control/controlclient: make mapSession not run in a separate goroutine (#21202)

Prior to 85bb5f8 mapSession was running directly as part of the
controlClient. That was changed previously to let discoKeys to flow into
the mapSession for TSMP learned keys, using the mapSession for filtering
of keys.

Since moving the TSMP key learning directly into the userspace engine,
and introducing the notion of two active disco keys for an enpoint, we
no longer need mapSession to run in a separate goroutine and have
separate paths of entry for data.

This commit effectively reverts the last changes that was previously
made in the controlClient to support TSMP.

Updates #20590

Signed-off-by: Claus Lensbøl <[email protected]>
This commit is contained in:
Claus Lensbøl authored and GitHub committed 2026-09-11 09:09:51 -04:00
1 parent 91bd4f99d2
commit 4b60ec876f
3 files changed
+13 -99

No files matched your search

+5 -20
View File
@@ -115,9 +115,8 @@ type Direct struct {
netinfo *tailcfg.NetInfo
endpoints []tailcfg.Endpoint
tkaHead string
lastPingURL string // last PingRequest.URL received, for dup suppression
connectionHandleForTest string // sent in MapRequest.ConnectionHandleForTest
streamingMapSession *mapSession // the one streaming mapSession instance
lastPingURL string // last PingRequest.URL received, for dup suppression
connectionHandleForTest string // sent in MapRequest.ConnectionHandleForTest
controlClientID int64 // Random ID used to differentiate clients for consumers of messages.
}
@@ -929,8 +928,8 @@ func (c *Direct) PollNetMap(ctx context.Context, nu NetmapUpdater) error {
// update it observed. It is used by tests and [NetmapFromMapResponseForDebug].
// It will report only the first netmap seen.
type rememberLastNetmapUpdater struct {
last *netmap.NetworkMap
done chan any
last *netmap.NetworkMap
done chan any
}
func (nu *rememberLastNetmapUpdater) UpdateFullNetmap(nm *netmap.NetworkMap) {
@@ -1197,22 +1196,8 @@ func (c *Direct) sendMapRequest(ctx context.Context, isStreaming bool, nu Netmap
return nil
}
if isStreaming && c.streamingMapSession != nil {
panic("mapSession is already set")
}
sess := newMapSession(persist.PrivateNodeKey(), nu, c.controlKnobs)
if isStreaming {
c.streamingMapSession = sess
defer func() {
sess.Close()
c.mu.Lock()
c.streamingMapSession = nil
c.mu.Unlock()
}()
} else {
defer sess.Close()
}
defer sess.Close()
sess.cancel = cancel
sess.logf = c.logf
sess.vlogf = vlogf
+3 -74
View File
@@ -98,10 +98,6 @@ type mapSession struct {
lastPopBrowserURL string
lastTKAInfo *tailcfg.TKAInfo
lastNetmapSummary string // from NetworkMap.VeryConcise
cqmu sync.Mutex
changeQueue chan *tailcfg.MapResponse
changeQueueClosed bool
processQueue sync.WaitGroup
// mu protects the peers map.
peersMu sync.RWMutex
@@ -128,48 +124,11 @@ func newMapSession(privateNodeKey key.NodePrivate, nu NetmapUpdater, controlKnob
cancel: func() {},
onDebug: func(context.Context, *tailcfg.Debug) error { return nil },
onSelfNodeChanged: func(*netmap.NetworkMap) {},
changeQueue: make(chan *tailcfg.MapResponse),
changeQueueClosed: false,
}
ms.sessionAliveCtx, ms.sessionAliveCtxClose = context.WithCancel(context.Background())
ms.processQueue.Add(1)
go ms.run()
return ms
}
// run starts the mapSession processing a queue of tailcfg.MapResponse one by
// one until close() is called on the mapSession.
// When the mapSession is closed, the remaining queue is locked and processed
// before the mapSession is done processing.
func (ms *mapSession) run() {
defer ms.processQueue.Done()
for {
select {
case change := <-ms.changeQueue:
ms.handleNonKeepAliveMapResponse(ms.sessionAliveCtx, change)
case <-ms.sessionAliveCtx.Done():
// Drain any remaining items in the queue before exiting.
// Lock the queue during this time to avoid updates through other channels
// to be overwritten. This is especially relevant for calls to
// updateDiscoForNode.
ms.cqmu.Lock()
ms.changeQueueClosed = true
ms.cqmu.Unlock()
for {
select {
case change := <-ms.changeQueue:
ms.handleNonKeepAliveMapResponse(ms.sessionAliveCtx, change)
default:
