chain loops back to %s", current),
				Subject:  current,
				Hint:     "Break the loop by pointing the last CNAME at an A/AAAA-bearing name.",
			})
			c.data.FinalTarget = current
			return
		}
		c.seenOwners[current] = true

		if i > c.maxLen {
			c.data.Findings = append(c.data.Findings, AliasFinding{
				Code:     "alias_chain_too_long",
				Severity: SeverityCrit,
				Message:  fmt.Sprintf("chain exceeds %d hops at %s; many resolvers will give up", c.maxLen, current),
				Subject:  current,
				Hint:     "Flatten intermediate CNAMEs so that the chain is at most a few hops long.",
			})
			c.data.FinalTarget = current
			return
		}

		q := dns.Question{Name: current, Qtype: dns.TypeCNAME, Qclass: dns.ClassINET}
		r, server, err := c.queryFor(ctx, currentServers, q)
		if err != nil {
			c.data.Findings = append(c.data.Findings, AliasFinding{
				Code:     "alias_query_failed",
				Severity: SeverityWarn,
				Message:  fmt.Sprintf("CNAME query for %s failed: %v", current, err),
				Subject:  current,
			})
			c.data.FinalTarget = current
			return
		}

		if r.Rcode != dns.RcodeSuccess {
			c.data.Rcode = rcodeText(r.Rcode)
			c.data.Findings = append(c.data.Findings, AliasFinding{
				Code:     "alias_rcode",
				Severity: SeverityCrit,
				Message:  fmt.Sprintf("server answered %s for %s", c.data.Rcode, current),
				Subject:  current,
				Hint:     "Ensure the zone publishes the expected record; NXDOMAIN/SERVFAIL mid-chain breaks the alias.",
			})
			c.data.FinalTarget = current
			return
		}

		cname, synthesizedFromDNAME, ttl := extractCNAME(r, current)
		if cname == "" {
			// No CNAME at this name: terminal hop, resolve A/AAAA.
			c.data.Chain = append(c.data.Chain, ChainHop{
				Owner:  current,
				Kind:   KindTarget,
				Server: server,
			})
			c.data.FinalTarget = current
			c.resolveFinal(ctx, current, currentServers)
			return
		}

		target := lowerFQDN(cname)
		kind := KindCNAME
		if synthesizedFromDNAME {
			kind = KindDNAME
		}
		c.data.Chain = append(c.data.Chain, ChainHop{
			Owner:       current,
			Kind:        kind,
			Target:      target,
			TTL:         ttl,
			Server:      server,
			Synthesized: synthesizedFromDNAME,
		})
		if ttl < c.minTTL {
			c.data.Findings = append(c.data.Findings, AliasFinding{
				Code:     "alias_low_ttl",
				Severity: SeverityWarn,
				Message:  fmt.Sprintf("hop %s → %s has TTL %ds (< %d)", current, target, ttl, c.minTTL),
				Subject:  current,
				Hint:     "Raise the CNAME TTL to improve cache efficiency (5-15 minutes is a common floor).",
			})
		}

		// Re-evaluate servers for the next hop: if target leaves the apex,
		// we need its own authoritative servers. Out-of-zone targets are
		// resolved via the system resolver (recursive path).
		if isSubdomain(target, c.apex) {
			currentServers = c.servers
		} else {
			zone, _, _ := findApex(ctx, target, c.recFallback)
			ns, err := resolveZoneNSAddrs(ctx, zone)
			if err != nil || len(ns) == 0 {
				currentServers = []string{c.recFallback}
			} else {
				currentServers = ns
			}
		}
		current = target
	}
}

// queryFor sends q, retrying via the recursive resolver if the authoritative
// set is empty (useful for foreign targets).
func (c *chainCtx) queryFor(ctx context.Context, servers []string, q dns.Question) (*dns.Msg, string, error) {
	if len(servers) == 0 {
		r, err := recursiveExchange(ctx, c.recFallback, q)
		return r, c.recFallback, err
	}
	return queryAtAuth(ctx, "", servers, q)
}

