cmd/tailscale: add start of "debug derp" subcommand

Updates #6526

Change-Id: I84e440a8bd837c383000ce0cec4ff36b24249e8b
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>

client/tailscale/localclient.go

@@ -981,6 +981,15 @@ func (lc *LocalClient) DeleteProfile(ctx context.Context, profile ipn.ProfileID)
 	return err
 }
 
+func (lc *LocalClient) DebugDERPRegion(ctx context.Context, regionIDOrCode string) (*ipnstate.DebugDERPRegionReport, error) {
+	v := url.Values{"region": {regionIDOrCode}}
+	body, err := lc.send(ctx, "POST", "/localapi/v0/debug-derp-region?"+v.Encode(), 200, nil)
+	if err != nil {
+		return nil, fmt.Errorf("error %w: %s", err, body)
+	}
+	return decodeJSON[*ipnstate.DebugDERPRegionReport](body)
+}
+
 // WatchIPNMask are filtering options for LocalClient.WatchIPNBus.
 //
 // The zero value is a valid WatchOpt that means to watch everything.

cmd/tailscale/cli/debug.go

@@ -37,6 +37,7 @@
 	"tailscale.com/tailcfg"
 	"tailscale.com/types/key"
 	"tailscale.com/types/logger"
+	"tailscale.com/util/must"
 	"tailscale.com/util/strs"
 )
 
@@ -160,6 +161,11 @@
 				return fs
 			})(),
 		},
+		{
+			Name:      "derp",
+			Exec:      runDebugDERP,
+			ShortHelp: "test a DERP configuration",
+		},
 	},
 }
 
@@ -610,3 +616,15 @@ func runDevStoreSet(ctx context.Context, args []string) error {
 	}
 	return localClient.SetDevStoreKeyValue(ctx, key, val)
 }
+
+func runDebugDERP(ctx context.Context, args []string) error {
+	if len(args) != 1 {
+		return errors.New("usage: debug derp <region>")
+	}
+	st, err := localClient.DebugDERPRegion(ctx, args[0])
+	if err != nil {
+		return err
+	}
+	fmt.Printf("%s\n", must.Get(json.MarshalIndent(st, "", " ")))
+	return nil
+}

cmd/tailscale/depaware.txt

@@ -108,6 +108,7 @@ tailscale.com/cmd/tailscale dependencies: (generated by github.com/tailscale/dep
         tailscale.com/util/lineread                                  from tailscale.com/net/interfaces+
         tailscale.com/util/mak                                       from tailscale.com/net/netcheck+
         tailscale.com/util/multierr                                  from tailscale.com/control/controlhttp
+        tailscale.com/util/must                                      from tailscale.com/cmd/tailscale/cli
         tailscale.com/util/quarantine                                from tailscale.com/cmd/tailscale/cli
         tailscale.com/util/singleflight                              from tailscale.com/net/dnscache
         tailscale.com/util/strs                                      from tailscale.com/hostinfo+

ipn/ipnstate/ipnstate.go

@@ -618,3 +618,11 @@ func sortKey(ps *PeerStatus) string {
 	raw := ps.PublicKey.Raw32()
 	return string(raw[:])
 }
+
+// DebugDERPRegionReport is the result of a "tailscale debug derp" command,
+// to let people debug a custom DERP setup.
+type DebugDERPRegionReport struct {
+	Info     []string
+	Warnings []string
+	Errors   []string
+}

ipn/localapi/debugderp.go

@@ -0,0 +1,92 @@
+// Copyright (c) 2022 Tailscale Inc & AUTHORS All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package localapi
+
+import (
+	"encoding/json"
+	"fmt"
+	"net/http"
+	"strconv"
+
+	"tailscale.com/ipn/ipnstate"
+	"tailscale.com/tailcfg"
+)
+
+func (h *Handler) serveDebugDERPRegion(w http.ResponseWriter, r *http.Request) {
+	if !h.PermitWrite {
+		http.Error(w, "debug access denied", http.StatusForbidden)
+		return
+	}
+	if r.Method != "POST" {
+		http.Error(w, "POST required", http.StatusMethodNotAllowed)
+		return
+	}
+	var st ipnstate.DebugDERPRegionReport
+	defer func() {
+		j, _ := json.Marshal(st)
+		w.Header().Set("Content-Type", "application/json")
+		w.Write(j)
+	}()
+
+	dm := h.b.DERPMap()
+	if dm == nil {
+		st.Errors = append(st.Errors, "no DERP map (not connected?)")
+		return
+	}
+	regStr := r.FormValue("region")
+	var reg *tailcfg.DERPRegion
+	if id, err := strconv.Atoi(regStr); err == nil {
+		reg = dm.Regions[id]
+	} else {
+		for _, r := range dm.Regions {
+			if r.RegionCode == regStr {
+				reg = r
+				break
+			}
+		}
+	}
+	if reg == nil {
+		st.Errors = append(st.Errors, fmt.Sprintf("no such region %q in DERP map", regStr))
+		return
+	}
+	st.Info = append(st.Info, fmt.Sprintf("Region %v == %q", reg.RegionID, reg.RegionCode))
+
+	if reg.Avoid {
+		st.Warnings = append(st.Warnings, "Region is marked with Avoid bit")
+	}
+	if len(reg.Nodes) == 0 {
+		st.Errors = append(st.Errors, "Region has no nodes defined")
+		return
+	}
+
+	// TODO(bradfitz): finish:
+	// * first try TCP connection
+	// * reconnect 4 or 5 times; see if we ever get a different server key.
+	//   if so, they're load balancing the wrong way. error.
+	// * try to DERP auth with new public key.
+	// * if rejected, add Info that it's likely the DERP server authz is on,
+	//   try with LocalBackend's node key instead.
+	// * if they have more then one node, try to relay a packet between them
+	//   and see if it works (like cmd/derpprobe). But if server authz is on,
+	//   we won't be able to, so just warn. Say to turn that off, try again,
+	//   then turn it back on. TODO(bradfitz): maybe add a debug frame to DERP
+	//   protocol to say how many peers it's meshed with. Should match count
+	//   in DERPRegion. Or maybe even list all their server pub keys that it's peered
+	//   with.
+	// * try STUN queries
+	// * warn about IPv6 only
+	// * If their certificate is bad, either expired or just wrongly
+	//   issued in the first place, tell them specifically that the
+	// 	 cert is bad not just that the connection failed.
+	// * If /generate_204 on port 80 cannot be reached, warn
+	// 	 that they won't get captive portal detection and
+	// 	 should allow port 80.
+	// * If they have exactly one DERP region because they
+	//   removed all of Tailscale's DERPs, warn that they have
+	//   a SPOF that will hamper even direct connections from
+	//   working. (warning, not error, as that's probably a likely
+	//   config for headscale users)
+	st.Info = append(st.Info, "TODO: 🦉")
+}

ipn/localapi/localapi.go

@@ -65,6 +65,7 @@
 	"check-prefs":             (*Handler).serveCheckPrefs,
 	"component-debug-logging": (*Handler).serveComponentDebugLogging,
 	"debug":                   (*Handler).serveDebug,
+	"debug-derp-region":       (*Handler).serveDebugDERPRegion,
 	"derpmap":                 (*Handler).serveDERPMap,
 	"dev-set-state-store":     (*Handler).serveDevSetStateStore,
 	"dial":                    (*Handler).serveDial,