// Queue is empty, close it and exit
close(ms.changeQueue)
return
}
}
}
}
}
// occasionallyPrintSummary logs summary at most once very 5 minutes. The
// summary is the Netmap.VeryConcise result from the last received map response.
func (ms *mapSession) occasionallyPrintSummary(summary string) {
@@ -190,7 +149,6 @@ func (ms *mapSession) clock() tstime.Clock {
func (ms *mapSession) Close() {
ms.sessionAliveCtxClose()
ms.processQueue.Wait()
}
var ErrChangeQueueClosed = errors.New("change queue closed")
@@ -203,44 +161,15 @@ var ErrChangeQueueClosed = errors.New("change queue closed")
//
// Debug messages are handled first, followed by pushing the response onto a
// queue for new updates handled sequentially.
//
// TODO(bradfitz): make this handle all fields later. For now (2023-08-20) this
// is [re]factoring progress enough.
func (ms *mapSession) HandleNonKeepAliveMapResponse(ctx context.Context, resp *tailcfg.MapResponse) error {
if debug := resp.Debug; debug != nil {
if err := ms.onDebug(ctx, debug); err != nil {
return err
}
}
ms.cqmu.Lock()
if ms.changeQueueClosed {
ms.cqmu.Unlock()
ms.processQueue.Wait()
return ErrChangeQueueClosed
}
defer ms.cqmu.Unlock()
return ms.addRespToQueue(resp)
}
func (ms *mapSession) addRespToQueue(resp *tailcfg.MapResponse) error {
select {
case ms.changeQueue <- resp:
return nil
case <-ms.sessionAliveCtx.Done():
return ErrChangeQueueClosed
}
}
// handleNonKeepAliveMapResponse handles a non-KeepAlive MapResponse (full or
// incremental).
//
// All fields that are valid on a KeepAlive MapResponse have already been
// handled.
//
// TODO(bradfitz): make this handle all fields later. For now (2023-08-20) this
// is [re]factoring progress enough.
func (ms *mapSession) handleNonKeepAliveMapResponse(ctx context.Context, resp *tailcfg.MapResponse) error {
if DevKnob.StripEndpoints() {
for _, p := range resp.Peers {
p.Endpoints = nil
+5 -5
View File
@@ -931,7 +931,7 @@ func TestExistingPeerReplacementHandledIncrementally(t *testing.T) {
AllowedIPs: []netip.Prefix{netip.MustParsePrefix("100.64.0.1/32")},
Hostinfo: (&tailcfg.Hostinfo{}).View(),
}
if err := ms.handleNonKeepAliveMapResponse(ctx, &tailcfg.MapResponse{
if err := ms.HandleNonKeepAliveMapResponse(ctx, &tailcfg.MapResponse{
Node: &tailcfg.Node{Name: "self.example.ts.net."},
Peers: []*tailcfg.Node{peer},
}); err != nil {
@@ -943,7 +943,7 @@ func TestExistingPeerReplacementHandledIncrementally(t *testing.T) {
replacement := peer.Clone()
replacement.AllowedIPs = append(replacement.AllowedIPs, netip.MustParsePrefix("100.64.0.2/32"))
if err := ms.handleNonKeepAliveMapResponse(ctx, &tailcfg.MapResponse{
if err := ms.HandleNonKeepAliveMapResponse(ctx, &tailcfg.MapResponse{
PeersChanged: []*tailcfg.Node{replacement},
}); err != nil {
t.Fatal(err)
@@ -995,7 +995,7 @@ func TestUpsertReplaysUserProfiles(t *testing.T) {
AllowedIPs: []netip.Prefix{netip.MustParsePrefix("100.64.0.1/32")},
Hostinfo: (&tailcfg.Hostinfo{}).View(),
}
if err := ms.handleNonKeepAliveMapResponse(ctx, &tailcfg.MapResponse{
if err := ms.HandleNonKeepAliveMapResponse(ctx, &tailcfg.MapResponse{
Node: &tailcfg.Node{Name: "self.example.ts.net."},
Peers: []*tailcfg.Node{peer},
UserProfiles: []tailcfg.UserProfile{
@@ -1014,7 +1014,7 @@ func TestUpsertReplaysUserProfiles(t *testing.T) {
// both profiles, before the delta lands.
replacement := peer.Clone()
replacement.AllowedIPs = append(replacement.AllowedIPs, netip.MustParsePrefix("100.64.0.2/32"))
if err := ms.handleNonKeepAliveMapResponse(ctx, &tailcfg.MapResponse{
if err := ms.HandleNonKeepAliveMapResponse(ctx, &tailcfg.MapResponse{
PeersChanged: []*tailcfg.Node{replacement},
}); err != nil {
t.Fatal(err)
@@ -1045,7 +1045,7 @@ func TestUpsertReplaysUserProfiles(t *testing.T) {
// A patch-only change (no upsert) must not replay any profiles.
patched := replacement.Clone()
patched.Endpoints = eps("10.0.0.1:1111")
if err := ms.handleNonKeepAliveMapResponse(ctx, &tailcfg.MapResponse{
if err := ms.HandleNonKeepAliveMapResponse(ctx, &tailcfg.MapResponse{
PeersChanged: []*tailcfg.Node{patched},
}); err != nil {
t.Fatal(err)