// extractCNAME returns the first CNAME target matched for owner, and reports
// whether it was synthesized from a DNAME present in the same response.
func extractCNAME(r *dns.Msg, owner string) (target string, fromDNAME bool, ttl uint32) {
	for _, rr := range r.Answer {
		if c, ok := rr.(*dns.CNAME); ok && strings.EqualFold(dns.Fqdn(c.Hdr.Name), dns.Fqdn(owner)) {
			target = c.Target
			ttl = c.Hdr.Ttl
			break
		}
	}
	if target == "" {
		return "", false, 0
	}
	for _, rr := range r.Answer {
		if _, ok := rr.(*dns.DNAME); ok {
			fromDNAME = true
			break
		}
	}
	return
}

// resolveFinal fetches A/AAAA of the final target and records them.
func (c *chainCtx) resolveFinal(ctx context.Context, name string, servers []string) {
	var wg sync.WaitGroup
	var finalA, finalAAAA []string
	var rcode string

	wg.Add(2)
	go func() {
		defer wg.Done()
		q := dns.Question{Name: dns.Fqdn(name), Qtype: dns.TypeA, Qclass: dns.ClassINET}
		var r *dns.Msg
		var err error
		if len(servers) > 0 {
			r, _, err = queryAtAuth(ctx, "", servers, q)
		} else {
			r, err = recursiveExchange(ctx, c.recFallback, q)
		}
		if err == nil && r != nil {
			if r.Rcode != dns.RcodeSuccess {
				rcode = rcodeText(r.Rcode)
			}
			for _, rr := range r.Answer {
				if a, ok := rr.(*dns.A); ok {
					finalA = append(finalA, a.A.String())
				}
			}
		}
	}()
	go func() {
		defer wg.Done()
		q := dns.Question{Name: dns.Fqdn(name), Qtype: dns.TypeAAAA, Qclass: dns.ClassINET}
		var r *dns.Msg
		var err error
		if len(servers) > 0 {
			r, _, err = queryAtAuth(ctx, "", servers, q)
		} else {
			r, err = recursiveExchange(ctx, c.recFallback, q)
		}
		if err == nil && r != nil {
			for _, rr := range r.Answer {
				if aaaa, ok := rr.(*dns.AAAA); ok {
					finalAAAA = append(finalAAAA, aaaa.AAAA.String())
				}
			}
		}
	}()
	wg.Wait()

	c.data.FinalA = append(c.data.FinalA, finalA...)
	c.data.FinalAAAA = append(c.data.FinalAAAA, finalAAAA...)
	if rcode != "" {
		c.data.Rcode = rcode
	}
}

// collectDNAMEs queries every label from owner up to (but excluding) apex for
// a DNAME record, returning any substitutions found.
func collectDNAMEs(ctx context.Context, servers []string, owner, apex string) []ChainHop {
	labels := dns.SplitDomainName(owner)
	apexLabels := dns.SplitDomainName(apex)
	stop := max(len(labels)-len(apexLabels), 0)

	results := make([][]ChainHop, stop)
	var wg sync.WaitGroup
	wg.Add(stop)
	for i := range stop {
		go func() {
			defer wg.Done()
			name := dns.Fqdn(strings.Join(labels[i:], "."))
			q := dns.Question{Name: name, Qtype: dns.TypeDNAME, Qclass: dns.ClassINET}
			r, server, err := queryAtAuth(ctx, "", servers, q)
			if err != nil || r == nil || r.Rcode != dns.RcodeSuccess {
				return
			}
			for _, rr := range r.Answer {
				if d, ok := rr.(*dns.DNAME); ok {
					results[i] = append(results[i], ChainHop{
						Owner:  lowerFQDN(d.Hdr.Name),
						Kind:   KindDNAME,
						Target: lowerFQDN(d.Target),
						TTL:    d.Hdr.Ttl,
						Server: server,
					})
				}
			}
		}()
	}
	wg.Wait()

	var out []ChainHop
	for _, hops := range results {
		out = append(out, hops...)
	}
	return out
}

