mapper, app: ship MagicDNS Routes as empty slices, not nil

policyChangeResponse already includes everything else; carry DNSConfig
too so the client's netmap DNS is anchored on every policy change
rather than relying on the previous snapshot.

Send the MagicDNS root domains as empty non-nil Resolver slices instead
of nil values. tailcfg.DNSConfig.Clone and net/dns.Config.Clone in
tailscale drop map entries whose value is nil (tailcfg_clone.go and
dns_clone.go both contain `if sv == nil { continue }`). On a major
LinkChange the client's wgengine handler clones lastDNSConfig and
re-applies it; with nil values the cloned config has Routes:{}, dns.Set
wipes Nameservers in /etc/resolv.conf, and curl-by-FQDN fails until
the next route-changing netmap, typically about six minutes later.
Empty slice survives Clone and carries the same "resolve locally"
semantics for Routes entries.

hscontrol/app.go

@@ -219,7 +219,19 @@ func NewHeadscale(cfg *types.Config) (*Headscale, error) {
 		}
 
 		for _, d := range magicDNSDomains {
-			app.cfg.TailcfgDNSConfig.Routes[d.WithoutTrailingDot()] = nil
+			// Empty non-nil slice rather than nil: tailcfg.DNSConfig.Clone
+			// and dns.Config.Clone in tailscale drop map entries whose
+			// value is nil (see tailscale.com/tailcfg/tailcfg_clone.go and
+			// tailscale.com/net/dns/dns_clone.go: `if sv == nil { continue }`).
+			// Sending nil here caused the client's wgengine LinkChange:major
+			// handler to clobber /etc/resolv.conf on every tunnel-IP rebind
+			// — the handler reapplies a Clone of lastDNSConfig and the magic
+			// DNS routes vanish, taking the resolver with them for ~6 min
+			// until the next route-changing netmap. Empty slice survives
+			// Clone and carries the same "resolve locally" semantics
+			// (tailscale.com/ipn/ipnlocal/node_backend.go:869 documents the
+			// empty-resolver Routes form for Issue 2706).
+			app.cfg.TailcfgDNSConfig.Routes[d.WithoutTrailingDot()] = []*dnstype.Resolver{}
 		}
 	}
 

hscontrol/mapper/mapper.go

@@ -314,11 +314,18 @@ func (m *mapper) selfMapResponse(
 
 // policyChangeResponse creates a MapResponse for policy changes.
 // It sends:
-// - PeersRemoved for peers that are no longer visible after the policy change
+//   - PeersRemoved for peers that are no longer visible after the policy change
-// - PeersChanged for remaining peers (their AllowedIPs may have changed due to policy)
+//   - PeersChanged for remaining peers (their AllowedIPs may have changed due to policy)
-// - Updated PacketFilters
+//   - Updated PacketFilters
-// - Updated SSHPolicy (SSH rules may reference users/groups that changed)
+//   - Updated SSHPolicy (SSH rules may reference users/groups that changed)
-// - Optionally, the node's own self info (when includeSelf is true)
+//   - DNSConfig so the client's resolver state stays anchored even when a
+//     policy-triggered wgengine reconfigure races a netmon LinkChange (the
+//     LinkChange handler reapplies dns.Manager.Set with the engine's
+//     lastDNSConfig; if that snapshot is stale, the OS resolver loses the
+//     MagicDNS reverse-DNS routes and Nameservers and curl-by-FQDN stops
+//     resolving for the rest of the policy window).
+//   - Optionally, the node's own self info (when includeSelf is true)
+//
 // This avoids the issue where an empty Peers slice is interpreted by Tailscale
 // clients as "no change" rather than "no peers".
 // When includeSelf is true, the node's self info is included so that a node
@@ -334,6 +341,7 @@ func (m *mapper) policyChangeResponse(
 	builder := m.NewMapResponseBuilder(nodeID).
 		WithDebugType(policyResponseDebug).
 		WithCapabilityVersion(capVer).
+		WithDNSConfig().
 		WithPacketFilters().
 		WithSSHPolicy()