control/controlclient, ipn: add client audit logging (#14950)

updates tailscale/corp#26435

Adds client support for sending audit logs to control via /machine/audit-log.
Specifically implements audit logging for user initiated disconnections.

This will require further work to optimize the peristant storage and exclusion
via build tags for mobile:
tailscale/corp#27011
tailscale/corp#27012

Signed-off-by: Jonathan Nobels <jonathan@tailscale.com>

cmd/k8s-operator/depaware.txt

@@ -814,6 +814,7 @@ tailscale.com/cmd/k8s-operator dependencies: (generated by github.com/tailscale/
         tailscale.com/internal/client/tailscale                      from tailscale.com/cmd/k8s-operator
         tailscale.com/internal/noiseconn                             from tailscale.com/control/controlclient
         tailscale.com/ipn                                            from tailscale.com/client/local+
+        tailscale.com/ipn/auditlog                                   from tailscale.com/ipn/ipnlocal+
         tailscale.com/ipn/conffile                                   from tailscale.com/ipn/ipnlocal+
      💣 tailscale.com/ipn/desktop                                    from tailscale.com/ipn/ipnlocal+
      💣 tailscale.com/ipn/ipnauth                                    from tailscale.com/ipn/ipnlocal+

cmd/tailscaled/depaware.txt

@@ -271,6 +271,7 @@ tailscale.com/cmd/tailscaled dependencies: (generated by github.com/tailscale/de
         tailscale.com/hostinfo                                       from tailscale.com/client/web+
         tailscale.com/internal/noiseconn                             from tailscale.com/control/controlclient
         tailscale.com/ipn                                            from tailscale.com/client/local+
+        tailscale.com/ipn/auditlog                                   from tailscale.com/ipn/ipnlocal+
         tailscale.com/ipn/conffile                                   from tailscale.com/cmd/tailscaled+
      💣 tailscale.com/ipn/desktop                                    from tailscale.com/cmd/tailscaled+
      💣 tailscale.com/ipn/ipnauth                                    from tailscale.com/ipn/ipnlocal+

control/controlclient/auto.go

@@ -119,6 +119,7 @@ type Auto struct {
 	updateCh      chan struct{} // readable when we should inform the server of a change
 	observer      Observer      // called to update Client status; always non-nil
 	observerQueue execqueue.ExecQueue
+	shutdownFn    func() // to be called prior to shutdown or nil
 
 	unregisterHealthWatch func()
 
@@ -189,6 +190,7 @@ func NewNoStart(opts Options) (_ *Auto, err error) {
 		mapDone:    make(chan struct{}),
 		updateDone: make(chan struct{}),
 		observer:   opts.Observer,
+		shutdownFn: opts.Shutdown,
 	}
 	c.authCtx, c.authCancel = context.WithCancel(context.Background())
 	c.authCtx = sockstats.WithSockStats(c.authCtx, sockstats.LabelControlClientAuto, opts.Logf)
@@ -755,6 +757,7 @@ func (c *Auto) Shutdown() {
 		return
 	}
 	c.logf("client.Shutdown ...")
+	shutdownFn := c.shutdownFn
 
 	direct := c.direct
 	c.closed = true
@@ -767,6 +770,10 @@ func (c *Auto) Shutdown() {
 	c.unpauseWaiters = nil
 	c.mu.Unlock()
 
+	if shutdownFn != nil {
+		shutdownFn()
+	}
+
 	c.unregisterHealthWatch()
 	<-c.authDone
 	<-c.mapDone

control/controlclient/controlclient_test.go

@@ -4,6 +4,8 @@
 package controlclient
 
 import (
+	"errors"
+	"fmt"
 	"io"
 	"reflect"
 	"slices"
@@ -147,3 +149,42 @@ func TestCanSkipStatus(t *testing.T) {
 		t.Errorf("Status fields = %q; this code was only written to handle fields %q", f, want)
 	}
 }
+
+func TestRetryableErrors(t *testing.T) {
+	errorTests := []struct {
+		err  error
+		want bool
+	}{
+		{errNoNoiseClient, true},
+		{errNoNodeKey, true},
+		{fmt.Errorf("%w: %w", errNoNoiseClient, errors.New("no noise")), true},
+		{fmt.Errorf("%w: %w", errHTTPPostFailure, errors.New("bad post")), true},
+		{fmt.Errorf("%w: %w", errNoNodeKey, errors.New("not node key")), true},
+		{errBadHTTPResponse(429, "too may requests"), true},
+		{errBadHTTPResponse(500, "internal server eror"), true},
+		{errBadHTTPResponse(502, "bad gateway"), true},
+		{errBadHTTPResponse(503, "service unavailable"), true},
+		{errBadHTTPResponse(504, "gateway timeout"), true},
+		{errBadHTTPResponse(1234, "random error"), false},
+	}
+
+	for _, tt := range errorTests {
+		t.Run(tt.err.Error(), func(t *testing.T) {
+			if isRetryableErrorForTest(tt.err) != tt.want {
+				t.Fatalf("retriable: got %v, want %v", tt.err, tt.want)
+			}
+		})
+	}
+}
+
+type retryableForTest interface {
+	Retryable() bool
+}
+
+func isRetryableErrorForTest(err error) bool {
+	var ae retryableForTest
+	if errors.As(err, &ae) {
+		return ae.Retryable()
+	}
+	return false
+}

control/controlclient/direct.go

@@ -156,6 +156,11 @@ type Options struct {
 	// If we receive a new DialPlan from the server, this value will be
 	// updated.
 	DialPlan ControlDialPlanner
+
+	// Shutdown is an optional function that will be called before client shutdown is
+	// attempted. It is used to allow the client to clean up any resources or complete any
+	// tasks that are dependent on a live client.
+	Shutdown func()
 }
 
 // ControlDialPlanner is the interface optionally supplied when creating a
@@ -1662,11 +1667,11 @@ func (c *Auto) SetDeviceAttrs(ctx context.Context, attrs tailcfg.AttrUpdate) err
 func (c *Direct) SetDeviceAttrs(ctx context.Context, attrs tailcfg.AttrUpdate) error {
 	nc, err := c.getNoiseClient()
 	if err != nil {
-		return err
+		return fmt.Errorf("%w: %w", errNoNoiseClient, err)
 	}
 	nodeKey, ok := c.GetPersist().PublicNodeKeyOK()
 	if !ok {
-		return errors.New("no node key")
+		return errNoNodeKey
 	}
 	if c.panicOnUse {
 		panic("tainted client")
@@ -1697,6 +1702,47 @@ func (c *Direct) SetDeviceAttrs(ctx context.Context, attrs tailcfg.AttrUpdate) e
 	return nil
 }
 
