diff --git a/build_dist.sh b/build_dist.sh
index f11d4aae2..fed37c264 100755
--- a/build_dist.sh
+++ b/build_dist.sh
@@ -41,7 +41,7 @@ while [ "$#" -gt 1 ]; do
 		fi
 		shift
 		ldflags="$ldflags -w -s"
-		tags="${tags:+$tags,}ts_omit_aws,ts_omit_bird,ts_omit_tap,ts_omit_kube,ts_omit_completion,ts_omit_ssh,ts_omit_wakeonlan,ts_omit_capture,ts_omit_relayserver,ts_omit_taildrop"
+		tags="${tags:+$tags,}ts_omit_aws,ts_omit_bird,ts_omit_tap,ts_omit_kube,ts_omit_completion,ts_omit_ssh,ts_omit_wakeonlan,ts_omit_capture,ts_omit_relayserver,ts_omit_taildrop,ts_omit_tpm"
 		;;
 	--box)
 		if [ ! -z "${TAGS:-}" ]; then
diff --git a/cmd/k8s-operator/depaware.txt b/cmd/k8s-operator/depaware.txt
index 4cc4a8d46..544fe9089 100644
--- a/cmd/k8s-operator/depaware.txt
+++ b/cmd/k8s-operator/depaware.txt
@@ -135,6 +135,13 @@ tailscale.com/cmd/k8s-operator dependencies: (generated by github.com/tailscale/
         github.com/google/go-cmp/cmp/internal/flags                  from github.com/google/go-cmp/cmp+
         github.com/google/go-cmp/cmp/internal/function               from github.com/google/go-cmp/cmp
      💣 github.com/google/go-cmp/cmp/internal/value                  from github.com/google/go-cmp/cmp
+        github.com/google/go-tpm/legacy/tpm2                         from github.com/google/go-tpm/tpm2/transport+
+        github.com/google/go-tpm/tpm2                                from tailscale.com/feature/tpm
+        github.com/google/go-tpm/tpm2/transport                      from github.com/google/go-tpm/tpm2/transport/linuxtpm+
+   L    github.com/google/go-tpm/tpm2/transport/linuxtpm             from tailscale.com/feature/tpm
+   W    github.com/google/go-tpm/tpm2/transport/windowstpm           from tailscale.com/feature/tpm
+        github.com/google/go-tpm/tpmutil                             from github.com/google/go-tpm/legacy/tpm2+
+   W 💣 github.com/google/go-tpm/tpmutil/tbs                         from github.com/google/go-tpm/legacy/tpm2+
         github.com/google/gofuzz                                     from k8s.io/apimachinery/pkg/apis/meta/v1+
         github.com/google/gofuzz/bytesource                          from github.com/google/gofuzz
    L    github.com/google/nftables                                   from tailscale.com/util/linuxfw
@@ -813,6 +820,7 @@ tailscale.com/cmd/k8s-operator dependencies: (generated by github.com/tailscale/
         tailscale.com/feature/relayserver                            from tailscale.com/feature/condregister
         tailscale.com/feature/taildrop                               from tailscale.com/feature/condregister
    L    tailscale.com/feature/tap                                    from tailscale.com/feature/condregister
+        tailscale.com/feature/tpm                                    from tailscale.com/feature/condregister
         tailscale.com/feature/wakeonlan                              from tailscale.com/feature/condregister
         tailscale.com/health                                         from tailscale.com/control/controlclient+
         tailscale.com/health/healthmsg                               from tailscale.com/ipn/ipnlocal
@@ -832,9 +840,10 @@ tailscale.com/cmd/k8s-operator dependencies: (generated by github.com/tailscale/
         tailscale.com/ipn/store/kubestore                            from tailscale.com/cmd/k8s-operator+
         tailscale.com/ipn/store/mem                                  from tailscale.com/ipn/ipnlocal+
         tailscale.com/k8s-operator                                   from tailscale.com/cmd/k8s-operator
+        tailscale.com/k8s-operator/api-proxy                         from tailscale.com/cmd/k8s-operator
         tailscale.com/k8s-operator/apis                              from tailscale.com/k8s-operator/apis/v1alpha1
         tailscale.com/k8s-operator/apis/v1alpha1                     from tailscale.com/cmd/k8s-operator+
-        tailscale.com/k8s-operator/sessionrecording                  from tailscale.com/cmd/k8s-operator
+        tailscale.com/k8s-operator/sessionrecording                  from tailscale.com/k8s-operator/api-proxy
         tailscale.com/k8s-operator/sessionrecording/spdy             from tailscale.com/k8s-operator/sessionrecording
         tailscale.com/k8s-operator/sessionrecording/tsrecorder       from tailscale.com/k8s-operator/sessionrecording+
         tailscale.com/k8s-operator/sessionrecording/ws               from tailscale.com/k8s-operator/sessionrecording
@@ -937,7 +946,7 @@ tailscale.com/cmd/k8s-operator dependencies: (generated by github.com/tailscale/
         tailscale.com/util/clientmetric                              from tailscale.com/cmd/k8s-operator+
         tailscale.com/util/cloudenv                                  from tailscale.com/hostinfo+
         tailscale.com/util/cmpver                                    from tailscale.com/clientupdate+
-        tailscale.com/util/ctxkey                                    from tailscale.com/cmd/k8s-operator+
+        tailscale.com/util/ctxkey                                    from tailscale.com/client/tailscale/apitype+
      💣 tailscale.com/util/deephash                                  from tailscale.com/ipn/ipnlocal+
    L 💣 tailscale.com/util/dirwalk                                   from tailscale.com/metrics+
         tailscale.com/util/dnsname                                   from tailscale.com/appc+
diff --git a/cmd/tailscale/cli/set.go b/cmd/tailscale/cli/set.go
index 37db252ad..f4ea674ec 100644
--- a/cmd/tailscale/cli/set.go
+++ b/cmd/tailscale/cli/set.go
@@ -83,7 +83,7 @@ func newSetFlagSet(goos string, setArgs *setArgsT) *flag.FlagSet {
 	setf.BoolVar(&setArgs.advertiseConnector, "advertise-connector", false, "offer to be an app connector for domain specific internet traffic for the tailnet")
 	setf.BoolVar(&setArgs.updateCheck, "update-check", true, "notify about available Tailscale updates")
 	setf.BoolVar(&setArgs.updateApply, "auto-update", false, "automatically update to the latest available version")
-	setf.BoolVar(&setArgs.postureChecking, "posture-checking", false, hidden+"allow management plane to gather device posture information")
+	setf.BoolVar(&setArgs.postureChecking, "posture-checking", false, "allow management plane to gather device posture information")
 	setf.BoolVar(&setArgs.runWebClient, "webclient", false, "expose the web interface for managing this node over Tailscale at port 5252")
 	setf.StringVar(&setArgs.relayServerPort, "relay-server-port", "", hidden+"UDP port number (0 will pick a random unused port) for the relay server to bind to, on all interfaces, or empty string to disable relay server functionality")
 
diff --git a/cmd/tailscaled/depaware.txt b/cmd/tailscaled/depaware.txt
index 329c00e93..c5d5a7b2d 100644
--- a/cmd/tailscaled/depaware.txt
+++ b/cmd/tailscaled/depaware.txt
@@ -109,6 +109,13 @@ tailscale.com/cmd/tailscaled dependencies: (generated by github.com/tailscale/de
    L 💣 github.com/godbus/dbus/v5                                    from tailscale.com/net/dns+
         github.com/golang/groupcache/lru                             from tailscale.com/net/dnscache
         github.com/google/btree                                      from gvisor.dev/gvisor/pkg/tcpip/header+
+        github.com/google/go-tpm/legacy/tpm2                         from github.com/google/go-tpm/tpm2/transport+
+        github.com/google/go-tpm/tpm2                                from tailscale.com/feature/tpm
+        github.com/google/go-tpm/tpm2/transport                      from github.com/google/go-tpm/tpm2/transport/linuxtpm+
+   L    github.com/google/go-tpm/tpm2/transport/linuxtpm             from tailscale.com/feature/tpm
+   W    github.com/google/go-tpm/tpm2/transport/windowstpm           from tailscale.com/feature/tpm
+        github.com/google/go-tpm/tpmutil                             from github.com/google/go-tpm/legacy/tpm2+
+   W 💣 github.com/google/go-tpm/tpmutil/tbs                         from github.com/google/go-tpm/legacy/tpm2+
    L    github.com/google/nftables                                   from tailscale.com/util/linuxfw
    L 💣 github.com/google/nftables/alignedbuff                       from github.com/google/nftables/xt
    L 💣 github.com/google/nftables/binaryutil                        from github.com/google/nftables+
@@ -271,6 +278,7 @@ tailscale.com/cmd/tailscaled dependencies: (generated by github.com/tailscale/de
         tailscale.com/feature/relayserver                            from tailscale.com/feature/condregister
         tailscale.com/feature/taildrop                               from tailscale.com/feature/condregister
    L    tailscale.com/feature/tap                                    from tailscale.com/feature/condregister
+        tailscale.com/feature/tpm                                    from tailscale.com/feature/condregister
         tailscale.com/feature/wakeonlan                              from tailscale.com/feature/condregister
         tailscale.com/health                                         from tailscale.com/control/controlclient+
         tailscale.com/health/healthmsg                               from tailscale.com/ipn/ipnlocal
diff --git a/cmd/tailscaled/tailscaled.go b/cmd/tailscaled/tailscaled.go
index 1c5236123..4b0dc95f9 100644
--- a/cmd/tailscaled/tailscaled.go
+++ b/cmd/tailscaled/tailscaled.go
@@ -573,7 +573,7 @@ func getLocalBackend(ctx context.Context, logf logger.Logf, logID logid.PublicID
 		if ms, ok := sys.MagicSock.GetOK(); ok {
 			debugMux.HandleFunc("/debug/magicsock", ms.ServeHTTPDebug)
 		}
-		go runDebugServer(debugMux, args.debug)
+		go runDebugServer(logf, debugMux, args.debug)
 	}
 
 	ns, err := newNetstack(logf, sys)
@@ -819,12 +819,20 @@ func servePrometheusMetrics(w http.ResponseWriter, r *http.Request) {
 	clientmetric.WritePrometheusExpositionFormat(w)
 }
 
-func runDebugServer(mux *http.ServeMux, addr string) {
+func runDebugServer(logf logger.Logf, mux *http.ServeMux, addr string) {
+	ln, err := net.Listen("tcp", addr)
+	if err != nil {
+		log.Fatalf("debug server: %v", err)
+	}
+	if strings.HasSuffix(addr, ":0") {
+		// Log kernel-selected port number so integration tests
+		// can find it portably.
+		logf("DEBUG-ADDR=%v", ln.Addr())
+	}
 	srv := &http.Server{
-		Addr:    addr,
 		Handler: mux,
 	}
-	if err := srv.ListenAndServe(); err != nil {
+	if err := srv.Serve(ln); err != nil {
 		log.Fatal(err)
 	}
 }
diff --git a/cmd/tsidp/README.md b/cmd/tsidp/README.md
index 29ce089df..61a81e8ae 100644
--- a/cmd/tsidp/README.md
+++ b/cmd/tsidp/README.md
@@ -35,7 +35,7 @@
 
