Internet exposure checks and response rate limiting

Jarl works out which views are reachable from the internet (match-clients
ending in a catch-all, not private-looking addresses: an internal view of a
global IPv6 prefix is still yours; a server without views: open allow-query
and a non-private listen address), with a per-view override (schema 9), and
checks public ones: rate limiting, open resolver, minimal-any, cache
snooping, open NOTIFY and transfers, version and host name, management
interfaces. Fixes are staged like any change, including rate limiting from
ISC's example in log-only mode with the secondaries exempt, then enforcing.
named logs what rate limiting limits (or would, in log-only mode, where its
counters don't move) as query-errors; Jarl routes those lines to its log
and counts them (rrl breakdown, ratelimited top lists, Prometheus
jarl_ratelimit_logged_total), shown on the new Exposure page.

A change that newly opens a public view up must be confirmed (accept_risk,
--accept-risk, a checkbox on the Changes page; 409 confirm_risk
otherwise). Findings can be acknowledged with a reason and stay visible as
info. GET /exposure, POST /exposure/fix|ack|unack, jarl exposure,
docs/exposure.md.

Also fixes query statistics skipping TSIG-signed queries ("IP#port/key
NAME" in named's log).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

internal/engine/exposure.go

@@ -0,0 +1,163 @@
+package engine
+
+import (
+	"net/netip"
+	"slices"
+	"strings"
+
+	"jarl/internal/model"
+	"jarl/internal/namedconf"
+)
+
+// ExposureFix returns the requests that fix an exposure finding of a view
+// (viewID 0: the server), and what they do, or nil when there is no
+// automatic fix: what to allow instead (your secondaries, your hosts)
+// is for you to decide. The requests go through a changeset like any
+// edit (docs/exposure.md).
+func ExposureFix(st *model.State, viewID int64, code string) ([]Request, string) {
+	v := st.Views[viewID]
+	scope, scopeID := model.ScopeOptions, int64(0)
+	if v != nil {
+		scope, scopeID = model.ScopeView, v.ID
+	}
+	set := func(name, value string) Request {
+		if o := st.FindOption(scope, scopeID, name); o != nil {
+			return NewRequest(OpOptionUpdate, o.ID, OptionInput{Value: &value})
+		}
+		return NewRequest(OpOptionCreate, 0, OptionInput{Scope: &scope, ScopeID: &scopeID, Name: &name, Value: &value})
+	}
+	server := func(name, value string) Request {
+		if o := st.FindOption(model.ScopeOptions, 0, name); o != nil {
+			return NewRequest(OpOptionUpdate, o.ID, OptionInput{Value: &value})
+		}
+		return NewRequest(OpOptionCreate, 0, OptionInput{Scope: model.Ptr(model.ScopeOptions), Name: &name, Value: &value})
+	}
+	switch code {
+	case "exposure_rate_limit":
+		value := RecommendedRateLimit(st, v)
+		return []Request{set("rate-limit", value)}, "start rate limiting in log-only mode: rate-limit " + value
+	case "exposure_rate_limit_log_only":
+		o := st.FindOption(scope, scopeID, "rate-limit")
+		if o == nil {
+			o = st.FindOption(model.ScopeOptions, 0, "rate-limit")
+		}
+		if o == nil {
+			return nil, ""
+		}
+		value := withoutLogOnly(o.Value)
+		return []Request{NewRequest(OpOptionUpdate, o.ID, OptionInput{Value: &value})}, "start enforcing the rate limit: rate-limit " + value
+	case "exposure_minimal_any":
+		return []Request{set("minimal-any", "yes")}, "answer ANY queries minimally: minimal-any yes"
+	case "exposure_cache_snooping":
+		return []Request{set("allow-query-cache", "{ none; }")}, "close the cache to everyone: allow-query-cache { none; }"