// checkApex verifies that a CNAME at apex does not break SOA/NS, and
// detects ALIAS/ANAME provider-side flattening.
func checkApex(ctx context.Context, data *AliasData, servers []string, apex string, allowApexCNAME, recognizeApex bool) {
	// Collect A/AAAA at apex.
	var hasA bool
	var hasAMu sync.Mutex
	var wg sync.WaitGroup
	wg.Add(2)
	for _, qt := range []uint16{dns.TypeA, dns.TypeAAAA} {
		go func() {
			defer wg.Done()
			q := dns.Question{Name: apex, Qtype: qt, Qclass: dns.ClassINET}
			r, _, err := queryAtAuth(ctx, "", servers, q)
			if err != nil || r == nil {
				return
			}
			for _, rr := range r.Answer {
				switch rr.(type) {
				case *dns.A, *dns.AAAA:
					hasAMu.Lock()
					hasA = true
					hasAMu.Unlock()
				}
			}
		}()
	}
	wg.Wait()

	// CNAME at apex?
	hasCNAME := false
	for _, h := range data.Chain {
		if h.Kind == KindCNAME && lowerFQDN(h.Owner) == lowerFQDN(apex) {
			hasCNAME = true
			break
		}
	}

	if hasCNAME {
		sev := SeverityCrit
		if allowApexCNAME {
			sev = SeverityWarn
		}
		data.Findings = append(data.Findings, AliasFinding{
			Code:     "alias_cname_at_apex",
			Severity: sev,
			Message:  fmt.Sprintf("CNAME at apex %s conflicts with the SOA/NS records a zone apex must carry (RFC 1912 §2.4)", apex),
			Subject:  apex,
			Hint:     "Use the provider's ALIAS/ANAME flattening, an HTTP redirect, or move content to a sub-label such as www.",
		})
	}

	if hasA && !hasCNAME {
		// A present at apex alongside SOA/NS, classic ALIAS/ANAME flattening.
		data.ApexFlattening = true
		if recognizeApex {
			data.Findings = append(data.Findings, AliasFinding{
				Code:     "alias_apex_flattening",
				Severity: SeverityInfo,
				Message:  fmt.Sprintf("apex %s serves A/AAAA directly (provider-side ALIAS/ANAME flattening)", apex),
				Subject:  apex,
				Hint:     "Keep the upstream target's TTL in mind: apex A/AAAA will only update as fast as the provider re-flattens.",
			})
		}
	}
}

// checkCoexistence verifies that a CNAME at owner is the only record type
// present (RFC 1034 §3.6.2, RFC 2181 §10.1).
func checkCoexistence(ctx context.Context, data *AliasData, servers []string, owner string, allowApexCNAME, recognizeApex bool) {
	hasCNAME := false
	for _, h := range data.Chain {
		if h.Kind == KindCNAME && lowerFQDN(h.Owner) == lowerFQDN(owner) {
			hasCNAME = true
			break
		}
	}
	if !hasCNAME {
		return
	}

	// Query a handful of common sibling types at owner.
	siblings := []uint16{
		dns.TypeA, dns.TypeAAAA, dns.TypeMX, dns.TypeTXT,
		dns.TypeNS, dns.TypeSRV, dns.TypeCAA,
	}
	seen := map[string]uint32{}
	var seenMu sync.Mutex
	var wg sync.WaitGroup
	wg.Add(len(siblings))
	for _, qt := range siblings {
		go func() {
			defer wg.Done()
			q := dns.Question{Name: owner, Qtype: qt, Qclass: dns.ClassINET}
			r, _, err := queryAtAuth(ctx, "", servers, q)
			if err != nil || r == nil {
				return
			}
			// A synthesized CNAME from DNAME will be present in Answer for any
			// type; only count answers whose owner matches and whose type is qt.
			for _, rr := range r.Answer {
				if rr.Header().Rrtype != qt {
					continue
				}
				if !strings.EqualFold(dns.Fqdn(rr.Header().Name), dns.Fqdn(owner)) {
					continue
				}
				seenMu.Lock()
				seen[dns.TypeToString[qt]] = rr.Header().Ttl
				seenMu.Unlock()
				break
			}
		}()
	}
	wg.Wait()