    ```bash
    docker run -d \
-     --name `tsidp` \
+     --name tsidp \
      -p 443:443 \
      -e TS_AUTHKEY=YOUR_TAILSCALE_AUTHKEY \
      -e TS_HOSTNAME=idp \
diff --git a/feature/condregister/maybe_tpm.go b/feature/condregister/maybe_tpm.go
new file mode 100644
index 000000000..caa57fef1
--- /dev/null
+++ b/feature/condregister/maybe_tpm.go
@@ -0,0 +1,8 @@
+// Copyright (c) Tailscale Inc & AUTHORS
+// SPDX-License-Identifier: BSD-3-Clause
+
+//go:build !ios && !ts_omit_tpm
+
+package condregister
+
+import _ "tailscale.com/feature/tpm"
diff --git a/feature/taildrop/integration_test.go b/feature/taildrop/integration_test.go
new file mode 100644
index 000000000..46768bb31
--- /dev/null
+++ b/feature/taildrop/integration_test.go
@@ -0,0 +1,170 @@
+// Copyright (c) Tailscale Inc & AUTHORS
+// SPDX-License-Identifier: BSD-3-Clause
+
+package taildrop_test
+
+import (
+	"bytes"
+	"context"
+	"errors"
+	"fmt"
+	"io"
+	"testing"
+	"time"
+
+	"tailscale.com/client/local"
+	"tailscale.com/client/tailscale/apitype"
+	"tailscale.com/tailcfg"
+	"tailscale.com/tstest"
+	"tailscale.com/tstest/integration"
+	"tailscale.com/tstest/integration/testcontrol"
+)
+
+// TODO(bradfitz): add test where control doesn't send tailcfg.CapabilityFileSharing
+// and verify that we get the "file sharing not enabled by Tailscale admin" error.
+
+// TODO(bradfitz): add test between different users with the peercap to permit that?
+
+func TestTaildropIntegration(t *testing.T) {
+	tstest.Parallel(t)
+	controlOpt := integration.ConfigureControl(func(s *testcontrol.Server) {
+		s.AllNodesSameUser = true // required for Taildrop
+	})
+	env := integration.NewTestEnv(t, controlOpt)
+
+	// Create two nodes:
+	n1 := integration.NewTestNode(t, env)
+	d1 := n1.StartDaemon()
+
+	n2 := integration.NewTestNode(t, env)
+	d2 := n2.StartDaemon()
+
+	n1.AwaitListening()
+	t.Logf("n1 is listening")
+	n2.AwaitListening()
+	t.Logf("n2 is listening")
+	n1.MustUp()
+	t.Logf("n1 is up")
+	n2.MustUp()
+	t.Logf("n2 is up")
+	n1.AwaitRunning()
+	t.Logf("n1 is running")
+	n2.AwaitRunning()
+	t.Logf("n2 is running")
+
+	var peerStableID tailcfg.StableNodeID
+
+	if err := tstest.WaitFor(5*time.Second, func() error {
+		st := n1.MustStatus()
+		if len(st.Peer) == 0 {
+			return errors.New("no peers")
+		}
+		if len(st.Peer) > 1 {
+			return fmt.Errorf("got %d peers; want 1", len(st.Peer))
+		}
+		peer := st.Peer[st.Peers()[0]]
+		peerStableID = peer.ID
+		if peer.ID == st.Self.ID {
+			return errors.New("peer is self")
+		}
+
+		if len(st.TailscaleIPs) == 0 {
+			return errors.New("no Tailscale IPs")
+		}
+
+		return nil
+	}); err != nil {
+		t.Fatal(err)
+	}
+
+	const timeout = 30 * time.Second
+	ctx, cancel := context.WithTimeout(context.Background(), timeout)
+	defer cancel()
+
+	c1 := n1.LocalClient()
+	c2 := n2.LocalClient()
+
+	wantNoWaitingFiles := func(c *local.Client) {
+		t.Helper()
+		files, err := c.WaitingFiles(ctx)
+		if err != nil {
+			t.Fatalf("WaitingFiles: %v", err)
+		}
+		if len(files) != 0 {
+			t.Fatalf("WaitingFiles: got %d files; want 0", len(files))
+		}
+	}
+
+	// Verify c2 has no files.
+	wantNoWaitingFiles(c2)
+
+	gotFile := make(chan bool, 1)
+	go func() {
+		v, err := c2.AwaitWaitingFiles(t.Context(), timeout)
+		if err != nil {
+			return
+		}
+		if len(v) != 0 {
+			gotFile <- true
+		}
+	}()
+
+	fileContents := []byte("hello world this is a file")
+
+	n2ID := n2.MustStatus().Self.ID
+	t.Logf("n2 self.ID = %q; n1's peer[0].ID = %q", n2ID, peerStableID)
+	t.Logf("Doing PushFile ...")
+	err := c1.PushFile(ctx, n2.MustStatus().Self.ID, int64(len(fileContents)), "test.txt", bytes.NewReader(fileContents))
+	if err != nil {
+		t.Fatalf("PushFile from n1->n2: %v", err)
+	}
+	t.Logf("PushFile done")
+
+	select {
+	case <-gotFile:
+		t.Logf("n2 saw AwaitWaitingFiles wake up")
+	case <-ctx.Done():
+		t.Fatalf("n2 timeout waiting for AwaitWaitingFiles")
+	}
+
+	files, err := c2.WaitingFiles(ctx)
+	if err != nil {
+		t.Fatalf("c2.WaitingFiles: %v", err)
+	}
+	if len(files) != 1 {
+		t.Fatalf("c2.WaitingFiles: got %d files; want 1", len(files))
+	}
+	got := files[0]
+	want := apitype.WaitingFile{
+		Name: "test.txt",
+		Size: int64(len(fileContents)),
+	}
+	if got != want {
+		t.Fatalf("c2.WaitingFiles: got %+v; want %+v", got, want)
+	}
+
+	// Download the file.
+	rc, size, err := c2.GetWaitingFile(ctx, got.Name)
+	if err != nil {
+		t.Fatalf("c2.GetWaitingFile: %v", err)
+	}
+	if size != int64(len(fileContents)) {
+		t.Fatalf("c2.GetWaitingFile: got size %d; want %d", size, len(fileContents))
+	}
+	gotBytes, err := io.ReadAll(rc)
+	if err != nil {
+		t.Fatalf("c2.GetWaitingFile: %v", err)
+	}
+	if !bytes.Equal(gotBytes, fileContents) {
+		t.Fatalf("c2.GetWaitingFile: got %q; want %q", gotBytes, fileContents)
+	}
+
+	// Now delete it.
+	if err := c2.DeleteWaitingFile(ctx, got.Name); err != nil {
+		t.Fatalf("c2.DeleteWaitingFile: %v", err)
+	}
+	wantNoWaitingFiles(c2)
+
+	d1.MustCleanShutdown(t)
+	d2.MustCleanShutdown(t)
+}
diff --git a/feature/taildrop/localapi.go b/feature/taildrop/localapi.go
index ce812514e..067a51f91 100644
--- a/feature/taildrop/localapi.go
+++ b/feature/taildrop/localapi.go
@@ -365,6 +365,7 @@ func serveFiles(h *localapi.Handler, w http.ResponseWriter, r *http.Request) {
 			return
 		}
 		ctx := r.Context()
+		var wfs []apitype.WaitingFile
 		if s := r.FormValue("waitsec"); s != "" && s != "0" {
 			d, err := strconv.Atoi(s)
 			if err != nil {
@@ -375,11 +376,18 @@ func serveFiles(h *localapi.Handler, w http.ResponseWriter, r *http.Request) {
 			var cancel context.CancelFunc
 			ctx, cancel = context.WithDeadline(ctx, deadline)
 			defer cancel()
-		}
-		wfs, err := lb.AwaitWaitingFiles(ctx)
-		if err != nil && ctx.Err() == nil {
-			http.Error(w, err.Error(), http.StatusInternalServerError)
-			return
+			wfs, err = lb.AwaitWaitingFiles(ctx)
+			if err != nil && ctx.Err() == nil {
+				http.Error(w, err.Error(), http.StatusInternalServerError)
+				return
+			}
+		} else {
+			var err error
+			wfs, err = lb.WaitingFiles()
+			if err != nil {
+				http.Error(w, err.Error(), http.StatusInternalServerError)
+				return
+			}
 		}
 		w.Header().Set("Content-Type", "application/json")
 		json.NewEncoder(w).Encode(wfs)
diff --git a/feature/tpm/tpm.go b/feature/tpm/tpm.go
new file mode 100644
index 000000000..18e56ae89
--- /dev/null
+++ b/feature/tpm/tpm.go
@@ -0,0 +1,83 @@
+// Copyright (c) Tailscale Inc & AUTHORS
+// SPDX-License-Identifier: BSD-3-Clause
+
+// Package tpm implements support for TPM 2.0 devices.
+package tpm
+
+import (
+	"slices"
+	"sync"
+
+	"github.com/google/go-tpm/tpm2"
+	"github.com/google/go-tpm/tpm2/transport"
+	"tailscale.com/feature"
+	"tailscale.com/hostinfo"
+	"tailscale.com/tailcfg"
+)
+
+var infoOnce = sync.OnceValue(info)
+
+func init() {
+	feature.Register("tpm")
+	hostinfo.RegisterHostinfoNewHook(func(hi *tailcfg.Hostinfo) {
+		hi.TPM = infoOnce()
+	})
+}
+
+//lint:ignore U1000 used in Linux and Windows builds only
+func infoFromCapabilities(tpm transport.TPM) *tailcfg.TPMInfo {
+	info := new(tailcfg.TPMInfo)
+	toStr := func(s *string) func(*tailcfg.TPMInfo, uint32) {
+		return func(info *tailcfg.TPMInfo, value uint32) {
+			*s += propToString(value)
+		}
+	}
+	for _, cap := range []struct {
+		prop  tpm2.TPMPT
+		apply func(info *tailcfg.TPMInfo, value uint32)
+	}{
+		{tpm2.TPMPTManufacturer, toStr(&info.Manufacturer)},
+		{tpm2.TPMPTVendorString1, toStr(&info.Vendor)},
+		{tpm2.TPMPTVendorString2, toStr(&info.Vendor)},
+		{tpm2.TPMPTVendorString3, toStr(&info.Vendor)},
+		{tpm2.TPMPTVendorString4, toStr(&info.Vendor)},
+		{tpm2.TPMPTRevision, func(info *tailcfg.TPMInfo, value uint32) { info.SpecRevision = int(value) }},
+		{tpm2.TPMPTVendorTPMType, func(info *tailcfg.TPMInfo, value uint32) { info.Model = int(value) }},
+		{tpm2.TPMPTFirmwareVersion1, func(info *tailcfg.TPMInfo, value uint32) { info.FirmwareVersion += uint64(value) << 32 }},
+		{tpm2.TPMPTFirmwareVersion2, func(info *tailcfg.TPMInfo, value uint32) { info.FirmwareVersion += uint64(value) }},
+	} {
+		resp, err := tpm2.GetCapability{
+			Capability:    tpm2.TPMCapTPMProperties,
+			Property:      uint32(cap.prop),
+			PropertyCount: 1,
+		}.Execute(tpm)
+		if err != nil {
+			continue
+		}
+		props, err := resp.CapabilityData.Data.TPMProperties()
+		if err != nil {
+			continue
+		}
+		if len(props.TPMProperty) == 0 {
+			continue
+		}
+		cap.apply(info, props.TPMProperty[0].Value)
+	}
+	return info
+}
+
+// propToString converts TPM_PT property value, which is a uint32, into a
+// string of up to 4 ASCII characters. This encoding applies only to some
+// properties, see
+// https://trustedcomputinggroup.org/resource/tpm-library-specification/ Part
+// 2, section 6.13.
+func propToString(v uint32) string {
+	chars := []byte{
+		byte(v >> 24),
+		byte(v >> 16),
+		byte(v >> 8),
+		byte(v),
+	}
+	// Delete any non-printable ASCII characters.
+	return string(slices.DeleteFunc(chars, func(b byte) bool { return b < ' ' || b > '~' }))
+}
diff --git a/feature/tpm/tpm_linux.go b/feature/tpm/tpm_linux.go
new file mode 100644
index 000000000..a90c0e153
--- /dev/null
+++ b/feature/tpm/tpm_linux.go
@@ -0,0 +1,18 @@
+// Copyright (c) Tailscale Inc & AUTHORS
+// SPDX-License-Identifier: BSD-3-Clause
+
+package tpm
+
+import (
+	"github.com/google/go-tpm/tpm2/transport/linuxtpm"
+	"tailscale.com/tailcfg"
+)
+
+func info() *tailcfg.TPMInfo {
+	t, err := linuxtpm.Open("/dev/tpm0")
+	if err != nil {
+		return nil
+	}
+	defer t.Close()
+	return infoFromCapabilities(t)
+}
diff --git a/feature/tpm/tpm_other.go b/feature/tpm/tpm_other.go
new file mode 100644
index 000000000..ba7c67621
--- /dev/null
+++ b/feature/tpm/tpm_other.go
@@ -0,0 +1,12 @@
+// Copyright (c) Tailscale Inc & AUTHORS
+// SPDX-License-Identifier: BSD-3-Clause
+
+//go:build !linux && !windows
+
+package tpm
+
+import "tailscale.com/tailcfg"
+
+func info() *tailcfg.TPMInfo {
+	return nil
+}
diff --git a/feature/tpm/tpm_test.go b/feature/tpm/tpm_test.go
new file mode 100644
index 000000000..fc0fc178c
--- /dev/null
+++ b/feature/tpm/tpm_test.go
@@ -0,0 +1,19 @@
+// Copyright (c) Tailscale Inc & AUTHORS
+// SPDX-License-Identifier: BSD-3-Clause
+
+package tpm
+
+import "testing"
+
+func TestPropToString(t *testing.T) {
+	for prop, want := range map[uint32]string{
+		0:          "",
+		0x4D534654: "MSFT",
+		0x414D4400: "AMD",
+		0x414D440D: "AMD",
+	} {
+		if got := propToString(prop); got != want {
+			t.Errorf("propToString(0x%x): got %q, want %q", prop, got, want)
+		}
+	}
+}
diff --git a/feature/tpm/tpm_windows.go b/feature/tpm/tpm_windows.go
new file mode 100644
index 000000000..578d687af
--- /dev/null
+++ b/feature/tpm/tpm_windows.go
@@ -0,0 +1,18 @@
+// Copyright (c) Tailscale Inc & AUTHORS
+// SPDX-License-Identifier: BSD-3-Clause
+
+package tpm
+
+import (
+	"github.com/google/go-tpm/tpm2/transport/windowstpm"
+	"tailscale.com/tailcfg"
+)
+
+func info() *tailcfg.TPMInfo {
+	t, err := windowstpm.Open()
+	if err != nil {
+		return nil
+	}
+	defer t.Close()
+	return infoFromCapabilities(t)
+}
diff --git a/go.mod b/go.mod
index 0c1224cf1..f346b1e40 100644
--- a/go.mod
+++ b/go.mod
@@ -44,6 +44,7 @@ require (
 	github.com/golangci/golangci-lint v1.57.1
 	github.com/google/go-cmp v0.6.0
 	github.com/google/go-containerregistry v0.20.2
+	github.com/google/go-tpm v0.9.4
 	github.com/google/gopacket v1.1.19
 	github.com/google/nftables v0.2.1-0.20240414091927-5e242ec57806
 	github.com/google/uuid v1.6.0
diff --git a/go.sum b/go.sum
index 8c8da8d14..bdbae11bb 100644
--- a/go.sum
+++ b/go.sum
@@ -486,6 +486,10 @@ github.com/google/go-cmp v0.6.0 h1:ofyhxvXcZhMsU5ulbFiLKl/XBFqE1GSq7atu8tAmTRI=
 github.com/google/go-cmp v0.6.0/go.mod h1:17dUlkBOakJ0+DkrSSNjCkIjxS6bF9zb3elmeNGIjoY=
 github.com/google/go-containerregistry v0.20.2 h1:B1wPJ1SN/S7pB+ZAimcciVD+r+yV/l/DSArMxlbwseo=
 github.com/google/go-containerregistry v0.20.2/go.mod h1:z38EKdKh4h7IP2gSfUUqEvalZBqs6AoLeWfUy34nQC8=
+github.com/google/go-tpm v0.9.4 h1:awZRf9FwOeTunQmHoDYSHJps3ie6f1UlhS1fOdPEt1I=
+github.com/google/go-tpm v0.9.4/go.mod h1:h9jEsEECg7gtLis0upRBQU+GhYVH6jMjrFxI8u6bVUY=
+github.com/google/go-tpm-tools v0.3.13-0.20230620182252-4639ecce2aba h1:qJEJcuLzH5KDR0gKc0zcktin6KSAwL7+jWKBYceddTc=
+github.com/google/go-tpm-tools v0.3.13-0.20230620182252-4639ecce2aba/go.mod h1:EFYHy8/1y2KfgTAsx7Luu7NGhoxtuVHnNo8jE7FikKc=
 github.com/google/gofuzz v1.0.0/go.mod h1:dBl0BpW6vV/+mYPU4Po3pmUjxk6FQPldtuIdl/M65Eg=
 github.com/google/gofuzz v1.2.0 h1:xRy4A+RhZaiKjJ1bPfwQ8sedCA+YS2YcCHW6ec7JMi0=
 github.com/google/gofuzz v1.2.0/go.mod h1:dBl0BpW6vV/+mYPU4Po3pmUjxk6FQPldtuIdl/M65Eg=
diff --git a/ipn/ipnlocal/local.go b/ipn/ipnlocal/local.go
index 95fe22641..b2998d11c 100644
--- a/ipn/ipnlocal/local.go
+++ b/ipn/ipnlocal/local.go
@@ -258,7 +258,7 @@ type LocalBackend struct {
 	// We intend to relax this in the future and only require holding b.mu when replacing it,
 	// but that requires a better (strictly ordered?) state machine and better management
 	// of [LocalBackend]'s own state that is not tied to the node context.
-	currentNodeAtomic atomic.Pointer[localNodeContext]
+	currentNodeAtomic atomic.Pointer[nodeBackend]
 
 	conf           *conffile.Config   // latest parsed config, or nil if not in declarative mode
 	pm             *profileManager    // mu guards access
@@ -519,7 +519,7 @@ func NewLocalBackend(logf logger.Logf, logID logid.PublicID, sys *tsd.System, lo
 		captiveCancel:         nil, // so that we start checkCaptivePortalLoop when Running
 		needsCaptiveDetection: make(chan bool),
 	}
-	b.currentNodeAtomic.Store(newLocalNodeContext())
+	b.currentNodeAtomic.Store(newNodeBackend())
 	mConn.SetNetInfoCallback(b.setNetInfo)
 
 	if sys.InitialConfig != nil {
@@ -594,12 +594,12 @@ func NewLocalBackend(logf logger.Logf, logID logid.PublicID, sys *tsd.System, lo
 func (b *LocalBackend) Clock() tstime.Clock { return b.clock }
 func (b *LocalBackend) Sys() *tsd.System    { return b.sys }
 
-func (b *LocalBackend) currentNode() *localNodeContext {
+func (b *LocalBackend) currentNode() *nodeBackend {
 	if v := b.currentNodeAtomic.Load(); v != nil || !testenv.InTest() {
 		return v
 	}
 	// Auto-init one in tests for LocalBackend created without the NewLocalBackend constructor...
-	v := newLocalNodeContext()
+	v := newNodeBackend()
 	b.currentNodeAtomic.CompareAndSwap(nil, v)
 	return b.currentNodeAtomic.Load()
 }
@@ -1463,15 +1463,30 @@ func (b *LocalBackend) PeerCaps(src netip.Addr) tailcfg.PeerCapMap {
 	return b.currentNode().PeerCaps(src)
 }
 
-func (b *localNodeContext) AppendMatchingPeers(base []tailcfg.NodeView, pred func(tailcfg.NodeView) bool) []tailcfg.NodeView {
-	b.mu.Lock()
-	defer b.mu.Unlock()
-	ret := base
-	if b.netMap == nil {
-		return ret
+// AppendMatchingPeers returns base with all peers that match pred appended.
+//
+// It acquires b.mu to read the netmap but releases it before calling pred.
+func (nb *nodeBackend) AppendMatchingPeers(base []tailcfg.NodeView, pred func(tailcfg.NodeView) bool) []tailcfg.NodeView {
+	var peers []tailcfg.NodeView
+
+	nb.mu.Lock()
+	if nb.netMap != nil {
+		// All fields on b.netMap are immutable, so this is
+		// safe to copy and use outside the lock.
+		peers = nb.netMap.Peers
 	}
-	for _, peer := range b.netMap.Peers {
-		if pred(peer) {
+	nb.mu.Unlock()
+
+	ret := base
+	for _, peer := range peers {
+		// The peers in b.netMap don't contain updates made via
+		// UpdateNetmapDelta. So only use PeerView in b.netMap for its NodeID,
+		// and then look up the latest copy in b.peers which is updated in
+		// response to UpdateNetmapDelta edits.
+		nb.mu.Lock()
+		peer, ok := nb.peers[peer.ID()]
+		nb.mu.Unlock()
+		if ok && pred(peer) {
 			ret = append(ret, peer)
 		}
 	}
@@ -1480,21 +1495,21 @@ func (b *localNodeContext) AppendMatchingPeers(base []tailcfg.NodeView, pred fun
 
 // PeerCaps returns the capabilities that remote src IP has to
 // ths current node.
-func (b *localNodeContext) PeerCaps(src netip.Addr) tailcfg.PeerCapMap {
-	b.mu.Lock()
-	defer b.mu.Unlock()
-	return b.peerCapsLocked(src)
+func (nb *nodeBackend) PeerCaps(src netip.Addr) tailcfg.PeerCapMap {
+	nb.mu.Lock()
+	defer nb.mu.Unlock()
+	return nb.peerCapsLocked(src)
 }
 
-func (b *localNodeContext) peerCapsLocked(src netip.Addr) tailcfg.PeerCapMap {
-	if b.netMap == nil {
+func (nb *nodeBackend) peerCapsLocked(src netip.Addr) tailcfg.PeerCapMap {
+	if nb.netMap == nil {
 		return nil
 	}
-	filt := b.filterAtomic.Load()
+	filt := nb.filterAtomic.Load()
 	if filt == nil {
 		return nil
 	}
-	addrs := b.netMap.GetAddresses()
+	addrs := nb.netMap.GetAddresses()
 	for i := range addrs.Len() {
 		a := addrs.At(i)
 		if !a.IsSingleIP() {
@@ -1508,8 +1523,8 @@ func (b *localNodeContext) peerCapsLocked(src netip.Addr) tailcfg.PeerCapMap {
 	return nil
 }
 
-func (b *localNodeContext) GetFilterForTest() *filter.Filter {
-	return b.filterAtomic.Load()
+func (nb *nodeBackend) GetFilterForTest() *filter.Filter {
+	return nb.filterAtomic.Load()
 }
 
 // SetControlClientStatus is the callback invoked by the control client whenever it posts a new status.
@@ -2019,14 +2034,14 @@ func (b *LocalBackend) UpdateNetmapDelta(muts []netmap.NodeMutation) (handled bo
 	return true
 }
 
-func (c *localNodeContext) netMapWithPeers() *netmap.NetworkMap {
-	c.mu.Lock()
-	defer c.mu.Unlock()
-	if c.netMap == nil {
+func (nb *nodeBackend) netMapWithPeers() *netmap.NetworkMap {
+	nb.mu.Lock()
+	defer nb.mu.Unlock()
+	if nb.netMap == nil {
 		return nil
 	}
-	nm := ptr.To(*c.netMap) // shallow clone
-	nm.Peers = slicesx.MapValues(c.peers)
+	nm := ptr.To(*nb.netMap) // shallow clone
+	nm.Peers = slicesx.MapValues(nb.peers)
 	slices.SortFunc(nm.Peers, func(a, b tailcfg.NodeView) int {
 		return cmp.Compare(a.ID(), b.ID())
 	})
@@ -2063,10 +2078,10 @@ func (b *LocalBackend) pickNewAutoExitNode() {
 	b.send(ipn.Notify{Prefs: &newPrefs})
 }
 
-func (c *localNodeContext) UpdateNetmapDelta(muts []netmap.NodeMutation) (handled bool) {
-	c.mu.Lock()
-	defer c.mu.Unlock()
-	if c.netMap == nil || len(c.peers) == 0 {
+func (nb *nodeBackend) UpdateNetmapDelta(muts []netmap.NodeMutation) (handled bool) {
+	nb.mu.Lock()
+	defer nb.mu.Unlock()
+	if nb.netMap == nil || len(nb.peers) == 0 {
 		return false
 	}
 
@@ -2078,7 +2093,7 @@ func (c *localNodeContext) UpdateNetmapDelta(muts []netmap.NodeMutation) (handle
 	for _, m := range muts {
 		n, ok := mutableNodes[m.NodeIDBeingMutated()]
 		if !ok {
-			nv, ok := c.peers[m.NodeIDBeingMutated()]
+			nv, ok := nb.peers[m.NodeIDBeingMutated()]
 			if !ok {
 				// TODO(bradfitz): unexpected metric?
 				return false
@@ -2089,7 +2104,7 @@ func (c *localNodeContext) UpdateNetmapDelta(muts []netmap.NodeMutation) (handle
 		m.Apply(n)
 	}
 	for nid, n := range mutableNodes {
-		c.peers[nid] = n.View()
+		nb.peers[nid] = n.View()
 	}
 	return true
 }
@@ -2250,10 +2265,10 @@ func (b *LocalBackend) PeersForTest() []tailcfg.NodeView {
 	return b.currentNode().PeersForTest()
 }
 
-func (b *localNodeContext) PeersForTest() []tailcfg.NodeView {
-	b.mu.Lock()
-	defer b.mu.Unlock()
-	ret := slicesx.MapValues(b.peers)
+func (nb *nodeBackend) PeersForTest() []tailcfg.NodeView {
+	nb.mu.Lock()
+	defer nb.mu.Unlock()
+	ret := slicesx.MapValues(nb.peers)
 	slices.SortFunc(ret, func(a, b tailcfg.NodeView) int {
 		return cmp.Compare(a.ID(), b.ID())
 	})
@@ -2532,12 +2547,12 @@ var invalidPacketFilterWarnable = health.Register(&health.Warnable{
 // b.mu must be held.
 func (b *LocalBackend) updateFilterLocked(prefs ipn.PrefsView) {
 	// TODO(nickkhyl) split this into two functions:
-	// - (*localNodeContext).RebuildFilters() (normalFilter, jailedFilter *filter.Filter, changed bool),
+	// - (*nodeBackend).RebuildFilters() (normalFilter, jailedFilter *filter.Filter, changed bool),
 	//   which would return packet filters for the current state and whether they changed since the last call.
 	// - (*LocalBackend).updateFilters(), which would use the above to update the engine with the new filters,
 	//    notify b.sshServer, etc.
 	//
-	// For this, we would need to plumb a few more things into the [localNodeContext]. Most importantly,
+	// For this, we would need to plumb a few more things into the [nodeBackend]. Most importantly,
 	// the current [ipn.PrefsView]), but also maybe also a b.logf and a b.health?
 	//
 	// NOTE(danderson): keep change detection as the first thing in
@@ -2823,8 +2838,8 @@ func (b *LocalBackend) setFilter(f *filter.Filter) {
 	b.e.SetFilter(f)
 }
 