+// SendAuditLog implements [auditlog.Transport] by sending an audit log synchronously to the control plane.
+//
+// See docs on [tailcfg.AuditLogRequest] and [auditlog.Logger] for background.
+func (c *Auto) SendAuditLog(ctx context.Context, auditLog tailcfg.AuditLogRequest) (err error) {
+	return c.direct.sendAuditLog(ctx, auditLog)
+}
+
+func (c *Direct) sendAuditLog(ctx context.Context, auditLog tailcfg.AuditLogRequest) (err error) {
+	nc, err := c.getNoiseClient()
+	if err != nil {
+		return fmt.Errorf("%w: %w", errNoNoiseClient, err)
+	}
+
+	nodeKey, ok := c.GetPersist().PublicNodeKeyOK()
+	if !ok {
+		return errNoNodeKey
+	}
+
+	req := &tailcfg.AuditLogRequest{
+		Version: tailcfg.CurrentCapabilityVersion,
+		NodeKey: nodeKey,
+		Action:  auditLog.Action,
+		Details: auditLog.Details,
+	}
+
+	if c.panicOnUse {
+		panic("tainted client")
+	}
+
+	res, err := nc.post(ctx, "/machine/audit-log", nodeKey, req)
+	if err != nil {
+		return fmt.Errorf("%w: %w", errHTTPPostFailure, err)
+	}
+	defer res.Body.Close()
+	if res.StatusCode != 200 {
+		all, _ := io.ReadAll(res.Body)
+		return errBadHTTPResponse(res.StatusCode, string(all))
+	}
+	return nil
+}
+
 func addLBHeader(req *http.Request, nodeKey key.NodePublic) {
 	if !nodeKey.IsZero() {
 		req.Header.Add(tailcfg.LBHeader, nodeKey.String())

control/controlclient/errors.go

@@ -0,0 +1,51 @@
+// Copyright (c) Tailscale Inc & AUTHORS
+// SPDX-License-Identifier: BSD-3-Clause
+
+package controlclient
+
+import (
+	"errors"
+	"fmt"
+	"net/http"
+)
+
+// apiResponseError is an error type that can be returned by controlclient
+// api requests.
+//
+// It wraps an underlying error and a flag for clients to query if the
+// error is retryable via the Retryable() method.
+type apiResponseError struct {
+	err       error
+	retryable bool
+}
+
+// Error implements [error].
+func (e *apiResponseError) Error() string {
+	return e.err.Error()
+}
+
+// Retryable reports whether the error is retryable.
+func (e *apiResponseError) Retryable() bool {
+	return e.retryable
+}
+
+func (e *apiResponseError) Unwrap() error { return e.err }
+
+var (
+	errNoNodeKey       = &apiResponseError{errors.New("no node key"), true}
+	errNoNoiseClient   = &apiResponseError{errors.New("no noise client"), true}
+	errHTTPPostFailure = &apiResponseError{errors.New("http failure"), true}
+)
+
+func errBadHTTPResponse(code int, msg string) error {
+	retryable := false
+	switch code {
+	case http.StatusTooManyRequests,
+		http.StatusInternalServerError,
+		http.StatusBadGateway,
+		http.StatusServiceUnavailable,
+		http.StatusGatewayTimeout:
+		retryable = true
+	}
+	return &apiResponseError{fmt.Errorf("http error %d: %s", code, msg), retryable}
+}