	// Apex with ALIAS/ANAME flattening is a known exception when requested.
	isApex := lowerFQDN(owner) == lowerFQDN(data.Apex)

	for t, ttl := range seen {
		// A/AAAA at apex alongside a CNAME is impossible in a standard zone;
		// a provider may still serve it through flattening. Still report it
		// as critical, two different owners cannot legally exist.
		if isApex && (t == "A" || t == "AAAA") && recognizeApex && data.ApexFlattening {
			continue
		}
		sev := SeverityCrit
		if isApex && allowApexCNAME {
			sev = SeverityWarn
		}
		data.Coexisting = append(data.Coexisting, CoexistingRRset{Type: t, TTL: ttl})
		data.Findings = append(data.Findings, AliasFinding{
			Code:     "alias_coexisting_rrset",
			Severity: sev,
			Message:  fmt.Sprintf("%s and CNAME both exist at %s (RFC 1034 §3.6.2 / RFC 2181 §10.1)", t, owner),
			Subject:  owner,
			Hint:     "Remove the sibling record or move it under a different label; a name cannot simultaneously carry a CNAME and other data.",
		})
	}
}

// checkDNSSEC verifies that, if the zone is signed, the CNAME at owner is
// properly signed (RRSIG covers it).
func checkDNSSEC(ctx context.Context, data *AliasData, servers []string, apex, owner string) {
	qk := dns.Question{Name: apex, Qtype: dns.TypeDNSKEY, Qclass: dns.ClassINET}
	r, _, err := queryAtAuth(ctx, "tcp", servers, qk)
	if err != nil || r == nil || r.Rcode != dns.RcodeSuccess {
		return
	}
	signed := false
	for _, rr := range r.Answer {
		if _, ok := rr.(*dns.DNSKEY); ok {
			signed = true
			break
		}
	}
	data.ZoneSigned = signed
	if !signed {
		return
	}

	// Query CNAME with DO; check for an RRSIG covering it.
	q := dns.Question{Name: owner, Qtype: dns.TypeCNAME, Qclass: dns.ClassINET}
	r, _, err = queryAtAuth(ctx, "tcp", servers, q)
	if err != nil || r == nil {
		return
	}
	sawCNAME := false
	sawSig := false
	for _, rr := range r.Answer {
		switch v := rr.(type) {
		case *dns.CNAME:
			sawCNAME = true
		case *dns.RRSIG:
			if v.TypeCovered == dns.TypeCNAME {
				sawSig = true
			}
		}
	}
	if sawCNAME {
		data.CNAMESigned = sawSig
		if !sawSig {
			data.Findings = append(data.Findings, AliasFinding{
				Code:     "alias_cname_not_signed",
				Severity: SeverityCrit,
				Message:  fmt.Sprintf("zone %s is DNSSEC-signed but CNAME at %s has no RRSIG", apex, owner),
				Subject:  owner,
				Hint:     "Re-sign the zone or verify your signer covers the alias RRset; unsigned answers in a signed zone SERVFAIL at validating resolvers.",
			})
		}
	}
}

// validateChain enforces global chain invariants.
func validateChain(data *AliasData, requireTarget bool) {
	if len(data.Chain) == 0 {
		return
	}

	// Target resolvability.
	if last := data.Chain[len(data.Chain)-1]; last.Kind == KindTarget {
		if len(data.FinalA) == 0 && len(data.FinalAAAA) == 0 {
			sev := SeverityWarn
			if requireTarget {
				sev = SeverityCrit
			}
			rcode := data.Rcode
			if rcode == "" {
				rcode = "no A/AAAA"
			}
			data.Findings = append(data.Findings, AliasFinding{
				Code:     "alias_target_unresolvable",
				Severity: sev,
				Message:  fmt.Sprintf("final target %s does not resolve to an address (%s)", last.Owner, rcode),
				Subject:  last.Owner,
				Hint:     "Point the alias at a name that publishes at least one A or AAAA record, or fix the upstream zone.",
			})
		}
	}

	// Multiple CNAME/DNAME kinds with same owner (malformed zone).
	seen := map[string]int{}
	for _, h := range data.Chain {
		if h.Kind == KindCNAME || h.Kind == KindDNAME {
			seen[h.Owner]++
		}
	}
	for o, n := range seen {
		if n > 1 {
			data.Findings = append(data.Findings, AliasFinding{
				Code:     "alias_multiple_records",
				Severity: SeverityCrit,
				Message:  fmt.Sprintf("%s carries %d CNAME/DNAME records in the chain; only one is legal per owner", o, n),
				Subject:  o,
				Hint:     "Keep a single CNAME per name; remove duplicates at the authoritative zone.",
			})
		}
	}
}