+	case "exposure_version":
+		return []Request{server("version", "none")}, "hide BIND's version: version none"
+	case "exposure_hostname":
+		return []Request{server("hostname", "none")}, "hide the host name: hostname none"
+	case "security_open_resolver":
+		if v == nil {
+			return nil, "" // a server without views is also your own resolver
+		}
+		return []Request{set("recursion", "no")}, "turn recursion off in view " + v.Name + ": recursion no"
+	}
+	return nil, ""
+}
+
+// RecommendedRateLimit is the rate-limit block Jarl suggests for a public
+// view (v nil: the server): ISC's example of 10 identical responses per
+// second per client network, in log-only mode until you have looked at
+// what it would limit, with your secondaries exempt so zone transfers
+// and their queries are never limited.
+func RecommendedRateLimit(st *model.State, v *model.View) string {
+	parts := []string{"responses-per-second 10;", "log-only yes;"}
+	if ex := secondaries(st, v); len(ex) > 0 {
+		parts = append(parts, "exempt-clients { "+strings.Join(ex, "; ")+"; };")
+	}
+	return "{ " + strings.Join(parts, " ") + " }"
+}
+
+// secondaries returns the addresses the view's zones notify and allow
+// transfers to (ACLs resolved), the servers that rate limiting must leave
+// alone.
+func secondaries(st *model.State, v *model.View) []string {
+	var opts []*model.Option
+	add := func(scope string, id int64) {
+		for _, name := range []string{"also-notify", "allow-transfer"} {
+			if o := st.FindOption(scope, id, name); o != nil {
+				opts = append(opts, o)
+			}
+		}
+	}
+	add(model.ScopeOptions, 0)
+	vid := int64(0)
+	if v != nil {
+		vid = v.ID
+		add(model.ScopeView, v.ID)
+	}
+	for _, z := range st.SortedZones() {
+		if z.ViewID == vid {
+			add(model.ScopeZone, z.ID)
+		}
+	}
+	seen := map[string]bool{}
+	var out []string
+	var addElems func(list []string, depth int)
+	addElems = func(list []string, depth int) {
+		for _, s := range list {
+			e, err := namedconf.ParseAMElement(s)
+			if err != nil || e.Negated {
+				continue
+			}
+			switch e.Kind {
+			case namedconf.AMPrefix, namedconf.AMRange:
+				if t := e.String(); !seen[t] {
+					seen[t] = true
+					out = append(out, t)
+				}
+			case namedconf.AMName:
+				if a := st.ACLByName(e.Name); a != nil && depth < 10 {
+					addElems(a.Elements, depth+1)
+				}
+			}
+		}
+	}
+	for _, o := range opts {
+		stmt, err := namedconf.ParseOptionValue(o.Name, o.Value)
+		if err != nil || stmt.FirstBlock() == nil {
+			continue
+		}
+		var list []string
+		for _, s := range stmt.FirstBlock().Stmts {
+			if o.Name == "also-notify" {
+				// "ADDRESS [port N] [key K]": the address only.
+				if len(s.Items) > 0 && s.Items[0].Value != nil {
+					if _, err := netip.ParseAddr(s.Items[0].Value.Text); err == nil {
+						list = append(list, s.Items[0].Value.Text)
+					}
+				}
+				continue
+			}
+			list = append(list, namedconf.Compact(s.Items))
+		}
+		addElems(list, 0)
+	}
+	slices.Sort(out)
+	return out
+}
+
+// withoutLogOnly removes "log-only ...;" from a rate-limit block.
+func withoutLogOnly(value string) string {
+	st, err := namedconf.ParseOptionValue("rate-limit", value)
+	if err != nil || st.FirstBlock() == nil {
+		return value
+	}
+	b := st.FirstBlock()
+	var keep []*namedconf.Stmt
+	for _, s := range b.Stmts {
+		if s.Keyword() != "log-only" {
+			keep = append(keep, s)
+		}
+	}
+	b.Stmts = keep
+	return namedconf.Compact(st.Items[1:])
+}