-func (c *localNodeContext) setFilter(f *filter.Filter) {
-	c.filterAtomic.Store(f)
+func (nb *nodeBackend) setFilter(f *filter.Filter) {
+	nb.filterAtomic.Store(f)
 }
 
 var removeFromDefaultRoute = []netip.Prefix{
@@ -3886,7 +3901,7 @@ func (b *LocalBackend) parseWgStatusLocked(s *wgengine.Status) (ret ipn.EngineSt
 // in Hostinfo. When the user preferences currently request "shields up"
 // mode, all inbound connections are refused, so services are not reported.
 // Otherwise, shouldUploadServices respects NetMap.CollectServices.
-// TODO(nickkhyl): move this into [localNodeContext]?
+// TODO(nickkhyl): move this into [nodeBackend]?
 func (b *LocalBackend) shouldUploadServices() bool {
 	b.mu.Lock()
 	defer b.mu.Unlock()
@@ -4758,10 +4773,10 @@ func (b *LocalBackend) NetMap() *netmap.NetworkMap {
 	return b.currentNode().NetMap()
 }
 
-func (c *localNodeContext) NetMap() *netmap.NetworkMap {
-	c.mu.Lock()
-	defer c.mu.Unlock()
-	return c.netMap
+func (nb *nodeBackend) NetMap() *netmap.NetworkMap {
+	nb.mu.Lock()
+	defer nb.mu.Unlock()
+	return nb.netMap
 }
 
 func (b *LocalBackend) isEngineBlocked() bool {
@@ -5003,10 +5018,10 @@ func shouldUseOneCGNATRoute(logf logger.Logf, mon *netmon.Monitor, controlKnobs
 	return false
 }
 
-func (c *localNodeContext) dnsConfigForNetmap(prefs ipn.PrefsView, selfExpired bool, logf logger.Logf, versionOS string) *dns.Config {
-	c.mu.Lock()
-	defer c.mu.Unlock()
-	return dnsConfigForNetmap(c.netMap, c.peers, prefs, selfExpired, logf, versionOS)
+func (nb *nodeBackend) dnsConfigForNetmap(prefs ipn.PrefsView, selfExpired bool, logf logger.Logf, versionOS string) *dns.Config {
+	nb.mu.Lock()
+	defer nb.mu.Unlock()
+	return dnsConfigForNetmap(nb.netMap, nb.peers, prefs, selfExpired, logf, versionOS)
 }
 
 // dnsConfigForNetmap returns a *dns.Config for the given netmap,
@@ -5041,6 +5056,8 @@ func dnsConfigForNetmap(nm *netmap.NetworkMap, peers map[tailcfg.NodeID]tailcfg.
 		!nm.GetAddresses().ContainsFunc(tsaddr.PrefixIs4)
 	dcfg.OnlyIPv6 = selfV6Only
 
+	wantAAAA := nm.AllCaps.Contains(tailcfg.NodeAttrMagicDNSPeerAAAA)
+
 	// Populate MagicDNS records. We do this unconditionally so that
 	// quad-100 can always respond to MagicDNS queries, even if the OS
 	// isn't configured to make MagicDNS resolution truly
@@ -5077,7 +5094,7 @@ func dnsConfigForNetmap(nm *netmap.NetworkMap, peers map[tailcfg.NodeID]tailcfg.
 			// https://github.com/tailscale/tailscale/issues/1152
 			// tracks adding the right capability reporting to
 			// enable AAAA in MagicDNS.
-			if addr.Addr().Is6() && have4 {
+			if addr.Addr().Is6() && have4 && !wantAAAA {
 				continue
 			}
 			ips = append(ips, addr.Addr())
@@ -6129,12 +6146,12 @@ func (b *LocalBackend) setAutoExitNodeIDLockedOnEntry(unlock unlockOnce) (newPre
 	return newPrefs
 }
 
-func (c *localNodeContext) SetNetMap(nm *netmap.NetworkMap) {
-	c.mu.Lock()
-	defer c.mu.Unlock()
-	c.netMap = nm
-	c.updateNodeByAddrLocked()
-	c.updatePeersLocked()
+func (nb *nodeBackend) SetNetMap(nm *netmap.NetworkMap) {
+	nb.mu.Lock()
+	defer nb.mu.Unlock()
+	nb.netMap = nm
+	nb.updateNodeByAddrLocked()
+	nb.updatePeersLocked()
 }
 
 // setNetMapLocked updates the LocalBackend state to reflect the newly
@@ -6209,25 +6226,25 @@ func (b *LocalBackend) setNetMapLocked(nm *netmap.NetworkMap) {
 	b.driveNotifyCurrentSharesLocked()
 }
 
-func (b *localNodeContext) updateNodeByAddrLocked() {
-	nm := b.netMap
+func (nb *nodeBackend) updateNodeByAddrLocked() {
+	nm := nb.netMap
 	if nm == nil {
-		b.nodeByAddr = nil
+		nb.nodeByAddr = nil
 		return
 	}
 
 	// Update the nodeByAddr index.
-	if b.nodeByAddr == nil {
-		b.nodeByAddr = map[netip.Addr]tailcfg.NodeID{}
+	if nb.nodeByAddr == nil {
+		nb.nodeByAddr = map[netip.Addr]tailcfg.NodeID{}
 	}
 	// First pass, mark everything unwanted.
-	for k := range b.nodeByAddr {
-		b.nodeByAddr[k] = 0
+	for k := range nb.nodeByAddr {
+		nb.nodeByAddr[k] = 0
 	}
 	addNode := func(n tailcfg.NodeView) {
 		for _, ipp := range n.Addresses().All() {
 			if ipp.IsSingleIP() {
-				b.nodeByAddr[ipp.Addr()] = n.ID()
+				nb.nodeByAddr[ipp.Addr()] = n.ID()
 			}
 		}
 	}
@@ -6238,34 +6255,34 @@ func (b *localNodeContext) updateNodeByAddrLocked() {
 		addNode(p)
 	}
 	// Third pass, actually delete the unwanted items.
-	for k, v := range b.nodeByAddr {
+	for k, v := range nb.nodeByAddr {
 		if v == 0 {
-			delete(b.nodeByAddr, k)
+			delete(nb.nodeByAddr, k)
 		}
 	}
 }
 
-func (b *localNodeContext) updatePeersLocked() {
-	nm := b.netMap
+func (nb *nodeBackend) updatePeersLocked() {
+	nm := nb.netMap
 	if nm == nil {
-		b.peers = nil
+		nb.peers = nil
 		return
 	}
 
 	// First pass, mark everything unwanted.
-	for k := range b.peers {
-		b.peers[k] = tailcfg.NodeView{}
+	for k := range nb.peers {
+		nb.peers[k] = tailcfg.NodeView{}
 	}
 
 	// Second pass, add everything wanted.
 	for _, p := range nm.Peers {
-		mak.Set(&b.peers, p.ID(), p)
+		mak.Set(&nb.peers, p.ID(), p)
 	}
 
 	// Third pass, remove deleted things.
-	for k, v := range b.peers {
+	for k, v := range nb.peers {
 		if !v.Valid() {
-			delete(b.peers, k)
+			delete(nb.peers, k)
 		}
 	}
 }
@@ -6652,14 +6669,14 @@ func (b *LocalBackend) TestOnlyPublicKeys() (machineKey key.MachinePublic, nodeK
 
 // PeerHasCap reports whether the peer with the given Tailscale IP addresses
 // contains the given capability string, with any value(s).
-func (b *localNodeContext) PeerHasCap(addr netip.Addr, wantCap tailcfg.PeerCapability) bool {
-	b.mu.Lock()
-	defer b.mu.Unlock()
-	return b.peerHasCapLocked(addr, wantCap)
+func (nb *nodeBackend) PeerHasCap(addr netip.Addr, wantCap tailcfg.PeerCapability) bool {
+	nb.mu.Lock()
+	defer nb.mu.Unlock()
+	return nb.peerHasCapLocked(addr, wantCap)
 }
 
-func (b *localNodeContext) peerHasCapLocked(addr netip.Addr, wantCap tailcfg.PeerCapability) bool {
-	return b.peerCapsLocked(addr).HasCapability(wantCap)
+func (nb *nodeBackend) peerHasCapLocked(addr netip.Addr, wantCap tailcfg.PeerCapability) bool {
+	return nb.peerCapsLocked(addr).HasCapability(wantCap)
 }
 
 // SetDNS adds a DNS record for the given domain name & TXT record
@@ -6722,16 +6739,16 @@ func peerAPIURL(ip netip.Addr, port uint16) string {
 	return fmt.Sprintf("http://%v", netip.AddrPortFrom(ip, port))
 }
 
-func (c *localNodeContext) PeerHasPeerAPI(p tailcfg.NodeView) bool {
-	return c.PeerAPIBase(p) != ""
+func (nb *nodeBackend) PeerHasPeerAPI(p tailcfg.NodeView) bool {
+	return nb.PeerAPIBase(p) != ""
 }
 
 // PeerAPIBase returns the "http://ip:port" URL base to reach peer's PeerAPI,
 // or the empty string if the peer is invalid or doesn't support PeerAPI.
-func (c *localNodeContext) PeerAPIBase(p tailcfg.NodeView) string {
-	c.mu.Lock()
-	nm := c.netMap
-	c.mu.Unlock()
+func (nb *nodeBackend) PeerAPIBase(p tailcfg.NodeView) string {
+	nb.mu.Lock()
+	nm := nb.netMap
+	nb.mu.Unlock()
 	return peerAPIBase(nm, p)
 }
 
@@ -6972,10 +6989,10 @@ func exitNodeCanProxyDNS(nm *netmap.NetworkMap, peers map[tailcfg.NodeID]tailcfg
 	return "", false
 }
 
-func (c *localNodeContext) exitNodeCanProxyDNS(exitNodeID tailcfg.StableNodeID) (dohURL string, ok bool) {
-	c.mu.Lock()
-	defer c.mu.Unlock()
-	return exitNodeCanProxyDNS(c.netMap, c.peers, exitNodeID)
+func (nb *nodeBackend) exitNodeCanProxyDNS(exitNodeID tailcfg.StableNodeID) (dohURL string, ok bool) {
+	nb.mu.Lock()
+	defer nb.mu.Unlock()
+	return exitNodeCanProxyDNS(nb.netMap, nb.peers, exitNodeID)
 }
 
 // wireguardExitNodeDNSResolvers returns the DNS resolvers to use for a
@@ -7396,7 +7413,7 @@ func (b *LocalBackend) resetForProfileChangeLockedOnEntry(unlock unlockOnce) err
 		// down, so no need to do any work.
 		return nil
 	}
-	b.currentNodeAtomic.Store(newLocalNodeContext())
+	b.currentNodeAtomic.Store(newNodeBackend())
 	b.setNetMapLocked(nil) // Reset netmap.
 	b.updateFilterLocked(ipn.PrefsView{})
 	// Reset the NetworkMap in the engine
@@ -8086,7 +8103,7 @@ func (b *LocalBackend) startAutoUpdate(logPrefix string) (retErr error) {
 // rules that require a source IP to have a certain node capability.
 //
 // TODO(bradfitz): optimize this later if/when it matters.
-// TODO(nickkhyl): move this into [localNodeContext] along with [LocalBackend.updateFilterLocked].
+// TODO(nickkhyl): move this into [nodeBackend] along with [LocalBackend.updateFilterLocked].
 func (b *LocalBackend) srcIPHasCapForFilter(srcIP netip.Addr, cap tailcfg.NodeCapability) bool {
 	if cap == "" {
 		// Shouldn't happen, but just in case.
diff --git a/ipn/ipnlocal/local_node_context.go b/ipn/ipnlocal/node_backend.go
similarity index 65%
rename from ipn/ipnlocal/local_node_context.go
rename to ipn/ipnlocal/node_backend.go
index 871880893..415c32ccf 100644
--- a/ipn/ipnlocal/local_node_context.go
+++ b/ipn/ipnlocal/node_backend.go
@@ -18,29 +18,29 @@ import (
 	"tailscale.com/wgengine/filter"
 )
 
-// localNodeContext holds the [LocalBackend]'s context tied to a local node (usually the current one).
+// nodeBackend is node-specific [LocalBackend] state. It is usually the current node.
 //
 // Its exported methods are safe for concurrent use, but the struct is not a snapshot of state at a given moment;
 // its state can change between calls. For example, asking for the same value (e.g., netmap or prefs) twice
 // may return different results. Returned values are immutable and safe for concurrent use.
 //
-// If both the [LocalBackend]'s internal mutex and the [localNodeContext] mutex must be held at the same time,
+// If both the [LocalBackend]'s internal mutex and the [nodeBackend] mutex must be held at the same time,
 // the [LocalBackend] mutex must be acquired first. See the comment on the [LocalBackend] field for more details.
 //
-// Two pointers to different [localNodeContext] instances represent different local nodes.
-// However, there's currently a bug where a new [localNodeContext] might not be created
+// Two pointers to different [nodeBackend] instances represent different local nodes.
+// However, there's currently a bug where a new [nodeBackend] might not be created
 // during an implicit node switch (see tailscale/corp#28014).
 
 // In the future, we might want to include at least the following in this struct (in addition to the current fields).
 // However, not everything should be exported or otherwise made available to the outside world (e.g. [ipnext] extensions,
 // peer API handlers, etc.).
-//   - [ipn.State]: when the LocalBackend switches to a different [localNodeContext], it can update the state of the old one.
+//   - [ipn.State]: when the LocalBackend switches to a different [nodeBackend], it can update the state of the old one.
 //   - [ipn.LoginProfileView] and [ipn.Prefs]: we should update them when the [profileManager] reports changes to them.
 //     In the future, [profileManager] (and the corresponding methods of the [LocalBackend]) can be made optional,
 //     and something else could be used to set them once or update them as needed.
 //   - [tailcfg.HostinfoView]: it includes certain fields that are tied to the current profile/node/prefs. We should also
 //     update to build it once instead of mutating it in twelvety different places.
-//   - [filter.Filter] (normal and jailed, along with the filterHash): the localNodeContext could have a method to (re-)build
+//   - [filter.Filter] (normal and jailed, along with the filterHash): the nodeBackend could have a method to (re-)build
 //     the filter for the current netmap/prefs (see [LocalBackend.updateFilterLocked]), and it needs to track the current
 //     filters and their hash.
 //   - Fields related to a requested or required (re-)auth: authURL, authURLTime, authActor, keyExpired, etc.
@@ -51,7 +51,7 @@ import (
 // It should not include any fields used by specific features that don't belong in [LocalBackend].
 // Even if they're tied to the local node, instead of moving them here, we should extract the entire feature
 // into a separate package and have it install proper hooks.
-type localNodeContext struct {
+type nodeBackend struct {
 	// filterAtomic is a stateful packet filter. Immutable once created, but can be
 	// replaced with a new one.
 	filterAtomic atomic.Pointer[filter.Filter]
@@ -71,33 +71,33 @@ type localNodeContext struct {
 	// peers is the set of current peers and their current values after applying
 	// delta node mutations as they come in (with mu held). The map values can be
 	// given out to callers, but the map itself can be mutated in place (with mu held)
-	// and must not escape the [localNodeContext].
+	// and must not escape the [nodeBackend].
 	peers map[tailcfg.NodeID]tailcfg.NodeView
 
 	// nodeByAddr maps nodes' own addresses (excluding subnet routes) to node IDs.
-	// It is mutated in place (with mu held) and must not escape the [localNodeContext].
+	// It is mutated in place (with mu held) and must not escape the [nodeBackend].
 	nodeByAddr map[netip.Addr]tailcfg.NodeID
 }
 
-func newLocalNodeContext() *localNodeContext {
-	cn := &localNodeContext{}
+func newNodeBackend() *nodeBackend {
+	cn := &nodeBackend{}
 	// Default filter blocks everything and logs nothing.
 	noneFilter := filter.NewAllowNone(logger.Discard, &netipx.IPSet{})
 	cn.filterAtomic.Store(noneFilter)
 	return cn
 }
 
-func (c *localNodeContext) Self() tailcfg.NodeView {
-	c.mu.Lock()
-	defer c.mu.Unlock()
-	if c.netMap == nil {
+func (nb *nodeBackend) Self() tailcfg.NodeView {
+	nb.mu.Lock()
+	defer nb.mu.Unlock()
+	if nb.netMap == nil {
 		return tailcfg.NodeView{}
 	}
-	return c.netMap.SelfNode
+	return nb.netMap.SelfNode
 }
 
-func (c *localNodeContext) SelfUserID() tailcfg.UserID {
-	self := c.Self()
+func (nb *nodeBackend) SelfUserID() tailcfg.UserID {
+	self := nb.Self()
 	if !self.Valid() {
 		return 0
 	}
@@ -105,59 +105,59 @@ func (c *localNodeContext) SelfUserID() tailcfg.UserID {
 }
 
 // SelfHasCap reports whether the specified capability was granted to the self node in the most recent netmap.
-func (c *localNodeContext) SelfHasCap(wantCap tailcfg.NodeCapability) bool {
-	return c.SelfHasCapOr(wantCap, false)
+func (nb *nodeBackend) SelfHasCap(wantCap tailcfg.NodeCapability) bool {
+	return nb.SelfHasCapOr(wantCap, false)
 }
 
-// SelfHasCapOr is like [localNodeContext.SelfHasCap], but returns the specified default value
+// SelfHasCapOr is like [nodeBackend.SelfHasCap], but returns the specified default value
 // if the netmap is not available yet.
-func (c *localNodeContext) SelfHasCapOr(wantCap tailcfg.NodeCapability, def bool) bool {
-	c.mu.Lock()
-	defer c.mu.Unlock()
-	if c.netMap == nil {
+func (nb *nodeBackend) SelfHasCapOr(wantCap tailcfg.NodeCapability, def bool) bool {
+	nb.mu.Lock()
+	defer nb.mu.Unlock()
+	if nb.netMap == nil {
 		return def
 	}
-	return c.netMap.AllCaps.Contains(wantCap)
+	return nb.netMap.AllCaps.Contains(wantCap)
 }
 
-func (c *localNodeContext) NetworkProfile() ipn.NetworkProfile {
-	c.mu.Lock()
-	defer c.mu.Unlock()
+func (nb *nodeBackend) NetworkProfile() ipn.NetworkProfile {
+	nb.mu.Lock()
+	defer nb.mu.Unlock()
 	return ipn.NetworkProfile{
 		// These are ok to call with nil netMap.
-		MagicDNSName: c.netMap.MagicDNSSuffix(),
-		DomainName:   c.netMap.DomainName(),
+		MagicDNSName: nb.netMap.MagicDNSSuffix(),
+		DomainName:   nb.netMap.DomainName(),
 	}
 }
 