ipn/auditlog/auditlog.go

@@ -0,0 +1,466 @@
+// Copyright (c) Tailscale Inc & AUTHORS
+// SPDX-License-Identifier: BSD-3-Clause
+
+// Package auditlog provides a mechanism for logging audit events.
+package auditlog
+
+import (
+	"context"
+	"encoding/json"
+	"errors"
+	"fmt"
+	"sort"
+	"sync"
+	"time"
+
+	"tailscale.com/ipn"
+	"tailscale.com/tailcfg"
+	"tailscale.com/types/logger"
+	"tailscale.com/util/rands"
+	"tailscale.com/util/set"
+)
+
+// transaction represents an audit log that has not yet been sent to the control plane.
+type transaction struct {
+	// EventID is the unique identifier for the event being logged.
+	// This is used on the client side only and is not sent to control.
+	EventID string `json:",omitempty"`
+	// Retries is the number of times the logger has attempted to send this log.
+	// This is used on the client side only and is not sent to control.
+	Retries int `json:",omitempty"`
+
+	// Action is the action to be logged. It must correspond to a known action in the control plane.
+	Action tailcfg.ClientAuditAction `json:",omitempty"`
+	// Details is an opaque string specific to the action being logged. Empty strings may not
+	// be valid depending on the action being logged.
+	Details string `json:",omitempty"`
+	// TimeStamp is the time at which the audit log was generated on the node.
+	TimeStamp time.Time `json:",omitzero"`
+}
+
+// Transport provides a means for a client to send audit logs to a consumer (typically the control plane).
+type Transport interface {
+	// SendAuditLog sends an audit log to a consumer of audit logs.
+	// Errors should be checked with [IsRetryableError] for retryability.
+	SendAuditLog(context.Context, tailcfg.AuditLogRequest) error
+}
+
+// LogStore provides a means for a [Logger] to persist logs to disk or memory.
+type LogStore interface {
+	// Save saves the given data to a persistent store. Save will overwrite existing data
+	// for the given key.
+	save(key ipn.ProfileID, txns []*transaction) error
+
+	// Load retrieves the data from a persistent store.  Returns a nil slice and
+	// no error if no data exists for the given key.
+	load(key ipn.ProfileID) ([]*transaction, error)
+}
+
+// Opts contains the configuration options for a [Logger].
+type Opts struct {
+	// RetryLimit is the maximum number of attempts the logger will make to send a log before giving up.
+	RetryLimit int
+	// Store is the persistent store used to save logs to disk. Must be non-nil.
+	Store LogStore
+	// Logf is the logger used to log messages from the audit logger. Must be non-nil.
+	Logf logger.Logf
+}
+
+// IsRetryableError returns true if the given error is retryable
+// See [controlclient.apiResponseError].  Potentially retryable errors implement the Retryable() method.
+func IsRetryableError(err error) bool {
+	var retryable interface{ Retryable() bool }
+	return errors.As(err, &retryable) && retryable.Retryable()
+}
+
+type backoffOpts struct {
+	min, max   time.Duration
+	multiplier float64
+}
+
+// .5, 1, 2, 4, 8, 10, 10, 10, 10, 10...
+var defaultBackoffOpts = backoffOpts{
+	min:        time.Millisecond * 500,
+	max:        10 * time.Second,
+	multiplier: 2,
+}
+
+// Logger provides a queue-based mechanism for submitting audit logs to the control plane - or
+// another suitable consumer. Logs are stored to disk and retried until they are successfully sent,
+// or until they permanently fail.
+//
+// Each individual profile/controlclient tuple should construct and manage a unique [Logger] instance.
+type Logger struct {
+	logf        logger.Logf
+	retryLimit  int                // the maximum number of attempts to send a log before giving up.
+	flusher     chan struct{}      // channel used to signal a flush operation.
+	done        chan struct{}      // closed when the flush worker exits.
+	ctx         context.Context    // canceled when the logger is stopped.
+	ctxCancel   context.CancelFunc // cancels ctx.
+	backoffOpts                    // backoff settings for retry operations.
+
+	// mu protects the fields below.
+	mu        sync.Mutex
+	store     LogStore      // persistent storage for unsent logs.
+	profileID ipn.ProfileID // empty if [Logger.SetProfileID] has not been called.
+	transport Transport     // nil until [Logger.Start] is called.
+}
+
+// NewLogger creates a new [Logger] with the given options.
+func NewLogger(opts Opts) *Logger {
+	ctx, cancel := context.WithCancel(context.Background())
+
+	al := &Logger{
+		retryLimit:  opts.RetryLimit,
+		logf:        logger.WithPrefix(opts.Logf, "auditlog: "),
+		store:       opts.Store,
+		flusher:     make(chan struct{}, 1),
+		done:        make(chan struct{}),
+		ctx:         ctx,
+		ctxCancel:   cancel,
+		backoffOpts: defaultBackoffOpts,
+	}
+	al.logf("created")
+	return al
+}
+
+// FlushAndStop synchronously flushes all pending logs and stops the audit logger.
+// This will block until a final flush operation completes or context is done.
+// If the logger is already stopped, this will return immediately.  All unsent
+// logs will be persisted to the store.
+func (al *Logger) FlushAndStop(ctx context.Context) {
+	al.stop()
+	al.flush(ctx)
+}
+
+// SetProfileID sets the profileID for the logger. This must be called before any logs can be enqueued.
+// The profileID of a logger cannot be changed once set.
+func (al *Logger) SetProfileID(profileID ipn.ProfileID) error {
+	al.mu.Lock()
+	defer al.mu.Unlock()
+	if al.profileID != "" {
+		return errors.New("profileID already set")
+	}
+
+	al.profileID = profileID
+	return nil
+}
+
+// Start starts the audit logger with the given transport.
+// It returns an error if the logger is already started.
+func (al *Logger) Start(t Transport) error {
+	al.mu.Lock()
+	defer al.mu.Unlock()
+
+	if al.transport != nil {
+		return errors.New("already started")
+	}
+
+	al.transport = t
+	pending, err := al.storedCountLocked()
+	if err != nil {
+		al.logf("[unexpected] failed to restore logs: %v", err)
+	}
+	go al.flushWorker()
+	if pending > 0 {
+		al.flushAsync()
+	}
+	return nil
+}
+
+// ErrAuditLogStorageFailure is returned when the logger fails to persist logs to the store.
+var ErrAuditLogStorageFailure = errors.New("audit log storage failure")
+
+// Enqueue queues an audit log to be sent to the control plane (or another suitable consumer/transport).
+// This will return an error if the underlying store fails to save the log or we fail to generate a unique
+// eventID for the log.
+func (al *Logger) Enqueue(action tailcfg.ClientAuditAction, details string) error {
+	txn := &transaction{
+		Action:    action,
+		Details:   details,
+		TimeStamp: time.Now(),
+	}
+	// Generate a suitably random eventID for the transaction.
+	txn.EventID = fmt.Sprint(txn.TimeStamp, rands.HexString(16))
+	return al.enqueue(txn)
+}
+
+// flushAsync requests an asynchronous flush.
+// It is a no-op if a flush is already pending.
+func (al *Logger) flushAsync() {
+	select {
+	case al.flusher <- struct{}{}:
+	default:
+	}
+}
+
+func (al *Logger) flushWorker() {
+	defer close(al.done)
+
+	var retryDelay time.Duration
+	retry := time.NewTimer(0)
+	retry.Stop()
+
+	for {
+		select {
+		case <-al.ctx.Done():
+			return
+		case <-al.flusher:
+			err := al.flush(al.ctx)
+			switch {
+			case errors.Is(err, context.Canceled):
