From eb23a645a53663afd3976ea6065129abc26a3cb0 Mon Sep 17 00:00:00 2001
From: Andrew Dunham <andrew@du.nham.ca>
Date: Wed, 26 Jul 2023 12:29:18 -0400
Subject: [PATCH] health, ipn/ipnlocal: add healthcheck for state/config

If the user sufficiently breaks their state file such that the local
machine or node keys are zero, the machine ends up in a broken state but
we don't surface this to the user. Add a healthcheck that tracks errors
when these key(s) are zero and surfaces them in 'tailscale status'/etc.

Signed-off-by: Andrew Dunham <andrew@du.nham.ca>
Change-Id: Ideffbf9510b09850ebf8daf440fac78109de7812
---
 health/health.go      | 10 ++++++++++
 ipn/ipnlocal/local.go | 22 ++++++++++++++++++++++
 2 files changed, 32 insertions(+)

diff --git a/health/health.go b/health/health.go
index bef416272..39d3edd87 100644
--- a/health/health.go
+++ b/health/health.go
@@ -59,6 +59,10 @@ const (
 	// the system, rather than one particular subsystem.
 	SysOverall = Subsystem("overall")
 
+	// SysConfig is the name of the subsystem that tracks health of the
+	// Tailscale state file and configuration.
+	SysConfig = Subsystem("config")
+
 	// SysRouter is the name of the wgengine/router subsystem.
 	SysRouter = Subsystem("router")
 
@@ -203,6 +207,12 @@ func SetTKAHealth(err error) { setErr(SysTKA, err) }
 // TKAHealth returns the tailnet key authority error state.
 func TKAHealth() error { return get(SysTKA) }
 
+// SetConfigHealth sets the health of the configuration.
+func SetConfigHealth(err error) { setErr(SysConfig, err) }
+
+// ConfigHealth returns the configuration error state.
+func ConfigHealth() error { return get(SysConfig) }
+
 // SetLocalLogConfigHealth sets the error state of this client's local log configuration.
 func SetLocalLogConfigHealth(err error) {
 	mu.Lock()
diff --git a/ipn/ipnlocal/local.go b/ipn/ipnlocal/local.go
index 813d01265..7ba8299b6 100644
--- a/ipn/ipnlocal/local.go
+++ b/ipn/ipnlocal/local.go
@@ -3667,6 +3667,27 @@ func (b *LocalBackend) enterStateLockedOnEntry(newState ipn.State) {
 			addrs = append(addrs, addr.Addr().String())
 		}
 		systemd.Status("Connected; %s; %s", activeLogin, strings.Join(addrs, " "))
+
+		var configErrs []error
+		if prefs.Valid() && prefs.Persist().Valid() {
+			pprefs := prefs.Persist()
+			if pprefs.PublicNodeKey().IsZero() {
+				configErrs = append(configErrs, fmt.Errorf("public node key is zero"))
+			}
+			if pprefs.PrivateNodeKey().IsZero() {
+				configErrs = append(configErrs, fmt.Errorf("private node key is zero"))
+			}
+		}
+		b.mu.Lock()
+		if b.machinePrivKey.IsZero() {
+			configErrs = append(configErrs, fmt.Errorf("private machine key is zero"))
+		}
+		b.mu.Unlock()
+
+		if len(configErrs) > 0 {
+			health.SetConfigHealth(multierr.New(configErrs...))
+		}
+
 	case ipn.NoState:
 		// Do nothing.
 	default:
@@ -4920,6 +4941,7 @@ func (b *LocalBackend) resetForProfileChangeLockedOnEntry() error {
 	b.serveConfig = ipn.ServeConfigView{}
 	b.enterStateLockedOnEntry(ipn.NoState) // Reset state.
 	health.SetLocalLogConfigHealth(nil)
+	health.SetConfigHealth(nil)
 	return b.Start(ipn.Options{})
 }