 // TODO(nickkhyl): update it to return a [tailcfg.DERPMapView]?
-func (c *localNodeContext) DERPMap() *tailcfg.DERPMap {
-	c.mu.Lock()
-	defer c.mu.Unlock()
-	if c.netMap == nil {
+func (nb *nodeBackend) DERPMap() *tailcfg.DERPMap {
+	nb.mu.Lock()
+	defer nb.mu.Unlock()
+	if nb.netMap == nil {
 		return nil
 	}
-	return c.netMap.DERPMap
+	return nb.netMap.DERPMap
 }
 
-func (c *localNodeContext) NodeByAddr(ip netip.Addr) (_ tailcfg.NodeID, ok bool) {
-	c.mu.Lock()
-	defer c.mu.Unlock()
-	nid, ok := c.nodeByAddr[ip]
+func (nb *nodeBackend) NodeByAddr(ip netip.Addr) (_ tailcfg.NodeID, ok bool) {
+	nb.mu.Lock()
+	defer nb.mu.Unlock()
+	nid, ok := nb.nodeByAddr[ip]
 	return nid, ok
 }
 
-func (c *localNodeContext) NodeByKey(k key.NodePublic) (_ tailcfg.NodeID, ok bool) {
-	c.mu.Lock()
-	defer c.mu.Unlock()
-	if c.netMap == nil {
+func (nb *nodeBackend) NodeByKey(k key.NodePublic) (_ tailcfg.NodeID, ok bool) {
+	nb.mu.Lock()
+	defer nb.mu.Unlock()
+	if nb.netMap == nil {
 		return 0, false
 	}
-	if self := c.netMap.SelfNode; self.Valid() && self.Key() == k {
+	if self := nb.netMap.SelfNode; self.Valid() && self.Key() == k {
 		return self.ID(), true
 	}
 	// TODO(bradfitz,nickkhyl): add nodeByKey like nodeByAddr instead of walking peers.
-	for _, n := range c.peers {
+	for _, n := range nb.peers {
 		if n.Key() == k {
 			return n.ID(), true
 		}
@@ -165,17 +165,17 @@ func (c *localNodeContext) NodeByKey(k key.NodePublic) (_ tailcfg.NodeID, ok boo
 	return 0, false
 }
 
-func (c *localNodeContext) PeerByID(id tailcfg.NodeID) (_ tailcfg.NodeView, ok bool) {
-	c.mu.Lock()
-	defer c.mu.Unlock()
-	n, ok := c.peers[id]
+func (nb *nodeBackend) PeerByID(id tailcfg.NodeID) (_ tailcfg.NodeView, ok bool) {
+	nb.mu.Lock()
+	defer nb.mu.Unlock()
+	n, ok := nb.peers[id]
 	return n, ok
 }
 
-func (c *localNodeContext) UserByID(id tailcfg.UserID) (_ tailcfg.UserProfileView, ok bool) {
-	c.mu.Lock()
-	nm := c.netMap
-	c.mu.Unlock()
+func (nb *nodeBackend) UserByID(id tailcfg.UserID) (_ tailcfg.UserProfileView, ok bool) {
+	nb.mu.Lock()
+	nm := nb.netMap
+	nb.mu.Unlock()
 	if nm == nil {
 		return tailcfg.UserProfileView{}, false
 	}
@@ -184,10 +184,10 @@ func (c *localNodeContext) UserByID(id tailcfg.UserID) (_ tailcfg.UserProfileVie
 }
 
 // Peers returns all the current peers in an undefined order.
-func (c *localNodeContext) Peers() []tailcfg.NodeView {
-	c.mu.Lock()
-	defer c.mu.Unlock()
-	return slicesx.MapValues(c.peers)
+func (nb *nodeBackend) Peers() []tailcfg.NodeView {
+	nb.mu.Lock()
+	defer nb.mu.Unlock()
+	return slicesx.MapValues(nb.peers)
 }
 
 // unlockedNodesPermitted reports whether any peer with theUnsignedPeerAPIOnly bool set true has any of its allowed IPs
@@ -195,13 +195,13 @@ func (c *localNodeContext) Peers() []tailcfg.NodeView {
 //
 // TODO(nickkhyl): It is here temporarily until we can move the whole [LocalBackend.updateFilterLocked] here,
 // but change it so it builds and returns a filter for the current netmap/prefs instead of re-configuring the engine filter.
-// Something like (*localNodeContext).RebuildFilters() (filter, jailedFilter *filter.Filter, changed bool) perhaps?
-func (c *localNodeContext) unlockedNodesPermitted(packetFilter []filter.Match) bool {
-	c.mu.Lock()
-	defer c.mu.Unlock()
-	return packetFilterPermitsUnlockedNodes(c.peers, packetFilter)
+// Something like (*nodeBackend).RebuildFilters() (filter, jailedFilter *filter.Filter, changed bool) perhaps?
+func (nb *nodeBackend) unlockedNodesPermitted(packetFilter []filter.Match) bool {
+	nb.mu.Lock()
+	defer nb.mu.Unlock()
+	return packetFilterPermitsUnlockedNodes(nb.peers, packetFilter)
 }
 
-func (c *localNodeContext) filter() *filter.Filter {
-	return c.filterAtomic.Load()
+func (nb *nodeBackend) filter() *filter.Filter {
+	return nb.filterAtomic.Load()
 }
diff --git a/ipn/ipnlocal/taildrop.go b/ipn/ipnlocal/taildrop.go
index 17ca40926..d8113d219 100644
--- a/ipn/ipnlocal/taildrop.go
+++ b/ipn/ipnlocal/taildrop.go
@@ -194,8 +194,8 @@ func (b *LocalBackend) FileTargets() ([]*apitype.FileTarget, error) {
 		if !p.Valid() || p.Hostinfo().OS() == "tvOS" {
 			return false
 		}
-		if self != p.User() {
-			return false
+		if self == p.User() {
+			return true
 		}
 		if p.Addresses().Len() != 0 && cn.PeerHasCap(p.Addresses().At(0).Addr(), tailcfg.PeerCapabilityFileSharingTarget) {
 			// Explicitly noted in the netmap ACL caps as a target.
diff --git a/logtail/logtail.go b/logtail/logtail.go
index a617397f9..b355addd2 100644
--- a/logtail/logtail.go
+++ b/logtail/logtail.go
@@ -15,9 +15,7 @@ import (
 	"log"
 	mrand "math/rand/v2"
 	"net/http"
-	"net/netip"
 	"os"
-	"regexp"
 	"runtime"
 	"slices"
 	"strconv"
@@ -29,7 +27,6 @@ import (
 	"tailscale.com/envknob"
 	"tailscale.com/net/netmon"
 	"tailscale.com/net/sockstats"
-	"tailscale.com/net/tsaddr"
 	"tailscale.com/tstime"
 	tslogger "tailscale.com/types/logger"
 	"tailscale.com/types/logid"
@@ -833,8 +830,6 @@ func (l *Logger) Logf(format string, args ...any) {
 	fmt.Fprintf(l, format, args...)
 }
 
-var obscureIPs = envknob.RegisterBool("TS_OBSCURE_LOGGED_IPS")
-
 // Write logs an encoded JSON blob.
 //
 // If the []byte passed to Write is not an encoded JSON blob,
@@ -859,10 +854,6 @@ func (l *Logger) Write(buf []byte) (int, error) {
 		}
 	}
 
-	if obscureIPs() {
-		buf = redactIPs(buf)
-	}
-
 	l.writeLock.Lock()
 	defer l.writeLock.Unlock()
 
@@ -871,40 +862,6 @@ func (l *Logger) Write(buf []byte) (int, error) {
 	return inLen, err
 }
 
-var (
-	regexMatchesIPv6 = regexp.MustCompile(`([0-9a-fA-F]{1,4}):([0-9a-fA-F]{1,4}):([0-9a-fA-F:]{1,4})*`)
-	regexMatchesIPv4 = regexp.MustCompile(`(\d{1,3})\.(\d{1,3})\.\d{1,3}\.\d{1,3}`)
-)
-
-// redactIPs is a helper function used in Write() to redact IPs (other than tailscale IPs).
-// This function takes a log line as a byte slice and
-// uses regex matching to parse and find IP addresses. Based on if the IP address is IPv4 or
-// IPv6, it parses and replaces the end of the addresses with an "x". This function returns the
-// log line with the IPs redacted.
-func redactIPs(buf []byte) []byte {
-	out := regexMatchesIPv6.ReplaceAllFunc(buf, func(b []byte) []byte {
-		ip, err := netip.ParseAddr(string(b))
-		if err != nil || tsaddr.IsTailscaleIP(ip) {
-			return b // don't change this one
-		}
-
-		prefix := bytes.Split(b, []byte(":"))
-		return bytes.Join(append(prefix[:2], []byte("x")), []byte(":"))
-	})
-
-	out = regexMatchesIPv4.ReplaceAllFunc(out, func(b []byte) []byte {
-		ip, err := netip.ParseAddr(string(b))
-		if err != nil || tsaddr.IsTailscaleIP(ip) {
-			return b // don't change this one
-		}
-
-		prefix := bytes.Split(b, []byte("."))
-		return bytes.Join(append(prefix[:2], []byte("x.x")), []byte("."))
-	})
-
-	return []byte(out)
-}
-
 var (
 	openBracketV = []byte("[v")
 	v1           = []byte("[v1] ")
diff --git a/logtail/logtail_test.go b/logtail/logtail_test.go
index 3ea630406..b8c46c448 100644
--- a/logtail/logtail_test.go
+++ b/logtail/logtail_test.go
@@ -15,7 +15,6 @@ import (
 	"time"
 
 	"github.com/go-json-experiment/json/jsontext"
-	"tailscale.com/envknob"
 	"tailscale.com/tstest"
 	"tailscale.com/tstime"
 	"tailscale.com/util/must"
@@ -316,85 +315,6 @@ func TestLoggerWriteResult(t *testing.T) {
 		t.Errorf("mismatch.\n got: %#q\nwant: %#q", back, want)
 	}
 }
-func TestRedact(t *testing.T) {
-	envknob.Setenv("TS_OBSCURE_LOGGED_IPS", "true")
-	tests := []struct {
-		in   string
-		want string
-	}{
-		// tests for ipv4 addresses
-		{
-			"120.100.30.47",
-			"120.100.x.x",
-		},
-		{
-			"192.167.0.1/65",
-			"192.167.x.x/65",
-		},
-		{
-			"node [5Btdd] d:e89a3384f526d251 now using 10.0.0.222:41641 mtu=1360 tx=d81a8a35a0ce",
-			"node [5Btdd] d:e89a3384f526d251 now using 10.0.x.x:41641 mtu=1360 tx=d81a8a35a0ce",
-		},
-		//tests for ipv6 addresses
-		{
-			"2001:0db8:85a3:0000:0000:8a2e:0370:7334",
-			"2001:0db8:x",
-		},
-		{
-			"2345:0425:2CA1:0000:0000:0567:5673:23b5",
-			"2345:0425:x",
-		},
-		{
-			"2601:645:8200:edf0::c9de/64",
-			"2601:645:x/64",
-		},
-		{
-			"node [5Btdd] d:e89a3384f526d251 now using 2051:0000:140F::875B:131C mtu=1360 tx=d81a8a35a0ce",
-			"node [5Btdd] d:e89a3384f526d251 now using 2051:0000:x mtu=1360 tx=d81a8a35a0ce",
-		},
-		{
-			"2601:645:8200:edf0::c9de/64 2601:645:8200:edf0:1ce9:b17d:71f5:f6a3/64",
-			"2601:645:x/64 2601:645:x/64",
-		},
-		//tests for tailscale ip addresses
-		{
-			"100.64.5.6",
-			"100.64.5.6",
-		},
-		{
-			"fd7a:115c:a1e0::/96",
-			"fd7a:115c:a1e0::/96",
-		},
-		//tests for ipv6 and ipv4 together
-		{
-			"192.167.0.1 2001:0db8:85a3:0000:0000:8a2e:0370:7334",
-			"192.167.x.x 2001:0db8:x",
-		},
-		{
-			"node [5Btdd] d:e89a3384f526d251 now using 10.0.0.222:41641 mtu=1360 tx=d81a8a35a0ce 2345:0425:2CA1::0567:5673:23b5",
-			"node [5Btdd] d:e89a3384f526d251 now using 10.0.x.x:41641 mtu=1360 tx=d81a8a35a0ce 2345:0425:x",
-		},
-		{
-			"100.64.5.6 2091:0db8:85a3:0000:0000:8a2e:0370:7334",
-			"100.64.5.6 2091:0db8:x",
-		},
-		{
-			"192.167.0.1 120.100.30.47 2041:0000:140F::875B:131B",
-			"192.167.x.x 120.100.x.x 2041:0000:x",
-		},
-		{
-			"fd7a:115c:a1e0::/96 192.167.0.1 2001:0db8:85a3:0000:0000:8a2e:0370:7334",
-			"fd7a:115c:a1e0::/96 192.167.x.x 2001:0db8:x",
-		},
-	}
-
-	for _, tt := range tests {
-		gotBuf := redactIPs([]byte(tt.in))
-		if string(gotBuf) != tt.want {
-			t.Errorf("for %q,\n got: %#q\nwant: %#q\n", tt.in, gotBuf, tt.want)
-		}
-	}
-}
 
 func TestAppendMetadata(t *testing.T) {
 	var l Logger
diff --git a/tailcfg/tailcfg.go b/tailcfg/tailcfg.go
index ada0df8fc..11a0d0830 100644
--- a/tailcfg/tailcfg.go
+++ b/tailcfg/tailcfg.go
@@ -160,7 +160,8 @@ type CapabilityVersion int
 //   - 113: 2025-01-20: Client communicates to control whether funnel is enabled by sending Hostinfo.IngressEnabled (#14688)
 //   - 114: 2025-01-30: NodeAttrMaxKeyDuration CapMap defined, clients might use it (no tailscaled code change) (#14829)
 //   - 115: 2025-03-07: Client understands DERPRegion.NoMeasureNoHome.
-const CurrentCapabilityVersion CapabilityVersion = 115
+//   - 116: 2025-05-05: Client serves MagicDNS "AAAA" if NodeAttrMagicDNSPeerAAAA set on self node
+const CurrentCapabilityVersion CapabilityVersion = 116
 
 // ID is an integer ID for a user, node, or login allocated by the
 // control plane.
@@ -875,10 +876,37 @@ type Hostinfo struct {
 	// explicitly declared by a node.
 	Location *Location `json:",omitempty"`
 
+	TPM *TPMInfo `json:",omitempty"` // TPM device metadata, if available
+
 	// NOTE: any new fields containing pointers in this type
 	//       require changes to Hostinfo.Equal.
 }
 
+// TPMInfo contains information about a TPM 2.0 device present on a node.
+// All fields are read from TPM_CAP_TPM_PROPERTIES, see Part 2, section 6.13 of
+// https://trustedcomputinggroup.org/resource/tpm-library-specification/.
+type TPMInfo struct {
+	// Manufacturer is a 4-letter code from section 4.1 of
+	// https://trustedcomputinggroup.org/resource/vendor-id-registry/,
+	// for example "MSFT" for Microsoft.
+	// Read from TPM_PT_MANUFACTURER.
+	Manufacturer string `json:",omitempty"`
+	// Vendor is a vendor ID string, up to 16 characters.
+	// Read from TPM_PT_VENDOR_STRING_*.
+	Vendor string `json:",omitempty"`
+	// Model is a vendor-defined TPM model.
+	// Read from TPM_PT_VENDOR_TPM_TYPE.
+	Model int `json:",omitempty"`
+	// FirmwareVersion is the version number of the firmware.
+	// Read from TPM_PT_FIRMWARE_VERSION_*.
+	FirmwareVersion uint64 `json:",omitempty"`
+	// SpecRevision is the TPM 2.0 spec revision encoded as a single number. All
+	// revisions can be found at
+	// https://trustedcomputinggroup.org/resource/tpm-library-specification/.
+	// Before revision 184, TCG used the "01.83" format for revision 183.
+	SpecRevision int `json:",omitempty"`
+}
+
 // ServiceName is the name of a service, of the form `svc:dns-label`. Services
 // represent some kind of application provided for users of the tailnet with a
 // MagicDNS name and possibly dedicated IP addresses. Currently (2024-01-21),
@@ -2466,6 +2494,10 @@ const (
 	// NodeAttrRelayClient permits the node to act as an underlay UDP relay
 	// client. There are no expected values for this key in NodeCapMap.
 	NodeAttrRelayClient NodeCapability = "relay:client"
+
+	// NodeAttrMagicDNSPeerAAAA is a capability that tells the node's MagicDNS
+	// server to answer AAAA queries about its peers. See tailscale/tailscale#1152.
+	NodeAttrMagicDNSPeerAAAA NodeCapability = "magicdns-aaaa"
 )
 
 // SetDNSRequest is a request to add a DNS record.
diff --git a/tailcfg/tailcfg_clone.go b/tailcfg/tailcfg_clone.go
index 3952f5f47..2c7941d51 100644
--- a/tailcfg/tailcfg_clone.go
+++ b/tailcfg/tailcfg_clone.go
@@ -141,6 +141,9 @@ func (src *Hostinfo) Clone() *Hostinfo {
 	if dst.Location != nil {
 		dst.Location = ptr.To(*src.Location)
 	}
+	if dst.TPM != nil {
+		dst.TPM = ptr.To(*src.TPM)
+	}
 	return dst
 }
 
@@ -184,6 +187,7 @@ var _HostinfoCloneNeedsRegeneration = Hostinfo(struct {
 	AppConnector    opt.Bool
 	ServicesHash    string
 	Location        *Location
+	TPM             *TPMInfo
 }{})
 
 // Clone makes a deep copy of NetInfo.
diff --git a/tailcfg/tailcfg_test.go b/tailcfg/tailcfg_test.go
index dd81af5d6..079162a15 100644
--- a/tailcfg/tailcfg_test.go
+++ b/tailcfg/tailcfg_test.go
@@ -68,6 +68,7 @@ func TestHostinfoEqual(t *testing.T) {
 		"AppConnector",
 		"ServicesHash",
 		"Location",
+		"TPM",
 	}
 	if have := fieldsOf(reflect.TypeFor[Hostinfo]()); !reflect.DeepEqual(have, hiHandles) {
 		t.Errorf("Hostinfo.Equal check might be out of sync\nfields: %q\nhandled: %q\n",
diff --git a/tailcfg/tailcfg_view.go b/tailcfg/tailcfg_view.go
index f8f9f865c..c76654887 100644
--- a/tailcfg/tailcfg_view.go
+++ b/tailcfg/tailcfg_view.go
@@ -301,7 +301,9 @@ func (v HostinfoView) UserspaceRouter() opt.Bool              { return v.ж.User
 func (v HostinfoView) AppConnector() opt.Bool                 { return v.ж.AppConnector }
 func (v HostinfoView) ServicesHash() string                   { return v.ж.ServicesHash }
 func (v HostinfoView) Location() LocationView                 { return v.ж.Location.View() }
-func (v HostinfoView) Equal(v2 HostinfoView) bool             { return v.ж.Equal(v2.ж) }
+func (v HostinfoView) TPM() views.ValuePointer[TPMInfo]       { return views.ValuePointerOf(v.ж.TPM) }
+
+func (v HostinfoView) Equal(v2 HostinfoView) bool { return v.ж.Equal(v2.ж) }
 