+				// The logger was stopped, no need to retry.
+				return
+			case err != nil:
+				retryDelay = max(al.backoffOpts.min, min(retryDelay*time.Duration(al.backoffOpts.multiplier), al.backoffOpts.max))
+				al.logf("retrying after %v, %v", retryDelay, err)
+				retry.Reset(retryDelay)
+			default:
+				retryDelay = 0
+				retry.Stop()
+			}
+		case <-retry.C:
+			al.flushAsync()
+		}
+	}
+}
+
+// flush attempts to send all pending logs to the control plane.
+// l.mu must not be held.
+func (al *Logger) flush(ctx context.Context) error {
+	al.mu.Lock()
+	pending, err := al.store.load(al.profileID)
+	t := al.transport
+	al.mu.Unlock()
+
+	if err != nil {
+		// This will catch nil profileIDs
+		return fmt.Errorf("failed to restore pending logs: %w", err)
+	}
+	if len(pending) == 0 {
+		return nil
+	}
+	if t == nil {
+		return errors.New("no transport")
+	}
+
+	complete, unsent := al.sendToTransport(ctx, pending, t)
+	al.markTransactionsDone(complete)
+
+	al.mu.Lock()
+	defer al.mu.Unlock()
+	if err = al.appendToStoreLocked(unsent); err != nil {
+		al.logf("[unexpected] failed to persist logs: %v", err)
+	}
+
+	if len(unsent) != 0 {
+		return fmt.Errorf("failed to send %d logs", len(unsent))
+	}
+
+	if len(complete) != 0 {
+		al.logf("complete %d audit log transactions", len(complete))
+	}
+	return nil
+}
+
+// sendToTransport sends all pending logs to the control plane. Returns a pair of slices
+// containing the logs that were successfully sent (or failed permanently) and those that were not.
+//
+// This may require multiple round trips to the control plane and can be a long running transaction.
+func (al *Logger) sendToTransport(ctx context.Context, pending []*transaction, t Transport) (complete []*transaction, unsent []*transaction) {
+	for i, txn := range pending {
+		req := tailcfg.AuditLogRequest{
+			Action:    tailcfg.ClientAuditAction(txn.Action),
+			Details:   txn.Details,
+			Timestamp: txn.TimeStamp,
+		}
+
+		if err := t.SendAuditLog(ctx, req); err != nil {
+			switch {
+			case errors.Is(err, context.Canceled) || errors.Is(err, context.DeadlineExceeded):
+				// The contex is done.  All further attempts will fail.
+				unsent = append(unsent, pending[i:]...)
+				return complete, unsent
+			case IsRetryableError(err) && txn.Retries+1 < al.retryLimit:
+				// We permit a maximum number of retries for each log. All retriable
+				// errors should be transient and we should be able to send the log eventually, but
+				// we don't want logs to be persisted indefinitely.
+				txn.Retries++
+				unsent = append(unsent, txn)
+			default:
+				complete = append(complete, txn)
+				al.logf("failed permanently: %v", err)
+			}
+		} else {
+			// No error - we're done.
+			complete = append(complete, txn)
+		}
+	}
+
+	return complete, unsent
+}
+
+func (al *Logger) stop() {
+	al.mu.Lock()
+	t := al.transport
+	al.mu.Unlock()
+
+	if t == nil {
+		// No transport means no worker goroutine and done will not be
+		// closed if we cancel the context.
+		return
+	}
+
+	al.ctxCancel()
+	<-al.done
+	al.logf("stopped for profileID: %v", al.profileID)
+}
+
+// appendToStoreLocked persists logs to the store.  This will deduplicate
+// logs so it is safe to call this with the same logs multiple time, to
+// requeue failed transactions for example.
+//
+// l.mu must be held.
+func (al *Logger) appendToStoreLocked(txns []*transaction) error {
+	if len(txns) == 0 {
+		return nil
+	}
+
+	if al.profileID == "" {
+		return errors.New("no logId set")
+	}
+
+	persisted, err := al.store.load(al.profileID)
+	if err != nil {
+		al.logf("[unexpected] append failed to restore logs: %v", err)
+	}
+
+	// The order is important here.  We want the latest transactions first, which will
+	// ensure when we dedup, the new transactions are seen and the older transactions
+	// are discarded.
+	txnsOut := append(txns, persisted...)
+	txnsOut = deduplicateAndSort(txnsOut)
+
+	return al.store.save(al.profileID, txnsOut)
+}
+
+// storedCountLocked returns the number of logs persisted to the store.
+// al.mu must be held.
+func (al *Logger) storedCountLocked() (int, error) {
+	persisted, err := al.store.load(al.profileID)
+	return len(persisted), err
+}
+
+// markTransactionsDone removes logs from the store that are complete (sent or failed permanently).
+// al.mu must not be held.
+func (al *Logger) markTransactionsDone(sent []*transaction) {
+	al.mu.Lock()
+	defer al.mu.Unlock()
+
+	ids := set.Set[string]{}
+	for _, txn := range sent {
+		ids.Add(txn.EventID)
+	}
+
+	persisted, err := al.store.load(al.profileID)
+	if err != nil {
+		al.logf("[unexpected] markTransactionsDone failed to restore logs: %v", err)
+	}
+	var unsent []*transaction
+	for _, txn := range persisted {
+		if !ids.Contains(txn.EventID) {
+			unsent = append(unsent, txn)
+		}
+	}
+	al.store.save(al.profileID, unsent)
+}
+
+// deduplicateAndSort removes duplicate logs from the given slice and sorts them by timestamp.
+// The first log entry in the slice will be retained, subsequent logs with the same EventID will be discarded.
+func deduplicateAndSort(txns []*transaction) []*transaction {
+	seen := set.Set[string]{}
+	deduped := make([]*transaction, 0, len(txns))
+	for _, txn := range txns {
+		if !seen.Contains(txn.EventID) {
+			deduped = append(deduped, txn)
+			seen.Add(txn.EventID)
+		}
+	}
+	// Sort logs by timestamp - oldest to newest. This will put the oldest logs at
+	// the front of the queue.
+	sort.Slice(deduped, func(i, j int) bool {
+		return deduped[i].TimeStamp.Before(deduped[j].TimeStamp)
+	})
+	return deduped
+}
+
+func (al *Logger) enqueue(txn *transaction) error {
+	al.mu.Lock()
+	defer al.mu.Unlock()
+
+	if err := al.appendToStoreLocked([]*transaction{txn}); err != nil {
+		return fmt.Errorf("%w: %w", ErrAuditLogStorageFailure, err)
+	}
+
+	// If a.transport is nil if the logger is stopped.
+	if al.transport != nil {
+		al.flushAsync()
+	}
+
+	return nil
+}
+
+var _ LogStore = (*logStateStore)(nil)
+
+// logStateStore is a concrete implementation of [LogStore]
+// using [ipn.StateStore] as the underlying storage.
+type logStateStore struct {
+	store ipn.StateStore
+}
+
+// NewLogStore creates a new LogStateStore with the given [ipn.StateStore].
+func NewLogStore(store ipn.StateStore) LogStore {
+	return &logStateStore{
+		store: store,
+	}
+}
+
+func (s *logStateStore) generateKey(key ipn.ProfileID) string {
+	return "auditlog-" + string(key)
+}
+
+// Save saves the given logs to an [ipn.StateStore]. This overwrites
+// any existing entries for the given key.
+func (s *logStateStore) save(key ipn.ProfileID, txns []*transaction) error {
+	if key == "" {
+		return errors.New("empty key")
+	}
+
+	data, err := json.Marshal(txns)
+	if err != nil {
+		return err
+	}
+	k := ipn.StateKey(s.generateKey(key))
+	return s.store.WriteState(k, data)
+}
+
+// Load retrieves the logs from an [ipn.StateStore].
+func (s *logStateStore) load(key ipn.ProfileID) ([]*transaction, error) {
+	if key == "" {
+		return nil, errors.New("empty key")
+	}
+
+	k := ipn.StateKey(s.generateKey(key))
+	data, err := s.store.ReadState(k)
+
+	switch {
+	case errors.Is(err, ipn.ErrStateNotExist):
+		return nil, nil
+	case err != nil:
+		return nil, err
+	}
+
+	var txns []*transaction
+	err = json.Unmarshal(data, &txns)
+	return txns, err
+}

ipn/auditlog/auditlog_test.go

@@ -0,0 +1,481 @@
+// Copyright (c) Tailscale Inc & AUTHORS
+// SPDX-License-Identifier: BSD-3-Clause
+
+package auditlog
+
+import (
+	"context"
+	"errors"
+	"fmt"
+	"sync"
+	"testing"
+	"time"
+
+	qt "github.com/frankban/quicktest"
+	"tailscale.com/ipn/store/mem"
+	"tailscale.com/tailcfg"
+	"tailscale.com/tstest"
+)
+
+// loggerForTest creates an auditLogger for you and cleans it up
+// (and ensures no goroutines are leaked) when the test is done.
+func loggerForTest(t *testing.T, opts Opts) *Logger {
+	t.Helper()
+	tstest.ResourceCheck(t)
+
+	if opts.Logf == nil {
+		opts.Logf = t.Logf
+	}
+
+	if opts.Store == nil {
+		t.Fatalf("opts.Store must be set")
+	}
+
+	a := NewLogger(opts)
+
+	t.Cleanup(func() {
+		a.FlushAndStop(context.Background())
+	})
+	return a
+}
+
+func TestNonRetryableErrors(t *testing.T) {
+	errorTests := []struct {
+		desc string
+		err  error
+		want bool
+	}{
+		{"DeadlineExceeded", context.DeadlineExceeded, false},
+		{"Canceled", context.Canceled, false},
+		{"Canceled wrapped", fmt.Errorf("%w: %w", context.Canceled, errors.New("ctx cancelled")), false},
+		{"Random error", errors.New("random error"), false},
+	}
+
+	for _, tt := range errorTests {
+		t.Run(tt.desc, func(t *testing.T) {
+			if IsRetryableError(tt.err) != tt.want {
+				t.Fatalf("retriable: got %v, want %v", !tt.want, tt.want)
+			}
+		})
+	}
+}
+
+// TestEnqueueAndFlush enqueues n logs and flushes them.
+// We expect all logs to be flushed and for no
+// logs to remain in the store once FlushAndStop returns.
+func TestEnqueueAndFlush(t *testing.T) {
+	c := qt.New(t)
+	mockTransport := newMockTransport(nil)
+	al := loggerForTest(t, Opts{
+		RetryLimit: 200,
+		Logf:       t.Logf,
+		Store:      NewLogStore(&mem.Store{}),
+	})
+
+	c.Assert(al.SetProfileID("test"), qt.IsNil)
+	c.Assert(al.Start(mockTransport), qt.IsNil)
+
+	wantSent := 10
+
+	for i := range wantSent {
+		err := al.Enqueue(tailcfg.AuditNodeDisconnect, fmt.Sprintf("log %d", i))
+		c.Assert(err, qt.IsNil)
+	}
+
+	al.FlushAndStop(context.Background())
+
+	al.mu.Lock()
+	defer al.mu.Unlock()
+	gotStored, err := al.storedCountLocked()
+	c.Assert(err, qt.IsNil)
+
+	if wantStored := 0; gotStored != wantStored {
+		t.Fatalf("stored: got %d, want %d", gotStored, wantStored)
+	}
+
+	if gotSent := mockTransport.sentCount(); gotSent != wantSent {
+		t.Fatalf("sent: got %d, want %d", gotSent, wantSent)
+	}
+}
+
+// TestEnqueueAndFlushWithFlushCancel calls FlushAndCancel with a cancelled
+// context.  We expect nothing to be sent and all logs to be stored.
+func TestEnqueueAndFlushWithFlushCancel(t *testing.T) {
+	c := qt.New(t)
+	mockTransport := newMockTransport(&retriableError)
+	al := loggerForTest(t, Opts{
+		RetryLimit: 200,
+		Logf:       t.Logf,
+		Store:      NewLogStore(&mem.Store{}),
+	})
+
+	c.Assert(al.SetProfileID("test"), qt.IsNil)
+	c.Assert(al.Start(mockTransport), qt.IsNil)
+
+	for i := range 10 {
+		err := al.Enqueue(tailcfg.AuditNodeDisconnect, fmt.Sprintf("log %d", i))
+		c.Assert(err, qt.IsNil)
+	}
+
+	// Cancel the context before calling FlushAndStop - nothing should get sent.
+	// This mimics a timeout before flush() has a chance to execute.
+	ctx, cancel := context.WithCancel(context.Background())
+	cancel()
+
+	al.FlushAndStop(ctx)
+
+	al.mu.Lock()
+	defer al.mu.Unlock()
+	gotStored, err := al.storedCountLocked()
+	c.Assert(err, qt.IsNil)
+
+	if wantStored := 10; gotStored != wantStored {
+		t.Fatalf("stored: got %d, want %d", gotStored, wantStored)
+	}
+
+	if gotSent, wantSent := mockTransport.sentCount(), 0; gotSent != wantSent {
+		t.Fatalf("sent: got %d, want %d", gotSent, wantSent)
+	}
+}
+
+// TestDeduplicateAndSort tests that the most recent log is kept when deduplicating logs
+func TestDeduplicateAndSort(t *testing.T) {
+	c := qt.New(t)
+	al := loggerForTest(t, Opts{
+		RetryLimit: 100,
+		Logf:       t.Logf,
+		Store:      NewLogStore(&mem.Store{}),
+	})
+
+	c.Assert(al.SetProfileID("test"), qt.IsNil)
+
+	logs := []*transaction{
+		{EventID: "1", Details: "log 1", TimeStamp: time.Now().Add(-time.Minute * 1), Retries: 1},
+	}
+
+	al.mu.Lock()
+	defer al.mu.Unlock()
+	al.appendToStoreLocked(logs)
+
+	// Update the transaction and re-append it
+	logs[0].Retries = 2
+	al.appendToStoreLocked(logs)
+
+	fromStore, err := al.store.load("test")
+	c.Assert(err, qt.IsNil)
+
+	// We should see only one transaction
+	if wantStored, gotStored := len(logs), len(fromStore); gotStored != wantStored {
+		t.Fatalf("stored: got %d, want %d", gotStored, wantStored)
+	}
+
+	// We should see the latest transaction
+	if wantRetryCount, gotRetryCount := 2, fromStore[0].Retries; gotRetryCount != wantRetryCount {
+		t.Fatalf("reties: got %d, want %d", gotRetryCount, wantRetryCount)
+	}
+}
+
+func TestChangeProfileId(t *testing.T) {
+	c := qt.New(t)
+	al := loggerForTest(t, Opts{
+		RetryLimit: 100,
+		Logf:       t.Logf,
+		Store:      NewLogStore(&mem.Store{}),
+	})
+	c.Assert(al.SetProfileID("test"), qt.IsNil)
+
+	// Changing a profile ID must fail
+	c.Assert(al.SetProfileID("test"), qt.IsNotNil)
+}
+
+// TestSendOnRestore pushes a n logs to the persistent store, and ensures they
+// are sent as soon as Start is called then checks to ensure the sent logs no
+// longer exist in the store.
+func TestSendOnRestore(t *testing.T) {
+	c := qt.New(t)
+	mockTransport := newMockTransport(nil)
+	al := loggerForTest(t, Opts{
+		RetryLimit: 100,
+		Logf:       t.Logf,
+		Store:      NewLogStore(&mem.Store{}),
+	})
+	al.SetProfileID("test")
+
+	wantTotal := 10
+
+	for range 10 {
+		al.Enqueue(tailcfg.AuditNodeDisconnect, "log")
+	}
+
+	c.Assert(al.Start(mockTransport), qt.IsNil)
+
+	al.FlushAndStop(context.Background())
+
+	al.mu.Lock()
+	defer al.mu.Unlock()
+	gotStored, err := al.storedCountLocked()
+	c.Assert(err, qt.IsNil)
+
+	if wantStored := 0; gotStored != wantStored {
+		t.Fatalf("stored: got %d, want %d", gotStored, wantStored)
+	}
+
+	if gotSent, wantSent := mockTransport.sentCount(), wantTotal; gotSent != wantSent {
+		t.Fatalf("sent: got %d, want %d", gotSent, wantSent)
+	}
+}
+
+// TestFailureExhaustion enqueues n logs,  with the transport in a failable state.
+// We then set it to a non-failing state, call FlushAndStop and expect all logs to be sent.
+func TestFailureExhaustion(t *testing.T) {
+	c := qt.New(t)
+	mockTransport := newMockTransport(&retriableError)
+
+	al := loggerForTest(t, Opts{
+		RetryLimit: 1,
+		Logf:       t.Logf,
+		Store:      NewLogStore(&mem.Store{}),
+	})
+
+	c.Assert(al.SetProfileID("test"), qt.IsNil)
+	c.Assert(al.Start(mockTransport), qt.IsNil)