 // A compilation failure here means this code must be regenerated, with the command at the top of this file.
 var _HostinfoViewNeedsRegeneration = Hostinfo(struct {
@@ -343,6 +345,7 @@ var _HostinfoViewNeedsRegeneration = Hostinfo(struct {
 	AppConnector    opt.Bool
 	ServicesHash    string
 	Location        *Location
+	TPM             *TPMInfo
 }{})
 
 // View returns a read-only view of NetInfo.
diff --git a/tool/gocross/gocross.go b/tool/gocross/gocross.go
index 8011c1095..d14ea0388 100644
--- a/tool/gocross/gocross.go
+++ b/tool/gocross/gocross.go
@@ -15,9 +15,9 @@ import (
 	"fmt"
 	"os"
 	"path/filepath"
+	"runtime/debug"
 
 	"tailscale.com/atomicfile"
-	"tailscale.com/version"
 )
 
 func main() {
@@ -28,8 +28,19 @@ func main() {
 		// any time.
 		switch os.Args[1] {
 		case "gocross-version":
-			fmt.Println(version.GetMeta().GitCommit)
-			os.Exit(0)
+			bi, ok := debug.ReadBuildInfo()
+			if !ok {
+				fmt.Fprintln(os.Stderr, "failed getting build info")
+				os.Exit(1)
+			}
+			for _, s := range bi.Settings {
+				if s.Key == "vcs.revision" {
+					fmt.Println(s.Value)
+					os.Exit(0)
+				}
+			}
+			fmt.Fprintln(os.Stderr, "did not find vcs.revision in build info")
+			os.Exit(1)
 		case "is-gocross":
 			// This subcommand exits with an error code when called on a
 			// regular go binary, so it can be used to detect when `go` is
@@ -85,9 +96,9 @@ func main() {
 		path := filepath.Join(toolchain, "bin") + string(os.PathListSeparator) + os.Getenv("PATH")
 		env.Set("PATH", path)
 
-		debug("Input: %s\n", formatArgv(os.Args))
-		debug("Command: %s\n", formatArgv(newArgv))
-		debug("Set the following flags/envvars:\n%s\n", env.Diff())
+		debugf("Input: %s\n", formatArgv(os.Args))
+		debugf("Command: %s\n", formatArgv(newArgv))
+		debugf("Set the following flags/envvars:\n%s\n", env.Diff())
 
 		args = newArgv
 		if err := env.Apply(); err != nil {
@@ -103,7 +114,7 @@ func main() {
 //go:embed gocross-wrapper.sh
 var wrapperScript []byte
 
-func debug(format string, args ...any) {
+func debugf(format string, args ...any) {
 	debug := os.Getenv("GOCROSS_DEBUG")
 	var (
 		out *os.File
diff --git a/tool/gocross/gocross_test.go b/tool/gocross/gocross_test.go
new file mode 100644
index 000000000..82afd268c
--- /dev/null
+++ b/tool/gocross/gocross_test.go
@@ -0,0 +1,19 @@
+// Copyright (c) Tailscale Inc & AUTHORS
+// SPDX-License-Identifier: BSD-3-Clause
+
+package main
+
+import (
+	"testing"
+
+	"tailscale.com/tstest/deptest"
+)
+
+func TestDeps(t *testing.T) {
+	deptest.DepChecker{
+		BadDeps: map[string]string{
+			"tailscale.com/tailcfg": "circular dependency via go generate",
+			"tailscale.com/version": "circular dependency via go generate",
+		},
+	}.Check(t)
+}
diff --git a/tsnet/tsnet.go b/tsnet/tsnet.go
index f97598075..1880b62b1 100644
--- a/tsnet/tsnet.go
+++ b/tsnet/tsnet.go
@@ -1101,13 +1101,33 @@ type FunnelOption interface {
 	funnelOption()
 }
 
-type funnelOnly int
+type funnelOnly struct{}
 
 func (funnelOnly) funnelOption() {}
 
 // FunnelOnly configures the listener to only respond to connections from Tailscale Funnel.
 // The local tailnet will not be able to connect to the listener.
-func FunnelOnly() FunnelOption { return funnelOnly(1) }
+func FunnelOnly() FunnelOption { return funnelOnly{} }
+
+type funnelTLSConfig struct{ conf *tls.Config }
+
+func (f funnelTLSConfig) funnelOption() {}
+
+// FunnelTLSConfig configures the TLS configuration for [Server.ListenFunnel]
+//
+// This is rarely needed but can permit requiring client certificates, specific
+// ciphers suites, etc.
+//
+// The provided conf should at least be able to get a certificate, setting
+// GetCertificate, Certificates or GetConfigForClient appropriately.
+// The most common configuration is to set GetCertificate to
+// Server.LocalClient's GetCertificate method.
+//
+// Unless [FunnelOnly] is also used, the configuration is also used for
+// in-tailnet connections that don't arrive over Funnel.
+func FunnelTLSConfig(conf *tls.Config) FunnelOption {
+	return funnelTLSConfig{conf: conf}
+}
 
 // ListenFunnel announces on the public internet using Tailscale Funnel.
 //
@@ -1140,6 +1160,26 @@ func (s *Server) ListenFunnel(network, addr string, opts ...FunnelOption) (net.L
 		return nil, err
 	}
 
+	// Process, validate opts.
+	lnOn := listenOnBoth
+	var tlsConfig *tls.Config
+	for _, opt := range opts {
+		switch v := opt.(type) {
+		case funnelTLSConfig:
+			if v.conf == nil {
+				return nil, errors.New("invalid nil FunnelTLSConfig")
+			}
+			tlsConfig = v.conf
+		case funnelOnly:
+			lnOn = listenOnFunnel
+		default:
+			return nil, fmt.Errorf("unknown opts FunnelOption type %T", v)
+		}
+	}
+	if tlsConfig == nil {
+		tlsConfig = &tls.Config{GetCertificate: s.getCert}
+	}
+
 	ctx := context.Background()
 	st, err := s.Up(ctx)
 	if err != nil {
@@ -1177,19 +1217,11 @@ func (s *Server) ListenFunnel(network, addr string, opts ...FunnelOption) (net.L
 	}
 
 	// Start a funnel listener.
-	lnOn := listenOnBoth
-	for _, opt := range opts {
-		if _, ok := opt.(funnelOnly); ok {
-			lnOn = listenOnFunnel
-		}
-	}
 	ln, err := s.listen(network, addr, lnOn)
 	if err != nil {
 		return nil, err
 	}
-	return tls.NewListener(ln, &tls.Config{
-		GetCertificate: s.getCert,
-	}), nil
+	return tls.NewListener(ln, tlsConfig), nil
 }
 
 type listenOn string
diff --git a/tstest/integration/integration.go b/tstest/integration/integration.go
index 9df536971..d64bfbbd9 100644
--- a/tstest/integration/integration.go
+++ b/tstest/integration/integration.go
@@ -33,6 +33,7 @@ import (
 	"time"
 
 	"go4.org/mem"
+	"tailscale.com/client/local"
 	"tailscale.com/derp"
 	"tailscale.com/derp/derphttp"
 	"tailscale.com/ipn"
@@ -64,61 +65,151 @@ var (
 // as a last ditch place to report errors.
 var MainError syncs.AtomicValue[error]
 
-// CleanupBinaries cleans up any resources created by calls to BinaryDir, TailscaleBinary, or TailscaledBinary.
-// It should be called from TestMain after all tests have completed.
-func CleanupBinaries() {
-	buildOnce.Do(func() {})
-	if binDir != "" {
-		os.RemoveAll(binDir)
+// Binaries contains the paths to the tailscale and tailscaled binaries.
+type Binaries struct {
+	Dir        string
+	Tailscale  BinaryInfo
+	Tailscaled BinaryInfo
+}
+
+// BinaryInfo describes a tailscale or tailscaled binary.
+type BinaryInfo struct {
+	Path string // abs path to tailscale or tailscaled binary
+	Size int64
+
+	// FD and FDmu are set on Unix to efficiently copy the binary to a new
+	// test's automatically-cleaned-up temp directory.
+	FD   *os.File // for Unix (macOS, Linux, ...)
+	FDMu sync.Locker
+
+	// Contents is used on Windows instead of FD to copy the binary between
+	// test directories. (On Windows you can't keep an FD open while an earlier
+	// test's temp directories are deleted.)
+	// This burns some memory and costs more in I/O, but oh well.
+	Contents []byte
+}
+
+func (b BinaryInfo) CopyTo(dir string) (BinaryInfo, error) {
+	ret := b
+	ret.Path = filepath.Join(dir, path.Base(b.Path))
+
+	switch runtime.GOOS {
+	case "linux":
+		// TODO(bradfitz): be fancy and use linkat with AT_EMPTY_PATH to avoid
+		// copying? I couldn't get it to work, though.
+		// For now, just do the same thing as every other Unix and copy
+		// the binary.
+		fallthrough
+	case "darwin", "freebsd", "openbsd", "netbsd":
+		f, err := os.OpenFile(ret.Path, os.O_RDWR|os.O_CREATE|os.O_EXCL, 0o755)
+		if err != nil {
+			return BinaryInfo{}, err
+		}
+		b.FDMu.Lock()
+		b.FD.Seek(0, 0)
+		size, err := io.Copy(f, b.FD)
+		b.FDMu.Unlock()
+		if err != nil {
+			f.Close()
+			return BinaryInfo{}, fmt.Errorf("copying %q: %w", b.Path, err)
+		}
+		if size != b.Size {
+			f.Close()
+			return BinaryInfo{}, fmt.Errorf("copy %q: size mismatch: %d != %d", b.Path, size, b.Size)
+		}
+		if err := f.Close(); err != nil {
+			return BinaryInfo{}, err
+		}
+		return ret, nil
+	case "windows":
+		return ret, os.WriteFile(ret.Path, b.Contents, 0o755)
+	default:
+		return BinaryInfo{}, fmt.Errorf("unsupported OS %q", runtime.GOOS)
 	}
 }
 
-// BinaryDir returns a directory containing test tailscale and tailscaled binaries.
-// If any test calls BinaryDir, there must be a TestMain function that calls
-// CleanupBinaries after all tests are complete.
-func BinaryDir(tb testing.TB) string {
+// GetBinaries create a temp directory using tb and builds (or copies previously
+// built) cmd/tailscale and cmd/tailscaled binaries into that directory.
+//
+// It fails tb if the build or binary copies fail.
+func GetBinaries(tb testing.TB) *Binaries {
+	dir := tb.TempDir()
 	buildOnce.Do(func() {
-		binDir, buildErr = buildTestBinaries()
+		buildErr = buildTestBinaries(dir)
 	})
 	if buildErr != nil {
 		tb.Fatal(buildErr)
 	}
-	return binDir
-}
-
-// TailscaleBinary returns the path to the test tailscale binary.
-// If any test calls TailscaleBinary, there must be a TestMain function that calls
-// CleanupBinaries after all tests are complete.
-func TailscaleBinary(tb testing.TB) string {
-	return filepath.Join(BinaryDir(tb), "tailscale"+exe())
-}
-
-// TailscaledBinary returns the path to the test tailscaled binary.
-// If any test calls TailscaleBinary, there must be a TestMain function that calls
-// CleanupBinaries after all tests are complete.
-func TailscaledBinary(tb testing.TB) string {
-	return filepath.Join(BinaryDir(tb), "tailscaled"+exe())
+	if binariesCache.Dir == dir {
+		return binariesCache
+	}
+	ts, err := binariesCache.Tailscale.CopyTo(dir)
+	if err != nil {
+		tb.Fatalf("copying tailscale binary: %v", err)
+	}
+	tsd, err := binariesCache.Tailscaled.CopyTo(dir)
+	if err != nil {
+		tb.Fatalf("copying tailscaled binary: %v", err)
+	}
+	return &Binaries{
+		Dir:        dir,
+		Tailscale:  ts,
+		Tailscaled: tsd,
+	}
 }
 
 var (
-	buildOnce sync.Once
-	buildErr  error
-	binDir    string
+	buildOnce     sync.Once
+	buildErr      error
+	binariesCache *Binaries
 )
 
 // buildTestBinaries builds tailscale and tailscaled.
-// It returns the dir containing the binaries.
-func buildTestBinaries() (string, error) {
-	bindir, err := os.MkdirTemp("", "")
-	if err != nil {
-		return "", err
+// On success, it initializes [binariesCache].
+func buildTestBinaries(dir string) error {
+	getBinaryInfo := func(name string) (BinaryInfo, error) {
+		bi := BinaryInfo{Path: filepath.Join(dir, name+exe())}
+		fi, err := os.Stat(bi.Path)
+		if err != nil {
+			return BinaryInfo{}, fmt.Errorf("stat %q: %v", bi.Path, err)
+		}
+		bi.Size = fi.Size()
+
+		switch runtime.GOOS {
+		case "windows":
+			bi.Contents, err = os.ReadFile(bi.Path)
+			if err != nil {
+				return BinaryInfo{}, fmt.Errorf("read %q: %v", bi.Path, err)
+			}
+		default:
+			bi.FD, err = os.OpenFile(bi.Path, os.O_RDONLY, 0)
+			if err != nil {
+				return BinaryInfo{}, fmt.Errorf("open %q: %v", bi.Path, err)
+			}
+			bi.FDMu = new(sync.Mutex)
+			// Note: bi.FD is copied around between tests but never closed, by
+			// design. It will be closed when the process exits, and that will
+			// close the inode that we're copying the bytes from for each test.
+		}
+		return bi, nil
 	}
-	err = build(bindir, "tailscale.com/cmd/tailscaled", "tailscale.com/cmd/tailscale")
+	err := build(dir, "tailscale.com/cmd/tailscaled", "tailscale.com/cmd/tailscale")
 	if err != nil {
-		os.RemoveAll(bindir)
-		return "", err
+		return err
 	}
-	return bindir, nil
+	b := &Binaries{
+		Dir: dir,
+	}
+	b.Tailscale, err = getBinaryInfo("tailscale")
+	if err != nil {
+		return err
+	}
+	b.Tailscaled, err = getBinaryInfo("tailscaled")
+	if err != nil {
+		return err
+	}
+	binariesCache = b
+	return nil
 }
 
 func build(outDir string, targets ...string) error {
@@ -436,14 +527,16 @@ func NewTestEnv(t testing.TB, opts ...TestEnvOpt) *TestEnv {
 	derpMap := RunDERPAndSTUN(t, logger.Discard, "127.0.0.1")
 	logc := new(LogCatcher)
 	control := &testcontrol.Server{
+		Logf:    logger.WithPrefix(t.Logf, "testcontrol: "),
 		DERPMap: derpMap,
 	}
 	control.HTTPTestServer = httptest.NewUnstartedServer(control)
 	trafficTrap := new(trafficTrap)
+	binaries := GetBinaries(t)
 	e := &TestEnv{
 		t:                 t,
-		cli:               TailscaleBinary(t),
-		daemon:            TailscaledBinary(t),
+		cli:               binaries.Tailscale.Path,
+		daemon:            binaries.Tailscaled.Path,
 		LogCatcher:        logc,
 		LogCatcherServer:  httptest.NewServer(logc),
 		Control:           control,
@@ -484,6 +577,7 @@ type TestNode struct {
 
 	mu        sync.Mutex
 	onLogLine []func([]byte)
+	lc        *local.Client
 }
 
 // NewTestNode allocates a temp directory for a new test node.
@@ -500,14 +594,18 @@ func NewTestNode(t *testing.T, env *TestEnv) *TestNode {
 		env:       env,
 		dir:       dir,
 		sockFile:  sockFile,
-		stateFile: filepath.Join(dir, "tailscale.state"),
+		stateFile: filepath.Join(dir, "tailscaled.state"), // matches what cmd/tailscaled uses
 	}
 
-	// Look for a data race. Once we see the start marker, start logging the rest.
+	// Look for a data race or panic.
+	// Once we see the start marker, start logging the rest.
 	var sawRace bool
 	var sawPanic bool
 	n.addLogLineHook(func(line []byte) {
 		lineB := mem.B(line)
+		if mem.Contains(lineB, mem.S("DEBUG-ADDR=")) {
+			t.Log(strings.TrimSpace(string(line)))
+		}
 		if mem.Contains(lineB, mem.S("WARNING: DATA RACE")) {
 			sawRace = true
 		}
@@ -522,6 +620,20 @@ func NewTestNode(t *testing.T, env *TestEnv) *TestNode {
 	return n
 }
 
+func (n *TestNode) LocalClient() *local.Client {
+	n.mu.Lock()
+	defer n.mu.Unlock()
+	if n.lc == nil {
+		tr := &http.Transport{}
+		n.lc = &local.Client{
+			Socket:        n.sockFile,
+			UseSocketOnly: true,
+		}
+		n.env.t.Cleanup(tr.CloseIdleConnections)
+	}
+	return n.lc
+}
+
 func (n *TestNode) diskPrefs() *ipn.Prefs {
 	t := n.env.t
 	t.Helper()
@@ -658,6 +770,27 @@ func (d *Daemon) MustCleanShutdown(t testing.TB) {
 	}
 }
 
+// awaitTailscaledRunnable tries to run `tailscaled --version` until it
+// works. This is an unsatisfying workaround for ETXTBSY we were seeing
+// on GitHub Actions that aren't understood. It's not clear what's holding
+// a writable fd to tailscaled after `go install` completes.
+// See https://github.com/tailscale/tailscale/issues/15868.
+func (n *TestNode) awaitTailscaledRunnable() error {
+	t := n.env.t
+	t.Helper()
+	if err := tstest.WaitFor(10*time.Second, func() error {
+		out, err := exec.Command(n.env.daemon, "--version").CombinedOutput()
+		if err == nil {
+			return nil
+		}
+		t.Logf("error running tailscaled --version: %v, %s", err, out)
+		return err
+	}); err != nil {
+		return fmt.Errorf("gave up trying to run tailscaled: %v", err)
+	}
+	return nil
+}
+
 // StartDaemon starts the node's tailscaled, failing if it fails to start.
 // StartDaemon ensures that the process will exit when the test completes.
 func (n *TestNode) StartDaemon() *Daemon {
@@ -666,11 +799,17 @@ func (n *TestNode) StartDaemon() *Daemon {
 
 func (n *TestNode) StartDaemonAsIPNGOOS(ipnGOOS string) *Daemon {
 	t := n.env.t
+
+	if err := n.awaitTailscaledRunnable(); err != nil {
+		t.Fatalf("awaitTailscaledRunnable: %v", err)
+	}
+
 	cmd := exec.Command(n.env.daemon)
 	cmd.Args = append(cmd.Args,
-		"--state="+n.stateFile,
+		"--statedir="+n.dir,
 		"--socket="+n.sockFile,
 		"--socks5-server=localhost:0",
+		"--debug=localhost:0",
 	)
 	if *verboseTailscaled {
 		cmd.Args = append(cmd.Args, "-verbose=2")
@@ -684,7 +823,6 @@ func (n *TestNode) StartDaemonAsIPNGOOS(ipnGOOS string) *Daemon {
 		cmd.Args = append(cmd.Args, "--config="+n.configFile)
 	}
 	cmd.Env = append(os.Environ(),
-		"TS_CONTROL_IS_PLAINTEXT_HTTP=1",
 		"TS_DEBUG_PERMIT_HTTP_C2N=1",
 		"TS_LOG_TARGET="+n.env.LogCatcherServer.URL,
 		"HTTP_PROXY="+n.env.TrafficTrapServer.URL,
diff --git a/tstest/integration/integration_test.go b/tstest/integration/integration_test.go
index 0da2e6086..90cc7e443 100644
--- a/tstest/integration/integration_test.go
+++ b/tstest/integration/integration_test.go
@@ -49,7 +49,6 @@ func TestMain(m *testing.M) {
 	os.Setenv("TS_DISABLE_UPNP", "true")
 	flag.Parse()
 	v := m.Run()
-	CleanupBinaries()
 	if v != 0 {
 		os.Exit(v)
 	}
@@ -278,15 +277,20 @@ func TestOneNodeUpAuth(t *testing.T) {
 	t.Logf("Running up --login-server=%s ...", env.ControlURL())
 
 	cmd := n1.Tailscale("up", "--login-server="+env.ControlURL())
-	var authCountAtomic int32
+	var authCountAtomic atomic.Int32
 	cmd.Stdout = &authURLParserWriter{fn: func(urlStr string) error {
+		t.Logf("saw auth URL %q", urlStr)
 		if env.Control.CompleteAuth(urlStr) {
-			atomic.AddInt32(&authCountAtomic, 1)
+			if authCountAtomic.Add(1) > 1 {
+				err := errors.New("completed multple auth URLs")
+				t.Error(err)
+				return err
+			}
 			t.Logf("completed auth path %s", urlStr)
 			return nil
 		}
 		err := fmt.Errorf("Failed to complete auth path to %q", urlStr)
-		t.Log(err)
+		t.Error(err)
 		return err
 	}}
 	cmd.Stderr = cmd.Stdout
@@ -297,7 +301,7 @@ func TestOneNodeUpAuth(t *testing.T) {
 
 	n1.AwaitRunning()
 
-	if n := atomic.LoadInt32(&authCountAtomic); n != 1 {
+	if n := authCountAtomic.Load(); n != 1 {
 		t.Errorf("Auth URLs completed = %d; want 1", n)
 	}
 
diff --git a/tstest/integration/testcontrol/testcontrol.go b/tstest/integration/testcontrol/testcontrol.go
index 52b96fe4d..71205f897 100644
--- a/tstest/integration/testcontrol/testcontrol.go
+++ b/tstest/integration/testcontrol/testcontrol.go
@@ -55,6 +55,10 @@ type Server struct {
 	MagicDNSDomain string
 	HandleC2N      http.Handler // if non-nil, used for /some-c2n-path/ in tests
 
+	// AllNodesSameUser, if true, makes all created nodes
+	// belong to the same user.
+	AllNodesSameUser bool
+
 	// ExplicitBaseURL or HTTPTestServer must be set.
 	ExplicitBaseURL string           // e.g. "http://127.0.0.1:1234" with no trailing URL
 	HTTPTestServer  *httptest.Server // if non-nil, used to get BaseURL
@@ -96,9 +100,9 @@ type Server struct {
 	logins        map[key.NodePublic]*tailcfg.Login
 	updates       map[tailcfg.NodeID]chan updateType
 	authPath      map[string]*AuthPath
-	nodeKeyAuthed map[key.NodePublic]bool // key => true once authenticated
-	msgToSend     map[key.NodePublic]any  // value is *tailcfg.PingRequest or entire *tailcfg.MapResponse
-	allExpired    bool                    // All nodes will be told their node key is expired.
+	nodeKeyAuthed set.Set[key.NodePublic]
+	msgToSend     map[key.NodePublic]any // value is *tailcfg.PingRequest or entire *tailcfg.MapResponse
+	allExpired    bool                   // All nodes will be told their node key is expired.
 }
 
 // BaseURL returns the server's base URL, without trailing slash.
@@ -522,6 +526,10 @@ func (s *Server) getUser(nodeKey key.NodePublic) (*tailcfg.User, *tailcfg.Login)
 		return u, s.logins[nodeKey]
 	}
 	id := tailcfg.UserID(len(s.users) + 1)
+	if s.AllNodesSameUser {
+		id = 123
+	}
+	s.logf("Created user %v for node %s", id, nodeKey)
 	loginName := fmt.Sprintf("user-%d@%s", id, domain)
 	displayName := fmt.Sprintf("User %d", id)
 	login := &tailcfg.Login{
@@ -582,10 +590,8 @@ func (s *Server) CompleteAuth(authPathOrURL string) bool {
 	if ap.nodeKey.IsZero() {
 		panic("zero AuthPath.NodeKey")
 	}
-	if s.nodeKeyAuthed == nil {
-		s.nodeKeyAuthed = map[key.NodePublic]bool{}
-	}
-	s.nodeKeyAuthed[ap.nodeKey] = true
+	s.nodeKeyAuthed.Make()
+	s.nodeKeyAuthed.Add(ap.nodeKey)
 	ap.CompleteSuccessfully()
 	return true
 }
@@ -645,36 +651,40 @@ func (s *Server) serveRegister(w http.ResponseWriter, r *http.Request, mkey key.
 	if s.nodes == nil {
 		s.nodes = map[key.NodePublic]*tailcfg.Node{}
 	}
-
+	_, ok := s.nodes[nk]
 	machineAuthorized := true // TODO: add Server.RequireMachineAuth
+	if !ok {
 
-	v4Prefix := netip.PrefixFrom(netaddr.IPv4(100, 64, uint8(tailcfg.NodeID(user.ID)>>8), uint8(tailcfg.NodeID(user.ID))), 32)
-	v6Prefix := netip.PrefixFrom(tsaddr.Tailscale4To6(v4Prefix.Addr()), 128)
+		nodeID := len(s.nodes) + 1
+		v4Prefix := netip.PrefixFrom(netaddr.IPv4(100, 64, uint8(nodeID>>8), uint8(nodeID)), 32)
+		v6Prefix := netip.PrefixFrom(tsaddr.Tailscale4To6(v4Prefix.Addr()), 128)
 
-	allowedIPs := []netip.Prefix{
-		v4Prefix,
-		v6Prefix,
-	}
-
-	s.nodes[nk] = &tailcfg.Node{
-		ID:                tailcfg.NodeID(user.ID),
-		StableID:          tailcfg.StableNodeID(fmt.Sprintf("TESTCTRL%08x", int(user.ID))),
-		User:              user.ID,
-		Machine:           mkey,
-		Key:               req.NodeKey,
-		MachineAuthorized: machineAuthorized,
-		Addresses:         allowedIPs,
-		AllowedIPs:        allowedIPs,
-		Hostinfo:          req.Hostinfo.View(),
-		Name:              req.Hostinfo.Hostname,
-		Capabilities: []tailcfg.NodeCapability{
-			tailcfg.CapabilityHTTPS,
-			tailcfg.NodeAttrFunnel,
-			tailcfg.CapabilityFunnelPorts + "?ports=8080,443",
-		},
+		allowedIPs := []netip.Prefix{
+			v4Prefix,
+			v6Prefix,
+		}
+		node := &tailcfg.Node{
+			ID:                tailcfg.NodeID(nodeID),
+			StableID:          tailcfg.StableNodeID(fmt.Sprintf("TESTCTRL%08x", int(nodeID))),
+			User:              user.ID,
+			Machine:           mkey,
+			Key:               req.NodeKey,
+			MachineAuthorized: machineAuthorized,
+			Addresses:         allowedIPs,
+			AllowedIPs:        allowedIPs,
+			Hostinfo:          req.Hostinfo.View(),
+			Name:              req.Hostinfo.Hostname,
+			Capabilities: []tailcfg.NodeCapability{
+				tailcfg.CapabilityHTTPS,
+				tailcfg.NodeAttrFunnel,
+				tailcfg.CapabilityFileSharing,
+				tailcfg.CapabilityFunnelPorts + "?ports=8080,443",
+			},
+		}
+		s.nodes[nk] = node
 	}
 	requireAuth := s.RequireAuth
-	if requireAuth && s.nodeKeyAuthed[nk] {
+	if requireAuth && s.nodeKeyAuthed.Contains(nk) {
 		requireAuth = false
 	}
 	allExpired := s.allExpired
@@ -951,7 +961,6 @@ func (s *Server) MapResponse(req *tailcfg.MapRequest) (res *tailcfg.MapResponse,
 	node.CapMap = nodeCapMap
 	node.Capabilities = append(node.Capabilities, tailcfg.NodeAttrDisableUPnP)
 
-	user, _ := s.getUser(nk)
 	t := time.Date(2020, 8, 3, 0, 0, 0, 1, time.UTC)
 	dns := s.DNSConfig
 	if dns != nil && s.MagicDNSDomain != "" {
@@ -1013,7 +1022,7 @@ func (s *Server) MapResponse(req *tailcfg.MapRequest) (res *tailcfg.MapResponse,
 	})
 	res.UserProfiles = s.allUserProfiles()
 
-	v4Prefix := netip.PrefixFrom(netaddr.IPv4(100, 64, uint8(tailcfg.NodeID(user.ID)>>8), uint8(tailcfg.NodeID(user.ID))), 32)
+	v4Prefix := netip.PrefixFrom(netaddr.IPv4(100, 64, uint8(node.ID>>8), uint8(node.ID)), 32)
 	v6Prefix := netip.PrefixFrom(tsaddr.Tailscale4To6(v4Prefix.Addr()), 128)
 
 	res.Node.Addresses = []netip.Prefix{
diff --git a/tstest/integration/vms/harness_test.go b/tstest/integration/vms/harness_test.go
index 1e080414d..256227d6c 100644
--- a/tstest/integration/vms/harness_test.go
+++ b/tstest/integration/vms/harness_test.go
@@ -134,11 +134,12 @@ func newHarness(t *testing.T) *Harness {
 	loginServer := fmt.Sprintf("http://%s", ln.Addr())
 	t.Logf("loginServer: %s", loginServer)
 
+	binaries := integration.GetBinaries(t)
 	h := &Harness{
 		pubKey:         string(pubkey),
-		binaryDir:      integration.BinaryDir(t),
-		cli:            integration.TailscaleBinary(t),
-		daemon:         integration.TailscaledBinary(t),
+		binaryDir:      binaries.Dir,
+		cli:            binaries.Tailscale.Path,
+		daemon:         binaries.Tailscaled.Path,
 		signer:         signer,
 		loginServerURL: loginServer,
 		cs:             cs,
diff --git a/tstest/integration/vms/vms_test.go b/tstest/integration/vms/vms_test.go
index 6d73a3f78..f71f2bdbf 100644
--- a/tstest/integration/vms/vms_test.go
+++ b/tstest/integration/vms/vms_test.go
@@ -28,7 +28,6 @@ import (
 	"golang.org/x/crypto/ssh"
 	"golang.org/x/sync/semaphore"
 	"tailscale.com/tstest"
-	"tailscale.com/tstest/integration"
 	"tailscale.com/types/logger"
 )
 
@@ -51,13 +50,6 @@ var (
 	}()
 )
 
-func TestMain(m *testing.M) {
-	flag.Parse()
-	v := m.Run()
-	integration.CleanupBinaries()
-	os.Exit(v)
-}
-
 func TestDownloadImages(t *testing.T) {
 	if !*runVMTests {
 		t.Skip("not running integration tests (need --run-vm-tests)")
diff --git a/util/linuxfw/nftables_runner.go b/util/linuxfw/nftables_runner.go
index 0f411521b..b87298c61 100644
--- a/util/linuxfw/nftables_runner.go
+++ b/util/linuxfw/nftables_runner.go
@@ -1710,55 +1710,43 @@ func (n *nftablesRunner) AddSNATRule() error {
 	return nil
 }
 
+func delMatchSubnetRouteMarkMasqRule(conn *nftables.Conn, table *nftables.Table, chain *nftables.Chain) error {
+
+	rule, err := createMatchSubnetRouteMarkRule(table, chain, Masq)
+	if err != nil {
+		return fmt.Errorf("create match subnet route mark rule: %w", err)
+	}
+
+	SNATRule, err := findRule(conn, rule)
+	if err != nil {
+		return fmt.Errorf("find SNAT rule v4: %w", err)
+	}
+
+	if SNATRule != nil {
+		_ = conn.DelRule(SNATRule)
+	}
+
+	if err := conn.Flush(); err != nil {
+		return fmt.Errorf("flush del SNAT rule: %w", err)
+	}
+
+	return nil
+}
+
 // DelSNATRule removes the netfilter rule to SNAT traffic destined for
 // local subnets. An error is returned if the rule does not exist.
 func (n *nftablesRunner) DelSNATRule() error {
 	conn := n.conn
 
-	hexTSFwmarkMask := getTailscaleFwmarkMask()
-	hexTSSubnetRouteMark := getTailscaleSubnetRouteMark()
-
-	exprs := []expr.Any{
-		&expr.Meta{Key: expr.MetaKeyMARK, Register: 1},
-		&expr.Bitwise{
-			SourceRegister: 1,
-			DestRegister:   1,
-			Len:            4,
-			Mask:           hexTSFwmarkMask,
-		},
-		&expr.Cmp{
-			Op:       expr.CmpOpEq,
-			Register: 1,
-			Data:     hexTSSubnetRouteMark,
-		},
-		&expr.Counter{},
-		&expr.Masq{},
-	}
-
 	for _, table := range n.getTables() {
 		chain, err := getChainFromTable(conn, table.Nat, chainNamePostrouting)
 		if err != nil {
-			return fmt.Errorf("get postrouting chain v4: %w", err)
+			return fmt.Errorf("get postrouting chain: %w", err)
 		}
-
-		rule := &nftables.Rule{
-			Table: table.Nat,
-			Chain: chain,
-			Exprs: exprs,
-		}
-
-		SNATRule, err := findRule(conn, rule)
+		err = delMatchSubnetRouteMarkMasqRule(conn, table.Nat, chain)
 		if err != nil {
-			return fmt.Errorf("find SNAT rule v4: %w", err)
+			return err
 		}
-
-		if SNATRule != nil {
-			_ = conn.DelRule(SNATRule)
-		}
-	}
-
-	if err := conn.Flush(); err != nil {
-		return fmt.Errorf("flush del SNAT rule: %w", err)
 	}
 
 	return nil
diff --git a/util/linuxfw/nftables_runner_test.go b/util/linuxfw/nftables_runner_test.go
index 712a7b939..6fb180ed6 100644
--- a/util/linuxfw/nftables_runner_test.go
+++ b/util/linuxfw/nftables_runner_test.go
@@ -12,6 +12,7 @@ import (
 	"net/netip"
 	"os"
 	"runtime"
+	"slices"
 	"strings"
 	"testing"
 
@@ -24,21 +25,21 @@ import (
 	"tailscale.com/types/logger"
 )
 
+func toAnySlice[T any](s []T) []any {
+	out := make([]any, len(s))
+	for i, v := range s {
+		out[i] = v
+	}
+	return out
+}
+
 // nfdump returns a hexdump of 4 bytes per line (like nft --debug=all), allowing
 // users to make sense of large byte literals more easily.
 func nfdump(b []byte) string {
 	var buf bytes.Buffer
-	i := 0
-	for ; i < len(b); i += 4 {
-		// TODO: show printable characters as ASCII
-		fmt.Fprintf(&buf, "%02x %02x %02x %02x\n",
-			b[i],
-			b[i+1],
-			b[i+2],
-			b[i+3])
-	}
-	for ; i < len(b); i++ {
-		fmt.Fprintf(&buf, "%02x ", b[i])
+	for c := range slices.Chunk(b, 4) {
+		format := strings.Repeat("%02x ", len(c))
+		fmt.Fprintf(&buf, format+"\n", toAnySlice(c)...)
 	}
 	return buf.String()
 }
@@ -75,7 +76,7 @@ func linediff(a, b string) string {
 	return buf.String()
 }
 
-func newTestConn(t *testing.T, want [][]byte) *nftables.Conn {
+func newTestConn(t *testing.T, want [][]byte, reply [][]netlink.Message) *nftables.Conn {
 	conn, err := nftables.New(nftables.WithTestDial(
 		func(req []netlink.Message) ([]netlink.Message, error) {
 			for idx, msg := range req {
@@ -96,7 +97,13 @@ func newTestConn(t *testing.T, want [][]byte) *nftables.Conn {
 				}
 				want = want[1:]
 			}
-			return req, nil
+			// no reply for batch end message
+			if len(want) == 0 {
+				return nil, nil
+			}
+			rep := reply[0]
+			reply = reply[1:]
+			return rep, nil
 		}))
 	if err != nil {
 		t.Fatal(err)
@@ -120,7 +127,7 @@ func TestInsertHookRule(t *testing.T) {
 		// batch end
 		[]byte("\x00\x00\x00\x0a"),
 	}
-	testConn := newTestConn(t, want)
+	testConn := newTestConn(t, want, nil)
 	table := testConn.AddTable(&nftables.Table{
 		Family: proto,
 		Name:   "ts-filter-test",
@@ -160,7 +167,7 @@ func TestInsertLoopbackRule(t *testing.T) {
 		// batch end
 		[]byte("\x00\x00\x00\x0a"),
 	}
-	testConn := newTestConn(t, want)
+	testConn := newTestConn(t, want, nil)
 	table := testConn.AddTable(&nftables.Table{
 		Family: proto,
 		Name:   "ts-filter-test",
@@ -196,7 +203,7 @@ func TestInsertLoopbackRuleV6(t *testing.T) {
 		// batch end
 		[]byte("\x00\x00\x00\x0a"),
 	}
-	testConn := newTestConn(t, want)
+	testConn := newTestConn(t, want, nil)
 	tableV6 := testConn.AddTable(&nftables.Table{
 		Family: protoV6,
 		Name:   "ts-filter-test",
@@ -232,7 +239,7 @@ func TestAddReturnChromeOSVMRangeRule(t *testing.T) {
 		// batch end
 		[]byte("\x00\x00\x00\x0a"),
 	}
-	testConn := newTestConn(t, want)
+	testConn := newTestConn(t, want, nil)
 	table := testConn.AddTable(&nftables.Table{
 		Family: proto,
 		Name:   "ts-filter-test",
@@ -264,7 +271,7 @@ func TestAddDropCGNATRangeRule(t *testing.T) {
 		// batch end
 		[]byte("\x00\x00\x00\x0a"),
 	}
-	testConn := newTestConn(t, want)
+	testConn := newTestConn(t, want, nil)
 	table := testConn.AddTable(&nftables.Table{
 		Family: proto,
 		Name:   "ts-filter-test",
@@ -296,7 +303,7 @@ func TestAddSetSubnetRouteMarkRule(t *testing.T) {
 		// batch end
 		[]byte("\x00\x00\x00\x0a"),
 	}
-	testConn := newTestConn(t, want)
+	testConn := newTestConn(t, want, nil)
 	table := testConn.AddTable(&nftables.Table{
 		Family: proto,
 		Name:   "ts-filter-test",
@@ -328,7 +335,7 @@ func TestAddDropOutgoingPacketFromCGNATRangeRuleWithTunname(t *testing.T) {
 		// batch end
 		[]byte("\x00\x00\x00\x0a"),
 	}
-	testConn := newTestConn(t, want)
+	testConn := newTestConn(t, want, nil)
 	table := testConn.AddTable(&nftables.Table{
 		Family: proto,
 		Name:   "ts-filter-test",
@@ -360,7 +367,7 @@ func TestAddAcceptOutgoingPacketRule(t *testing.T) {
 		// batch end
 		[]byte("\x00\x00\x00\x0a"),
 	}
-	testConn := newTestConn(t, want)
+	testConn := newTestConn(t, want, nil)
 	table := testConn.AddTable(&nftables.Table{
 		Family: proto,
 		Name:   "ts-filter-test",
@@ -392,7 +399,7 @@ func TestAddAcceptIncomingPacketRule(t *testing.T) {
 		// batch end
 		[]byte("\x00\x00\x00\x0a"),
 	}
-	testConn := newTestConn(t, want)
+	testConn := newTestConn(t, want, nil)
 	table := testConn.AddTable(&nftables.Table{
 		Family: proto,
 		Name:   "ts-filter-test",
@@ -420,11 +427,11 @@ func TestAddMatchSubnetRouteMarkRuleMasq(t *testing.T) {
 		// nft add chain ip ts-nat-test ts-postrouting-test { type nat hook postrouting priority 100; }
 		[]byte("\x02\x00\x00\x00\x10\x00\x01\x00\x74\x73\x2d\x6e\x61\x74\x2d\x74\x65\x73\x74\x00\x18\x00\x03\x00\x74\x73\x2d\x70\x6f\x73\x74\x72\x6f\x75\x74\x69\x6e\x67\x2d\x74\x65\x73\x74\x00\x14\x00\x04\x80\x08\x00\x01\x00\x00\x00\x00\x04\x08\x00\x02\x00\x00\x00\x00\x64\x08\x00\x07\x00\x6e\x61\x74\x00"),
 		// nft add rule ip ts-nat-test ts-postrouting-test meta mark & 0x00ff0000 == 0x00040000 counter masquerade
-		[]byte("\x02\x00\x00\x00\x10\x00\x01\x00\x74\x73\x2d\x6e\x61\x74\x2d\x74\x65\x73\x74\x00\x18\x00\x02\x00\x74\x73\x2d\x70\x6f\x73\x74\x72\x6f\x75\x74\x69\x6e\x67\x2d\x74\x65\x73\x74\x00\xf4\x00\x04\x80\x24\x00\x01\x80\x09\x00\x01\x00\x6d\x65\x74\x61\x00\x00\x00\x00\x14\x00\x02\x80\x08\x00\x02\x00\x00\x00\x00\x03\x08\x00\x01\x00\x00\x00\x00\x01\x44\x00\x01\x80\x0c\x00\x01\x00\x62\x69\x74\x77\x69\x73\x65\x00\x34\x00\x02\x80\x08\x00\x01\x00\x00\x00\x00\x01\x08\x00\x02\x00\x00\x00\x00\x01\x08\x00\x03\x00\x00\x00\x00\x04\x0c\x00\x04\x80\x08\x00\x01\x00\x00\xff\x00\x00\x0c\x00\x05\x80\x08\x00\x01\x00\x00\x00\x00\x00\x2c\x00\x01\x80\x08\x00\x01\x00\x63\x6d\x70\x00\x20\x00\x02\x80\x08\x00\x01\x00\x00\x00\x00\x01\x08\x00\x02\x00\x00\x00\x00\x00\x0c\x00\x03\x80\x08\x00\x01\x00\x00\x04\x00\x00\x2c\x00\x01\x80\x0c\x00\x01\x00\x63\x6f\x75\x6e\x74\x65\x72\x00\x1c\x00\x02\x80\x0c\x00\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x0c\x00\x02\x00\x00\x00\x00\x00\x00\x00\x00\x00\x30\x00\x01\x80\x0e\x00\x01\x00\x69\x6d\x6d\x65\x64\x69\x61\x74\x65\x00\x00\x00\x1c\x00\x02\x80\x08\x00\x01\x00\x00\x00\x00\x00\x10\x00\x02\x80\x0c\x00\x02\x80\x08\x00\x01\x00\x00\x00\x00\x01"),
+		[]byte("\x02\x00\x00\x00\x10\x00\x01\x00\x74\x73\x2d\x6e\x61\x74\x2d\x74\x65\x73\x74\x00\x18\x00\x02\x00\x74\x73\x2d\x70\x6f\x73\x74\x72\x6f\x75\x74\x69\x6e\x67\x2d\x74\x65\x73\x74\x00\xd8\x00\x04\x80\x24\x00\x01\x80\x09\x00\x01\x00\x6d\x65\x74\x61\x00\x00\x00\x00\x14\x00\x02\x80\x08\x00\x02\x00\x00\x00\x00\x03\x08\x00\x01\x00\x00\x00\x00\x01\x44\x00\x01\x80\x0c\x00\x01\x00\x62\x69\x74\x77\x69\x73\x65\x00\x34\x00\x02\x80\x08\x00\x01\x00\x00\x00\x00\x01\x08\x00\x02\x00\x00\x00\x00\x01\x08\x00\x03\x00\x00\x00\x00\x04\x0c\x00\x04\x80\x08\x00\x01\x00\x00\xff\x00\x00\x0c\x00\x05\x80\x08\x00\x01\x00\x00\x00\x00\x00\x2c\x00\x01\x80\x08\x00\x01\x00\x63\x6d\x70\x00\x20\x00\x02\x80\x08\x00\x01\x00\x00\x00\x00\x01\x08\x00\x02\x00\x00\x00\x00\x00\x0c\x00\x03\x80\x08\x00\x01\x00\x00\x04\x00\x00\x2c\x00\x01\x80\x0c\x00\x01\x00\x63\x6f\x75\x6e\x74\x65\x72\x00\x1c\x00\x02\x80\x0c\x00\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x0c\x00\x02\x00\x00\x00\x00\x00\x00\x00\x00\x00\x14\x00\x01\x80\x09\x00\x01\x00\x6d\x61\x73\x71\x00\x00\x00\x00\x04\x00\x02\x80"),
 		// batch end
 		[]byte("\x00\x00\x00\x0a"),
 	}
-	testConn := newTestConn(t, want)
+	testConn := newTestConn(t, want, nil)
 	table := testConn.AddTable(&nftables.Table{
 		Family: proto,
 		Name:   "ts-nat-test",
@@ -436,7 +443,46 @@ func TestAddMatchSubnetRouteMarkRuleMasq(t *testing.T) {
 		Hooknum:  nftables.ChainHookPostrouting,
 		Priority: nftables.ChainPriorityNATSource,
 	})
-	err := addMatchSubnetRouteMarkRule(testConn, table, chain, Accept)
+	err := addMatchSubnetRouteMarkRule(testConn, table, chain, Masq)
+	if err != nil {
+		t.Fatal(err)
+	}
+}
+
+func TestDelMatchSubnetRouteMarkMasqRule(t *testing.T) {
+	proto := nftables.TableFamilyIPv4
+	reply := [][]netlink.Message{
+		nil,
+		{{Header: netlink.Header{Length: 0x128, Type: 0xa06, Flags: 0x802, Sequence: 0xa213d55d, PID: 0x11e79}, Data: []uint8{0x2, 0x0, 0x0, 0x8c, 0xd, 0x0, 0x1, 0x0, 0x6e, 0x61, 0x74, 0x2d, 0x74, 0x65, 0x73, 0x74, 0x0, 0x0, 0x0, 0x0, 0x18, 0x0, 0x2, 0x0, 0x74, 0x73, 0x2d, 0x70, 0x6f, 0x73, 0x74, 0x72, 0x6f, 0x75, 0x74, 0x69, 0x6e, 0x67, 0x2d, 0x74, 0x65, 0x73, 0x74, 0x0, 0xc, 0x0, 0x3, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x4, 0xe0, 0x0, 0x4, 0x0, 0x24, 0x0, 0x1, 0x0, 0x9, 0x0, 0x1, 0x0, 0x6d, 0x65, 0x74, 0x61, 0x0, 0x0, 0x0, 0x0, 0x14, 0x0, 0x2, 0x0, 0x8, 0x0, 0x2, 0x0, 0x0, 0x0, 0x0, 0x3, 0x8, 0x0, 0x1, 0x0, 0x0, 0x0, 0x0, 0x1, 0x4c, 0x0, 0x1, 0x0, 0xc, 0x0, 0x1, 0x0, 0x62, 0x69, 0x74, 0x77, 0x69, 0x73, 0x65, 0x0, 0x3c, 0x0, 0x2, 0x0, 0x8, 0x0, 0x1, 0x0, 0x0, 0x0, 0x0, 0x1, 0x8, 0x0, 0x2, 0x0, 0x0, 0x0, 0x0, 0x1, 0x8, 0x0, 0x3, 0x0, 0x0, 0x0, 0x0, 0x4, 0x8, 0x0, 0x6, 0x0, 0x0, 0x0, 0x0, 0x0, 0xc, 0x0, 0x4, 0x0, 0x8, 0x0, 0x1, 0x0, 0x0, 0xff, 0x0, 0x0, 0xc, 0x0, 0x5, 0x0, 0x8, 0x0, 0x1, 0x0, 0x0, 0x0, 0x0, 0x0, 0x2c, 0x0, 0x1, 0x0, 0x8, 0x0, 0x1, 0x0, 0x63, 0x6d, 0x70, 0x0, 0x20, 0x0, 0x2, 0x0, 0x8, 0x0, 0x1, 0x0, 0x0, 0x0, 0x0, 0x1, 0x8, 0x0, 0x2, 0x0, 0x0, 0x0, 0x0, 0x0, 0xc, 0x0, 0x3, 0x0, 0x8, 0x0, 0x1, 0x0, 0x0, 0x4, 0x0, 0x0, 0x2c, 0x0, 0x1, 0x0, 0xc, 0x0, 0x1, 0x0, 0x63, 0x6f, 0x75, 0x6e, 0x74, 0x65, 0x72, 0x0, 0x1c, 0x0, 0x2, 0x0, 0xc, 0x0, 0x1, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0xc, 0x0, 0x2, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x14, 0x0, 0x1, 0x0, 0x9, 0x0, 0x1, 0x0, 0x6d, 0x61, 0x73, 0x71, 0x0, 0x0, 0x0, 0x0, 0x4, 0x0, 0x2, 0x0}}},
+		{{Header: netlink.Header{Length: 0x14, Type: 0x3, Flags: 0x2, Sequence: 0x311fdccb, PID: 0x11e79}, Data: []uint8{0x0, 0x0, 0x0, 0x0}}},
+		{{Header: netlink.Header{Length: 0x24, Type: 0x2, Flags: 0x100, Sequence: 0x311fdccb, PID: 0x11e79}, Data: []uint8{0x0, 0x0, 0x0, 0x0, 0x48, 0x0, 0x0, 0x0, 0x8, 0xa, 0x5, 0x0, 0xcb, 0xdc, 0x1f, 0x31, 0x79, 0x1e, 0x1, 0x0}}},
+	}
+	want := [][]byte{
+		// get rules in nat-test table ts-postrouting-test chain
+		[]byte("\x02\x00\x00\x00\x0d\x00\x01\x00\x6e\x61\x74\x2d\x74\x65\x73\x74\x00\x00\x00\x00\x18\x00\x02\x00\x74\x73\x2d\x70\x6f\x73\x74\x72\x6f\x75\x74\x69\x6e\x67\x2d\x74\x65\x73\x74\x00"),
+		// batch begin
+		[]byte("\x00\x00\x00\x0a"),
+		// nft delete rule ip nat-test ts-postrouting-test handle 4
+		[]byte("\x02\x00\x00\x00\x0d\x00\x01\x00\x6e\x61\x74\x2d\x74\x65\x73\x74\x00\x00\x00\x00\x18\x00\x02\x00\x74\x73\x2d\x70\x6f\x73\x74\x72\x6f\x75\x74\x69\x6e\x67\x2d\x74\x65\x73\x74\x00\x0c\x00\x03\x00\x00\x00\x00\x00\x00\x00\x00\x04"),
+		// batch end
+		[]byte("\x00\x00\x00\x0a"),
+	}
+
+	conn := newTestConn(t, want, reply)
+
+	table := &nftables.Table{
+		Family: proto,
+		Name:   "nat-test",
+	}
+	chain := &nftables.Chain{
+		Name:     "ts-postrouting-test",
+		Table:    table,
+		Type:     nftables.ChainTypeNAT,
+		Hooknum:  nftables.ChainHookPostrouting,
+		Priority: nftables.ChainPriorityNATSource,
+	}
+
+	err := delMatchSubnetRouteMarkMasqRule(conn, table, chain)
 	if err != nil {
 		t.Fatal(err)
 	}
@@ -456,7 +502,7 @@ func TestAddMatchSubnetRouteMarkRuleAccept(t *testing.T) {
 		// batch end
 		[]byte("\x00\x00\x00\x0a"),
 	}
-	testConn := newTestConn(t, want)
+	testConn := newTestConn(t, want, nil)
 	table := testConn.AddTable(&nftables.Table{
 		Family: proto,
 		Name:   "ts-filter-test",
diff --git a/wgengine/magicsock/magicsock.go b/wgengine/magicsock/magicsock.go
index 31bf66b2b..7df46f76c 100644
--- a/wgengine/magicsock/magicsock.go
+++ b/wgengine/magicsock/magicsock.go
@@ -9,6 +9,7 @@ import (
 	"bufio"
 	"bytes"
 	"context"
+	"encoding/binary"
 	"errors"
 	"expvar"
 	"fmt"
@@ -316,7 +317,11 @@ type Conn struct {
 	// by node key, node ID, and discovery key.
 	peerMap peerMap
 
-	// discoInfo is the state for an active DiscoKey.
+	// relayManager manages allocation and handshaking of
+	// [tailscale.com/net/udprelay.Server] endpoints.
+	relayManager relayManager
+
+	// discoInfo is the state for an active peer DiscoKey.
 	discoInfo map[key.DiscoPublic]*discoInfo
 
 	// netInfoFunc is a callback that provides a tailcfg.NetInfo when
@@ -1624,6 +1629,27 @@ func (c *Conn) sendDiscoMessage(dst netip.AddrPort, geneveVNI *uint32, dstKey ke
 		c.mu.Unlock()
 		return false, errConnClosed
 	}
+	var di *discoInfo
+	switch {
+	case isRelayHandshakeMsg:
+		var ok bool
+		di, ok = c.relayManager.discoInfo(dstDisco)
+		if !ok {
+			c.mu.Unlock()
+			return false, errors.New("unknown relay server")
+		}
+	case c.peerMap.knownPeerDiscoKey(dstDisco):
+		di = c.discoInfoForKnownPeerLocked(dstDisco)
+	default:
+		// This is an attempt to send to an unknown peer that is not a relay
+		// server. This can happen when a call to the current function, which is
+		// often via a new goroutine, races with applying a change in the
+		// netmap, e.g. the associated peer(s) for dstDisco goes away.
+		c.mu.Unlock()
+		return false, errors.New("unknown peer")
+	}
+	c.mu.Unlock()
+
 	pkt := make([]byte, 0, 512) // TODO: size it correctly? pool? if it matters.
 	if geneveVNI != nil {
 		gh := packet.GeneveHeader{
@@ -1640,23 +1666,6 @@ func (c *Conn) sendDiscoMessage(dst netip.AddrPort, geneveVNI *uint32, dstKey ke
 	}
 	pkt = append(pkt, disco.Magic...)
 	pkt = c.discoPublic.AppendTo(pkt)
-	var di *discoInfo
-	if !isRelayHandshakeMsg {
-		di = c.discoInfoLocked(dstDisco)
-	} else {
-		// c.discoInfoLocked() caches [*discoInfo] for dstDisco. It assumes that
-		// dstDisco is a known Tailscale peer, and will be cleaned around
-		// network map changes. In the case of a relay handshake message,
-		// dstDisco belongs to a relay server with a disco key that is
-		// discovered at endpoint allocation time or [disco.CallMeMaybeVia]
-		// reception time. There is no clear ending to its lifetime, so we
-		// can't cache with the same strategy. Instead, generate the shared
-		// key on the fly for now.
-		di = &discoInfo{
-			sharedKey: c.discoPrivate.Shared(dstDisco),
-		}
-	}
-	c.mu.Unlock()
 
 	if isDERP {
 		metricSendDiscoDERP.Add(1)
@@ -1707,6 +1716,45 @@ const (
 	discoRXPathRawSocket discoRXPath = "raw socket"
 )
 
+const discoHeaderLen = len(disco.Magic) + key.DiscoPublicRawLen
+
+// isDiscoMaybeGeneve reports whether msg is a Tailscale Disco protocol
+// message, and if true, whether it is encapsulated by a Geneve header.
+//
+// isGeneveEncap is only relevant when isDiscoMsg is true.
+//
+// Naked Disco, Geneve followed by Disco, and naked WireGuard can be confidently
+// distinguished based on the following:
+//  1. [disco.Magic] is sufficiently non-overlapping with a Geneve protocol
+//     field value of [packet.GeneveProtocolDisco].
+//  2. [disco.Magic] is sufficiently non-overlapping with the first 4 bytes of
+//     a WireGuard packet.
+//  3. [packet.GeneveHeader] with a Geneve protocol field value of
+//     [packet.GeneveProtocolDisco] is sufficiently non-overlapping with the
+//     first 4 bytes of a WireGuard packet.
+func isDiscoMaybeGeneve(msg []byte) (isDiscoMsg bool, isGeneveEncap bool) {
+	if len(msg) < discoHeaderLen {
+		return false, false
+	}
+	if string(msg[:len(disco.Magic)]) == disco.Magic {
+		return true, false
+	}
+	if len(msg) < packet.GeneveFixedHeaderLength+discoHeaderLen {
+		return false, false
+	}
+	if msg[0]&0xC0 != 0 || // version bits that we always transmit as 0s
+		msg[1]&0x3F != 0 || // reserved bits that we always transmit as 0s
+		binary.BigEndian.Uint16(msg[2:4]) != packet.GeneveProtocolDisco ||
+		msg[7] != 0 { // reserved byte that we always transmit as 0
+		return false, false
+	}
+	msg = msg[packet.GeneveFixedHeaderLength:]
+	if string(msg[:len(disco.Magic)]) == disco.Magic {
+		return true, true
+	}
+	return false, false
+}
+
 // handleDiscoMessage handles a discovery message and reports whether
 // msg was a Tailscale inter-node discovery message.
 //
@@ -1722,18 +1770,28 @@ const (
 // it was received from at the DERP layer. derpNodeSrc is zero when received
 // over UDP.
 func (c *Conn) handleDiscoMessage(msg []byte, src netip.AddrPort, derpNodeSrc key.NodePublic, via discoRXPath) (isDiscoMsg bool) {
-	const headerLen = len(disco.Magic) + key.DiscoPublicRawLen
-	if len(msg) < headerLen || string(msg[:len(disco.Magic)]) != disco.Magic {
-		return false
+	isDiscoMsg, isGeneveEncap := isDiscoMaybeGeneve(msg)
+	if !isDiscoMsg {
+		return
 	}
+	var geneve packet.GeneveHeader
+	if isGeneveEncap {
+		err := geneve.Decode(msg)
+		if err != nil {
+			// Decode only returns an error when 'msg' is too short, and
+			// 'isGeneveEncap' indicates it's a sufficient length.
+			c.logf("[unexpected] geneve header decoding error: %v", err)
+			return
+		}
+		msg = msg[packet.GeneveFixedHeaderLength:]
+	}
+	// The control bit should only be set for relay handshake messages
+	// terminating on or originating from a UDP relay server. We have yet to
+	// open the encrypted payload to determine the [disco.MessageType], but
+	// we assert it should be handshake-related.
+	shouldBeRelayHandshakeMsg := isGeneveEncap && geneve.Control
 
-	// If the first four parts are the prefix of disco.Magic
-	// (0x5453f09f) then it's definitely not a valid WireGuard
-	// packet (which starts with little-endian uint32 1, 2, 3, 4).
-	// Use naked returns for all following paths.
-	isDiscoMsg = true
-
-	sender := key.DiscoPublicFromRaw32(mem.B(msg[len(disco.Magic):headerLen]))
+	sender := key.DiscoPublicFromRaw32(mem.B(msg[len(disco.Magic):discoHeaderLen]))
 
 	c.mu.Lock()
 	defer c.mu.Unlock()
@@ -1750,7 +1808,20 @@ func (c *Conn) handleDiscoMessage(msg []byte, src netip.AddrPort, derpNodeSrc ke
 		return
 	}
 
-	if !c.peerMap.knownPeerDiscoKey(sender) {
+	var di *discoInfo
+	switch {
+	case shouldBeRelayHandshakeMsg:
+		var ok bool
+		di, ok = c.relayManager.discoInfo(sender)
+		if !ok {
+			if debugDisco() {
+				c.logf("magicsock: disco: ignoring disco-looking relay handshake frame, no active handshakes with key %v over VNI %d", sender.ShortString(), geneve.VNI)
+			}
+			return
+		}
+	case c.peerMap.knownPeerDiscoKey(sender):
+		di = c.discoInfoForKnownPeerLocked(sender)
+	default:
 		metricRecvDiscoBadPeer.Add(1)
 		if debugDisco() {
 			c.logf("magicsock: disco: ignoring disco-looking frame, don't know of key %v", sender.ShortString())
@@ -1759,7 +1830,7 @@ func (c *Conn) handleDiscoMessage(msg []byte, src netip.AddrPort, derpNodeSrc ke
 	}
 
 	isDERP := src.Addr() == tailcfg.DerpMagicIPAddr
-	if !isDERP {
+	if !isDERP && !shouldBeRelayHandshakeMsg {
 		// Record receive time for UDP transport packets.
 		pi, ok := c.peerMap.byIPPort[src]
 		if ok {
@@ -1767,17 +1838,13 @@ func (c *Conn) handleDiscoMessage(msg []byte, src netip.AddrPort, derpNodeSrc ke
 		}
 	}
 
-	// We're now reasonably sure we're expecting communication from
-	// this peer, do the heavy crypto lifting to see what they want.
-	//
-	// From here on, peerNode and de are non-nil.
+	// We're now reasonably sure we're expecting communication from 'sender',
+	// do the heavy crypto lifting to see what they want.
 
-	di := c.discoInfoLocked(sender)
-
-	sealedBox := msg[headerLen:]
+	sealedBox := msg[discoHeaderLen:]
 	payload, ok := di.sharedKey.Open(sealedBox)
 	if !ok {
-		// This might be have been intended for a previous
+		// This might have been intended for a previous
 		// disco key.  When we restart we get a new disco key
 		// and old packets might've still been in flight (or
 		// scheduled). This is particularly the case for LANs
@@ -1820,6 +1887,19 @@ func (c *Conn) handleDiscoMessage(msg []byte, src netip.AddrPort, derpNodeSrc ke
 		metricRecvDiscoUDP.Add(1)
 	}
 
+	if shouldBeRelayHandshakeMsg {
+		challenge, ok := dm.(*disco.BindUDPRelayEndpointChallenge)
+		if !ok {
+			// We successfully parsed the disco message, but it wasn't a
+			// challenge. We should never receive other message types
+			// from a relay server with the Geneve header control bit set.
+			c.logf("[unexpected] %T packets should not come from a relay server with Geneve control bit set", dm)
+			return
+		}
+		c.relayManager.handleBindUDPRelayEndpointChallenge(challenge, di, src, geneve.VNI)
+		return
+	}
+
 	switch dm := dm.(type) {
 	case *disco.Ping:
 		metricRecvDiscoPing.Add(1)
@@ -1835,18 +1915,28 @@ func (c *Conn) handleDiscoMessage(msg []byte, src netip.AddrPort, derpNodeSrc ke
 			}
 			return true
 		})
-	case *disco.CallMeMaybe:
+	case *disco.CallMeMaybe, *disco.CallMeMaybeVia:
+		var via *disco.CallMeMaybeVia
+		isVia := false
+		msgType := "CallMeMaybe"
+		cmm, ok := dm.(*disco.CallMeMaybe)
+		if !ok {
+			via = dm.(*disco.CallMeMaybeVia)
+			msgType = "CallMeMaybeVia"
+			isVia = true
+		}
+
 		metricRecvDiscoCallMeMaybe.Add(1)
 		if !isDERP || derpNodeSrc.IsZero() {
-			// CallMeMaybe messages should only come via DERP.
-			c.logf("[unexpected] CallMeMaybe packets should only come via DERP")
+			// CallMeMaybe{Via} messages should only come via DERP.
+			c.logf("[unexpected] %s packets should only come via DERP", msgType)
 			return
 		}
 		nodeKey := derpNodeSrc
 		ep, ok := c.peerMap.endpointForNodeKey(nodeKey)
 		if !ok {
 			metricRecvDiscoCallMeMaybeBadNode.Add(1)
-			c.logf("magicsock: disco: ignoring CallMeMaybe from %v; %v is unknown", sender.ShortString(), derpNodeSrc.ShortString())
+			c.logf("magicsock: disco: ignoring %s from %v; %v is unknown", msgType, sender.ShortString(), derpNodeSrc.ShortString())
 			return
 		}
 		epDisco := ep.disco.Load()
@@ -1855,14 +1945,23 @@ func (c *Conn) handleDiscoMessage(msg []byte, src netip.AddrPort, derpNodeSrc ke
 		}
 		if epDisco.key != di.discoKey {
 			metricRecvDiscoCallMeMaybeBadDisco.Add(1)
-			c.logf("[unexpected] CallMeMaybe from peer via DERP whose netmap discokey != disco source")
+			c.logf("[unexpected] %s from peer via DERP whose netmap discokey != disco source", msgType)
 			return
 		}
-		c.dlogf("[v1] magicsock: disco: %v<-%v (%v, %v)  got call-me-maybe, %d endpoints",
-			c.discoShort, epDisco.short,
-			ep.publicKey.ShortString(), derpStr(src.String()),
-			len(dm.MyNumber))
-		go ep.handleCallMeMaybe(dm)
+		if isVia {
+			c.dlogf("[v1] magicsock: disco: %v<-%v via %v (%v, %v)  got call-me-maybe-via, %d endpoints",
+				c.discoShort, epDisco.short, via.ServerDisco.ShortString(),
+				ep.publicKey.ShortString(), derpStr(src.String()),
+				len(via.AddrPorts))
+			c.relayManager.handleCallMeMaybeVia(via)
+		} else {
+			c.dlogf("[v1] magicsock: disco: %v<-%v (%v, %v)  got call-me-maybe, %d endpoints",
+				c.discoShort, epDisco.short,
+				ep.publicKey.ShortString(), derpStr(src.String()),
+				len(cmm.MyNumber))
+			go ep.handleCallMeMaybe(cmm)
+		}
+
 	}
 	return
 }
@@ -2034,10 +2133,15 @@ func (c *Conn) enqueueCallMeMaybe(derpAddr netip.AddrPort, de *endpoint) {
 	}
 }
 
-// discoInfoLocked returns the previous or new discoInfo for k.
+// discoInfoForKnownPeerLocked returns the previous or new discoInfo for k.
+//
+// Callers must only pass key.DiscoPublic's that are present in and
+// lifetime-managed via [Conn].peerMap. UDP relay server disco keys are discovered
+// at relay endpoint allocation time or [disco.CallMeMaybeVia] reception time
+// and therefore must never pass through this method.
 //
 // c.mu must be held.
-func (c *Conn) discoInfoLocked(k key.DiscoPublic) *discoInfo {
+func (c *Conn) discoInfoForKnownPeerLocked(k key.DiscoPublic) *discoInfo {
 	di, ok := c.discoInfo[k]
 	if !ok {
 		di = &discoInfo{
diff --git a/wgengine/magicsock/magicsock_test.go b/wgengine/magicsock/magicsock_test.go
index f50f21f56..1a899ea22 100644
--- a/wgengine/magicsock/magicsock_test.go
+++ b/wgengine/magicsock/magicsock_test.go
@@ -3155,3 +3155,165 @@ func TestNetworkDownSendErrors(t *testing.T) {
 		t.Errorf("expected NetworkDown to increment packet dropped metric; got %q", resp.Body.String())
 	}
 }
+
+func Test_isDiscoMaybeGeneve(t *testing.T) {
+	discoPub := key.DiscoPublicFromRaw32(mem.B([]byte{1: 1, 30: 30, 31: 31}))
+	nakedDisco := make([]byte, 0, 512)
+	nakedDisco = append(nakedDisco, disco.Magic...)
+	nakedDisco = discoPub.AppendTo(nakedDisco)
+
+	geneveEncapDisco := make([]byte, packet.GeneveFixedHeaderLength+len(nakedDisco))
+	gh := packet.GeneveHeader{
+		Version:  0,
+		Protocol: packet.GeneveProtocolDisco,
+		VNI:      1,
+		Control:  true,
+	}
+	err := gh.Encode(geneveEncapDisco)
+	if err != nil {
+		t.Fatal(err)
+	}
+	copy(geneveEncapDisco[packet.GeneveFixedHeaderLength:], nakedDisco)
+
+	nakedWireGuardInitiation := make([]byte, len(geneveEncapDisco))
+	binary.LittleEndian.PutUint32(nakedWireGuardInitiation, device.MessageInitiationType)
+	nakedWireGuardResponse := make([]byte, len(geneveEncapDisco))
+	binary.LittleEndian.PutUint32(nakedWireGuardResponse, device.MessageResponseType)
+	nakedWireGuardCookieReply := make([]byte, len(geneveEncapDisco))
+	binary.LittleEndian.PutUint32(nakedWireGuardCookieReply, device.MessageCookieReplyType)
+	nakedWireGuardTransport := make([]byte, len(geneveEncapDisco))
+	binary.LittleEndian.PutUint32(nakedWireGuardTransport, device.MessageTransportType)
+
+	geneveEncapWireGuard := make([]byte, packet.GeneveFixedHeaderLength+len(nakedWireGuardInitiation))
+	gh = packet.GeneveHeader{
+		Version:  0,
+		Protocol: packet.GeneveProtocolWireGuard,
+		VNI:      1,
+		Control:  true,
+	}
+	err = gh.Encode(geneveEncapWireGuard)
+	if err != nil {
+		t.Fatal(err)
+	}
+	copy(geneveEncapWireGuard[packet.GeneveFixedHeaderLength:], nakedWireGuardInitiation)
+
+	geneveEncapDiscoNonZeroGeneveVersion := make([]byte, packet.GeneveFixedHeaderLength+len(nakedDisco))
+	gh = packet.GeneveHeader{
+		Version:  1,
+		Protocol: packet.GeneveProtocolDisco,
+		VNI:      1,
+		Control:  true,
+	}
+	err = gh.Encode(geneveEncapDiscoNonZeroGeneveVersion)
+	if err != nil {
+		t.Fatal(err)
+	}
+	copy(geneveEncapDiscoNonZeroGeneveVersion[packet.GeneveFixedHeaderLength:], nakedDisco)
+
+	geneveEncapDiscoNonZeroGeneveReservedBits := make([]byte, packet.GeneveFixedHeaderLength+len(nakedDisco))
+	gh = packet.GeneveHeader{
+		Version:  0,
+		Protocol: packet.GeneveProtocolDisco,
+		VNI:      1,
+		Control:  true,
+	}
+	err = gh.Encode(geneveEncapDiscoNonZeroGeneveReservedBits)
+	if err != nil {
+		t.Fatal(err)
+	}
+	geneveEncapDiscoNonZeroGeneveReservedBits[1] |= 0x3F
+	copy(geneveEncapDiscoNonZeroGeneveReservedBits[packet.GeneveFixedHeaderLength:], nakedDisco)
+
+	geneveEncapDiscoNonZeroGeneveVNILSB := make([]byte, packet.GeneveFixedHeaderLength+len(nakedDisco))
+	gh = packet.GeneveHeader{
+		Version:  0,
+		Protocol: packet.GeneveProtocolDisco,
+		VNI:      1,
+		Control:  true,
+	}
+	err = gh.Encode(geneveEncapDiscoNonZeroGeneveVNILSB)
+	if err != nil {
+		t.Fatal(err)
+	}
+	geneveEncapDiscoNonZeroGeneveVNILSB[7] |= 0xFF
+	copy(geneveEncapDiscoNonZeroGeneveVNILSB[packet.GeneveFixedHeaderLength:], nakedDisco)
+
+	tests := []struct {
+		name              string
+		msg               []byte
+		wantIsDiscoMsg    bool
+		wantIsGeneveEncap bool
+	}{
+		{
+			name:              "naked disco",
+			msg:               nakedDisco,
+			wantIsDiscoMsg:    true,
+			wantIsGeneveEncap: false,
+		},
+		{
+			name:              "geneve encap disco",
+			msg:               geneveEncapDisco,
+			wantIsDiscoMsg:    true,
+			wantIsGeneveEncap: true,
+		},
+		{
+			name:              "geneve encap disco nonzero geneve version",
+			msg:               geneveEncapDiscoNonZeroGeneveVersion,
+			wantIsDiscoMsg:    false,
+			wantIsGeneveEncap: false,
+		},
+		{
+			name:              "geneve encap disco nonzero geneve reserved bits",
+			msg:               geneveEncapDiscoNonZeroGeneveReservedBits,
+			wantIsDiscoMsg:    false,
+			wantIsGeneveEncap: false,
+		},
+		{
+			name:              "geneve encap disco nonzero geneve vni lsb",
+			msg:               geneveEncapDiscoNonZeroGeneveVNILSB,
+			wantIsDiscoMsg:    false,
+			wantIsGeneveEncap: false,
+		},
+		{
+			name:              "geneve encap wireguard",
+			msg:               geneveEncapWireGuard,
+			wantIsDiscoMsg:    false,
+			wantIsGeneveEncap: false,
+		},
+		{
+			name:              "naked WireGuard Initiation type",
+			msg:               nakedWireGuardInitiation,
+			wantIsDiscoMsg:    false,
+			wantIsGeneveEncap: false,
+		},
+		{
+			name:              "naked WireGuard Response type",
+			msg:               nakedWireGuardResponse,
+			wantIsDiscoMsg:    false,
+			wantIsGeneveEncap: false,
+		},
+		{
+			name:              "naked WireGuard Cookie Reply type",
+			msg:               nakedWireGuardCookieReply,
+			wantIsDiscoMsg:    false,
+			wantIsGeneveEncap: false,
+		},
+		{
+			name:              "naked WireGuard Transport type",
+			msg:               nakedWireGuardTransport,
+			wantIsDiscoMsg:    false,
+			wantIsGeneveEncap: false,
+		},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			gotIsDiscoMsg, gotIsGeneveEncap := isDiscoMaybeGeneve(tt.msg)
+			if gotIsDiscoMsg != tt.wantIsDiscoMsg {
+				t.Errorf("isDiscoMaybeGeneve() gotIsDiscoMsg = %v, want %v", gotIsDiscoMsg, tt.wantIsDiscoMsg)
+			}
+			if gotIsGeneveEncap != tt.wantIsGeneveEncap {
+				t.Errorf("isDiscoMaybeGeneve() gotIsGeneveEncap = %v, want %v", gotIsGeneveEncap, tt.wantIsGeneveEncap)
+			}
+		})
+	}
+}
diff --git a/wgengine/magicsock/relaymanager.go b/wgengine/magicsock/relaymanager.go
new file mode 100644
index 000000000..bf737b078
--- /dev/null
+++ b/wgengine/magicsock/relaymanager.go
@@ -0,0 +1,51 @@
+// Copyright (c) Tailscale Inc & AUTHORS
+// SPDX-License-Identifier: BSD-3-Clause
+
+package magicsock
+
+import (
+	"net/netip"
+	"sync"
+
+	"tailscale.com/disco"
+	"tailscale.com/types/key"
+)
+
+// relayManager manages allocation and handshaking of
+// [tailscale.com/net/udprelay.Server] endpoints. The zero value is ready for
+// use.
+type relayManager struct {
+	mu                     sync.Mutex // guards the following fields
+	discoInfoByServerDisco map[key.DiscoPublic]*discoInfo
+}
+
+func (h *relayManager) initLocked() {
+	if h.discoInfoByServerDisco != nil {
+		return
+	}
+	h.discoInfoByServerDisco = make(map[key.DiscoPublic]*discoInfo)
+}
+
+// discoInfo returns a [*discoInfo] for 'serverDisco' if there is an
+// active/ongoing handshake with it, otherwise it returns nil, false.
+func (h *relayManager) discoInfo(serverDisco key.DiscoPublic) (_ *discoInfo, ok bool) {
+	h.mu.Lock()
+	defer h.mu.Unlock()
+	h.initLocked()
+	di, ok := h.discoInfoByServerDisco[serverDisco]
+	return di, ok
+}
+
+func (h *relayManager) handleCallMeMaybeVia(dm *disco.CallMeMaybeVia) {
+	h.mu.Lock()
+	defer h.mu.Unlock()
+	h.initLocked()
+	// TODO(jwhited): implement
+}
+
+func (h *relayManager) handleBindUDPRelayEndpointChallenge(dm *disco.BindUDPRelayEndpointChallenge, di *discoInfo, src netip.AddrPort, vni uint32) {
+	h.mu.Lock()
+	defer h.mu.Unlock()
+	h.initLocked()
+	// TODO(jwhited): implement
+}