+
+	for range 10 {
+		err := al.Enqueue(tailcfg.AuditNodeDisconnect, "log")
+		c.Assert(err, qt.IsNil)
+	}
+
+	al.FlushAndStop(context.Background())
+	al.mu.Lock()
+	defer al.mu.Unlock()
+	gotStored, err := al.storedCountLocked()
+	c.Assert(err, qt.IsNil)
+
+	if wantStored := 0; gotStored != wantStored {
+		t.Fatalf("stored: got %d, want %d", gotStored, wantStored)
+	}
+
+	if gotSent, wantSent := mockTransport.sentCount(), 0; gotSent != wantSent {
+		t.Fatalf("sent: got %d, want %d", gotSent, wantSent)
+	}
+}
+
+// TestEnqueueAndFailNoRetry enqueues a set of logs, all of which will fail and are not
+// retriable. We then call FlushAndStop and expect all to be unsent.
+func TestEnqueueAndFailNoRetry(t *testing.T) {
+	c := qt.New(t)
+	mockTransport := newMockTransport(&nonRetriableError)
+
+	al := loggerForTest(t, Opts{
+		RetryLimit: 100,
+		Logf:       t.Logf,
+		Store:      NewLogStore(&mem.Store{}),
+	})
+
+	c.Assert(al.SetProfileID("test"), qt.IsNil)
+	c.Assert(al.Start(mockTransport), qt.IsNil)
+
+	for i := range 10 {
+		err := al.Enqueue(tailcfg.AuditNodeDisconnect, fmt.Sprintf("log %d", i))
+		c.Assert(err, qt.IsNil)
+	}
+
+	al.FlushAndStop(context.Background())
+	al.mu.Lock()
+	defer al.mu.Unlock()
+	gotStored, err := al.storedCountLocked()
+	c.Assert(err, qt.IsNil)
+
+	if wantStored := 0; gotStored != wantStored {
+		t.Fatalf("stored: got %d, want %d", gotStored, wantStored)
+	}
+
+	if gotSent, wantSent := mockTransport.sentCount(), 0; gotSent != wantSent {
+		t.Fatalf("sent: got %d, want %d", gotSent, wantSent)
+	}
+}
+
+// TestEnqueueAndRetry enqueues a set of logs, all of which will fail and are retriable.
+// Mid-test, we set the transport to not-fail and expect the queue to flush properly
+// We set the backoff parameters to 0 seconds so retries are immediate.
+func TestEnqueueAndRetry(t *testing.T) {
+	c := qt.New(t)
+	mockTransport := newMockTransport(&retriableError)
+
+	al := loggerForTest(t, Opts{
+		RetryLimit: 100,
+		Logf:       t.Logf,
+		Store:      NewLogStore(&mem.Store{}),
+	})
+
+	al.backoffOpts = backoffOpts{
+		min:        1 * time.Millisecond,
+		max:        4 * time.Millisecond,
+		multiplier: 2.0,
+	}
+
+	c.Assert(al.SetProfileID("test"), qt.IsNil)
+	c.Assert(al.Start(mockTransport), qt.IsNil)
+
+	err := al.Enqueue(tailcfg.AuditNodeDisconnect, fmt.Sprintf("log 1"))
+	c.Assert(err, qt.IsNil)
+
+	// This will wait for at least 2 retries
+	gotRetried, wantRetried := mockTransport.waitForSendAttemptsToReach(3), true
+	if gotRetried != wantRetried {
+		t.Fatalf("retried: got %v, want %v", gotRetried, wantRetried)
+	}
+
+	mockTransport.setErrorCondition(nil)
+
+	al.FlushAndStop(context.Background())
+	al.mu.Lock()
+	defer al.mu.Unlock()
+
+	gotStored, err := al.storedCountLocked()
+	c.Assert(err, qt.IsNil)
+
+	if wantStored := 0; gotStored != wantStored {
+		t.Fatalf("stored: got %d, want %d", gotStored, wantStored)
+	}
+
+	if gotSent, wantSent := mockTransport.sentCount(), 1; gotSent != wantSent {
+		t.Fatalf("sent: got %d, want %d", gotSent, wantSent)
+	}
+}
+
+// TestEnqueueBeforeSetProfileID tests that logs enqueued before SetProfileId are not sent
+func TestEnqueueBeforeSetProfileID(t *testing.T) {
+	c := qt.New(t)
+	al := loggerForTest(t, Opts{
+		RetryLimit: 100,
+		Logf:       t.Logf,
+		Store:      NewLogStore(&mem.Store{}),
+	})
+
+	err := al.Enqueue(tailcfg.AuditNodeDisconnect, "log")
+	c.Assert(err, qt.IsNotNil)
+	al.FlushAndStop(context.Background())
+
+	al.mu.Lock()
+	defer al.mu.Unlock()
+	gotStored, err := al.storedCountLocked()
+	c.Assert(err, qt.IsNotNil)
+
+	if wantStored := 0; gotStored != wantStored {
+		t.Fatalf("stored: got %d, want %d", gotStored, wantStored)
+	}
+}
+
+// TestLogStoring tests that audit logs are persisted sorted by timestamp, oldest to newest
+func TestLogSorting(t *testing.T) {
+	c := qt.New(t)
+	mockStore := NewLogStore(&mem.Store{})
+
+	logs := []*transaction{
+		{EventID: "1", Details: "log 3", TimeStamp: time.Now().Add(-time.Minute * 1)},
+		{EventID: "1", Details: "log 3", TimeStamp: time.Now().Add(-time.Minute * 2)},
+		{EventID: "2", Details: "log 2", TimeStamp: time.Now().Add(-time.Minute * 3)},
+		{EventID: "3", Details: "log 1", TimeStamp: time.Now().Add(-time.Minute * 4)},
+	}
+
+	wantLogs := []transaction{
+		{Details: "log 1"},
+		{Details: "log 2"},
+		{Details: "log 3"},
+	}
+
+	mockStore.save("test", logs)
+
+	gotLogs, err := mockStore.load("test")
+	c.Assert(err, qt.IsNil)
+	gotLogs = deduplicateAndSort(gotLogs)
+
+	for i := range gotLogs {
+		if want, got := wantLogs[i].Details, gotLogs[i].Details; want != got {
+			t.Fatalf("Details: got %v, want %v", got, want)
+		}
+	}
+}
+
+// mock implementations for testing
+
+// newMockTransport returns a mock transport for testing
+// If err is no nil, SendAuditLog will return this error if the send is attempted
+// before the context is cancelled.
+func newMockTransport(err error) *mockAuditLogTransport {
+	return &mockAuditLogTransport{
+		err:      err,
+		attempts: make(chan int, 1),
+	}
+}
+
+type mockAuditLogTransport struct {
+	attempts chan int // channel to notify of send attempts
+
+	mu          sync.Mutex
+	sendAttmpts int   // number of attempts to send logs
+	sendCount   int   // number of logs sent by the transport
+	err         error // error to return when sending logs
+}
+
+// waitForSendAttemptsToReach blocks until the number of send attempts reaches n
+// This should be use only in tests where the transport is expected to retry sending logs
+func (t *mockAuditLogTransport) waitForSendAttemptsToReach(n int) bool {
+	for attempts := range t.attempts {
+		if attempts >= n {
+			return true
+		}
+	}
+	return false
+}
+
+func (t *mockAuditLogTransport) setErrorCondition(err error) {
+	t.mu.Lock()
+	defer t.mu.Unlock()
+	t.err = err
+}
+
+func (t *mockAuditLogTransport) sentCount() int {
+	t.mu.Lock()
+	defer t.mu.Unlock()
+	return t.sendCount
+}
+
+func (t *mockAuditLogTransport) SendAuditLog(ctx context.Context, _ tailcfg.AuditLogRequest) (err error) {
+	t.mu.Lock()
+	t.sendAttmpts += 1
+	defer func() {
+		a := t.sendAttmpts
+		t.mu.Unlock()
+		select {
+		case t.attempts <- a:
+		default:
+		}
+	}()
+
+	select {
+	case <-ctx.Done():
+		return ctx.Err()
+	default:
+	}
+
+	if t.err != nil {
+		return t.err
+	}
+	t.sendCount += 1
+	return nil
+}
+
+var (
+	retriableError    = mockError{errors.New("retriable error")}
+	nonRetriableError = mockError{errors.New("permanent failure error")}
+)
+
+type mockError struct {
+	error
+}
+
+func (e mockError) Retryable() bool {
+	return e == retriableError
+}

ipn/ipnauth/actor.go

@@ -10,12 +10,11 @@ import (
 
 	"tailscale.com/client/tailscale/apitype"
 	"tailscale.com/ipn"
+	"tailscale.com/tailcfg"
 )
 
 // AuditLogFunc is any function that can be used to log audit actions performed by an [Actor].
-//
-// TODO(nickkhyl,barnstar): define a named string type for the action (in tailcfg?) and use it here.
-type AuditLogFunc func(action, details string)
+type AuditLogFunc func(action tailcfg.ClientAuditAction, details string) error
 
 // Actor is any actor using the [ipnlocal.LocalBackend].
 //
@@ -45,7 +44,7 @@ type Actor interface {
 	//
 	// If the auditLogger is non-nil, it is used to write details about the action
 	// to the audit log when required by the policy.
-	CheckProfileAccess(profile ipn.LoginProfileView, requestedAccess ProfileAccess, auditLogger AuditLogFunc) error
+	CheckProfileAccess(profile ipn.LoginProfileView, requestedAccess ProfileAccess, auditLogFn AuditLogFunc) error
 
 	// IsLocalSystem reports whether the actor is the Windows' Local System account.
 	//

ipn/ipnauth/policy.go

@@ -9,6 +9,7 @@ import (
 
 	"tailscale.com/client/tailscale/apitype"
 	"tailscale.com/ipn"
+	"tailscale.com/tailcfg"
 	"tailscale.com/util/syspolicy"
 )
 
@@ -48,7 +49,7 @@ func (a actorWithPolicyChecks) CheckProfileAccess(profile ipn.LoginProfileView,
 //
 // TODO(nickkhyl): unexport it when we move [ipn.Actor] implementations from [ipnserver]
 // and corp to this package.
-func CheckDisconnectPolicy(actor Actor, profile ipn.LoginProfileView, reason string, auditLogger AuditLogFunc) error {
+func CheckDisconnectPolicy(actor Actor, profile ipn.LoginProfileView, reason string, auditFn AuditLogFunc) error {
 	if alwaysOn, _ := syspolicy.GetBoolean(syspolicy.AlwaysOn, false); !alwaysOn {
 		return nil
 	}
@@ -58,15 +59,16 @@ func CheckDisconnectPolicy(actor Actor, profile ipn.LoginProfileView, reason str
 	if reason == "" {
 		return errors.New("disconnect not allowed: reason required")
 	}
-	if auditLogger != nil {
+	if auditFn != nil {
 		var details string
 		if username, _ := actor.Username(); username != "" { // best-effort; we don't have it on all platforms
 			details = fmt.Sprintf("%q is being disconnected by %q: %v", profile.Name(), username, reason)
 		} else {
 			details = fmt.Sprintf("%q is being disconnected: %v", profile.Name(), reason)
 		}
-		// TODO(nickkhyl,barnstar): use a const for DISCONNECT_NODE.
-		auditLogger("DISCONNECT_NODE", details)
+		if err := auditFn(tailcfg.AuditNodeDisconnect, details); err != nil {
+			return err
+		}
 	}
 	return nil
 }

ipn/ipnlocal/local.go

@@ -57,10 +57,12 @@ import (
 	"tailscale.com/health/healthmsg"
 	"tailscale.com/hostinfo"
 	"tailscale.com/ipn"
+	"tailscale.com/ipn/auditlog"
 	"tailscale.com/ipn/conffile"
 	"tailscale.com/ipn/ipnauth"
 	"tailscale.com/ipn/ipnstate"
 	"tailscale.com/ipn/policy"
+	memstore "tailscale.com/ipn/store/mem"
 	"tailscale.com/log/sockstatlog"
 	"tailscale.com/logpolicy"
 	"tailscale.com/net/captivedetection"
@@ -450,6 +452,12 @@ type LocalBackend struct {
 	// Each callback is called exactly once in unspecified order and without b.mu held.
 	// Returned errors are logged but otherwise ignored and do not affect the shutdown process.
 	shutdownCbs set.HandleSet[func() error]
+
+	// auditLogger, if non-nil, manages audit logging for the backend.
+	//
+	// It queues, persists, and sends audit logs
+	// to the control client.  auditLogger has the same lifespan as b.cc.
+	auditLogger *auditlog.Logger
 }
 
 // HealthTracker returns the health tracker for the backend.
@@ -1679,6 +1687,15 @@ func (b *LocalBackend) SetControlClientStatus(c controlclient.Client, st control
 			b.logf("Failed to save new controlclient state: %v", err)
 		}
 	}
+
+	// Update the audit logger with the current profile ID.
+	if b.auditLogger != nil && prefsChanged {
+		pid := b.pm.CurrentProfile().ID()
+		if err := b.auditLogger.SetProfileID(pid); err != nil {
+			b.logf("Failed to set profile ID in audit logger: %v", err)
+		}
+	}
+
 	// initTKALocked is dependent on CurrentProfile.ID, which is initialized
 	// (for new profiles) on the first call to b.pm.SetPrefs.
 	if err := b.initTKALocked(); err != nil {
@@ -2386,6 +2403,27 @@ func (b *LocalBackend) Start(opts ipn.Options) error {
 		debugFlags = append([]string{"netstack"}, debugFlags...)
 	}
 
+	var auditLogShutdown func()
+	// Audit logging is only available if the client has set up a proper persistent
+	// store for the logs in sys.
+	store, ok := b.sys.AuditLogStore.GetOK()
+	if !ok {
+		b.logf("auditlog: [unexpected] no persistent audit log storage configured.  using memory store.")
+		store = auditlog.NewLogStore(&memstore.Store{})
+	}
+
+	al := auditlog.NewLogger(auditlog.Opts{
+		Logf:       b.logf,
+		RetryLimit: 32,
+		Store:      store,
+	})
+	b.auditLogger = al
+	auditLogShutdown = func() {
+		ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+		defer cancel()
+		al.FlushAndStop(ctx)
+	}
+
 	// TODO(apenwarr): The only way to change the ServerURL is to
 	// re-run b.Start, because this is the only place we create a
 	// new controlclient. EditPrefs allows you to overwrite ServerURL,
@@ -2411,6 +2449,7 @@ func (b *LocalBackend) Start(opts ipn.Options) error {
 		C2NHandler:                 http.HandlerFunc(b.handleC2N),
 		DialPlan:                   &b.dialPlan, // pointer because it can't be copied
 		ControlKnobs:               b.sys.ControlKnobs(),
+		Shutdown:                   auditLogShutdown,
 
 		// Don't warn about broken Linux IP forwarding when
 		// netstack is being used.
@@ -4263,6 +4302,21 @@ func (b *LocalBackend) MaybeClearAppConnector(mp *ipn.MaskedPrefs) error {
 	return err
 }
 
+var errNoAuditLogger = errors.New("no audit logger configured")
+
+func (b *LocalBackend) getAuditLoggerLocked() ipnauth.AuditLogFunc {
+	logger := b.auditLogger
+	return func(action tailcfg.ClientAuditAction, details string) error {
+		if logger == nil {
+			return errNoAuditLogger
+		}
+		if err := logger.Enqueue(action, details); err != nil {
+			return fmt.Errorf("failed to enqueue audit log %v %q: %w", action, details, err)
+		}
+		return nil
+	}
+}
+
 // EditPrefs applies the changes in mp to the current prefs,
 // acting as the tailscaled itself rather than a specific user.
 func (b *LocalBackend) EditPrefs(mp *ipn.MaskedPrefs) (ipn.PrefsView, error) {
@@ -4288,9 +4342,8 @@ func (b *LocalBackend) EditPrefsAs(mp *ipn.MaskedPrefs, actor ipnauth.Actor) (ip
 	unlock := b.lockAndGetUnlock()
 	defer unlock()
 	if mp.WantRunningSet && !mp.WantRunning && b.pm.CurrentPrefs().WantRunning() {
-		// TODO(barnstar,nickkhyl): replace loggerFn with the actual audit logger.
-		loggerFn := func(action, details string) { b.logf("[audit]: %s: %s", action, details) }
-		if err := actor.CheckProfileAccess(b.pm.CurrentProfile(), ipnauth.Disconnect, loggerFn); err != nil {
+		if err := actor.CheckProfileAccess(b.pm.CurrentProfile(), ipnauth.Disconnect, b.getAuditLoggerLocked()); err != nil {
+			b.logf("check profile access failed: %v", err)
 			return ipn.PrefsView{}, err
 		}
 
@@ -5874,6 +5927,15 @@ func (b *LocalBackend) requestEngineStatusAndWait() {
 func (b *LocalBackend) setControlClientLocked(cc controlclient.Client) {
 	b.cc = cc
 	b.ccAuto, _ = cc.(*controlclient.Auto)
+	if b.auditLogger != nil {
+		if err := b.auditLogger.SetProfileID(b.pm.CurrentProfile().ID()); err != nil {
+			b.logf("audit logger set profile ID failure: %v", err)
+		}
+
+		if err := b.auditLogger.Start(b.ccAuto); err != nil {
+			b.logf("audit logger start failure: %v", err)
+		}
+	}
 }
 
 // resetControlClientLocked sets b.cc to nil and returns the old value. If the

tailcfg/tailcfg.go

@@ -2982,3 +2982,33 @@ const LBHeader = "Ts-Lb"
 // correspond to those IPs. Any services that don't correspond to a service
 // this client is hosting can be ignored.
 type ServiceIPMappings map[ServiceName][]netip.Addr
+
+// ClientAuditAction represents an auditable action that a client can report to the
+// control plane.  These actions must correspond to the supported actions
+// in the control plane.
+type ClientAuditAction string
+
+const (
+	// AuditNodeDisconnect action is sent when a node has disconnected
+	// from the control plane.  The details must include a reason in the Details
+	// field, either generated, or entered by the user.
+	AuditNodeDisconnect = ClientAuditAction("DISCONNECT_NODE")
+)
+
+// AuditLogRequest represents an audit log request to be sent to the control plane.
+//
+// This is JSON-encoded and sent over the control plane connection to:
+// POST https://<control-plane>/machine/audit-log
+type AuditLogRequest struct {
+	// Version is the client's current CapabilityVersion.
+	Version CapabilityVersion `json:",omitempty"`
+	// NodeKey is the client's current node key.
+	NodeKey key.NodePublic `json:",omitzero"`
+	// Action is the action to be logged. It must correspond to a known action in the control plane.
+	Action ClientAuditAction `json:",omitempty"`
+	// Details is an opaque string, specific to the action being logged.  Empty strings may not
+	// be valid depending on the action being logged.
+	Details string `json:",omitempty"`
+	// Timestamp is the time at which the audit log was generated on the node.
+	Timestamp time.Time `json:",omitzero"`
+}

tsd/tsd.go

@@ -25,6 +25,7 @@ import (
 	"tailscale.com/drive"
 	"tailscale.com/health"
 	"tailscale.com/ipn"
+	"tailscale.com/ipn/auditlog"
 	"tailscale.com/ipn/conffile"
 	"tailscale.com/ipn/desktop"
 	"tailscale.com/net/dns"
@@ -50,6 +51,7 @@ type System struct {
 	Router         SubSystem[router.Router]
 	Tun            SubSystem[*tstun.Wrapper]
 	StateStore     SubSystem[ipn.StateStore]
+	AuditLogStore  SubSystem[auditlog.LogStore]
 	Netstack       SubSystem[NetstackImpl] // actually a *netstack.Impl
 	DriveForLocal  SubSystem[drive.FileSystemForLocal]
 	DriveForRemote SubSystem[drive.FileSystemForRemote]
@@ -106,6 +108,8 @@ func (s *System) Set(v any) {
 		s.MagicSock.Set(v)
 	case ipn.StateStore:
 		s.StateStore.Set(v)
+	case auditlog.LogStore:
+		s.AuditLogStore.Set(v)
 	case NetstackImpl:
 		s.Netstack.Set(v)
 	case drive.FileSystemForLocal: