webnet/tailscale
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

control/controlclient,ipn/ipnlocal,wgengine: avoid restarting wireguard when key is learned via tsmp (#19142)

Browse Source 
When disco keys are learned on a node that is connected to control and
has a mapSession, wgengine will see the key as having changed, and
assume that any existing connections will need to be reset.

For keys learned via TSMP, the connection should not be reset as that
key is learned via an active wireguard connection. If wgengine resets
that connetion, a 15s timeout will occur.

This change adds a map to track new keys coming in via TSMP, and removes
them from the list of keys that needs to trigger wireguard resets. This
is done with an interface chain from controlclient down via localBackend
to userspaceEngine via the watchdog.

Once a key has been actively used for preventing a wireguard reset, the
key is removed from the map.

If mapSession becomes a long lived process instead of being dependent on
having a connection to control. This interface chain can be removed, and
the event sequence from wrap->controlClient->userspaceEngine, can be
changed to wrap->userspaceEngine->controlClient as we know the map will
not be gunked up with stale TSMP entries.

Updates #12639

Signed-off-by: Claus Lensbøl <claus@tailscale.com>
This commit is contained in:
Claus Lensbøl
2026-03-30 14:26:08 -04:00
committed by GitHub
parent 99f8039101
commit bf467727fc
8 changed files with 323 additions and 23 deletions
Download Patch File Download Diff File Expand all files Collapse all files
control/controlclient/map.go
+50 -14
View File
@@ -37,6 +37,11 @@ import (
"tailscale.com/wgengine/filter"
)
type responseWithSource struct {
response *tailcfg.MapResponse
viaTSMP bool
}
// mapSession holds the state over a long-polled "map" request to the
// control plane.
//
@@ -98,7 +103,7 @@ type mapSession struct {
lastTKAInfo *tailcfg.TKAInfo
lastNetmapSummary string // from NetworkMap.VeryConcise
cqmu sync.Mutex
changeQueue chan (*tailcfg.MapResponse)
changeQueue chan responseWithSource
changeQueueClosed bool
processQueue sync.WaitGroup
}
@@ -123,7 +128,7 @@ 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),
changeQueue: make(chan responseWithSource),
changeQueueClosed: false,
}
ms.sessionAliveCtx, ms.sessionAliveCtxClose = context.WithCancel(context.Background())
@@ -142,7 +147,7 @@ func (ms *mapSession) run() {
for {
select {
case change := <-ms.changeQueue:
ms.handleNonKeepAliveMapResponse(ms.sessionAliveCtx, change)
ms.handleNonKeepAliveMapResponse(ms.sessionAliveCtx, change.response, change.viaTSMP)
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
@@ -154,7 +159,7 @@ func (ms *mapSession) run() {
for {
select {
case change := <-ms.changeQueue:
ms.handleNonKeepAliveMapResponse(ms.sessionAliveCtx, change)
ms.handleNonKeepAliveMapResponse(ms.sessionAliveCtx, change.response, change.viaTSMP)
default:
// Queue is empty, close it and exit
close(ms.changeQueue)
@@ -190,7 +195,7 @@ func (ms *mapSession) Close() {
var ErrChangeQueueClosed = errors.New("change queue closed")
func (ms *mapSession) updateDiscoForNode(id tailcfg.NodeID, key key.DiscoPublic, lastSeen time.Time, online bool) error {
func (ms *mapSession) updateDiscoForNode(id tailcfg.NodeID, key key.NodePublic, discoKey key.DiscoPublic, lastSeen time.Time, online bool) error {
ms.cqmu.Lock()
if ms.changeQueueClosed {
@@ -199,13 +204,17 @@ func (ms *mapSession) updateDiscoForNode(id tailcfg.NodeID, key key.DiscoPublic,
return ErrChangeQueueClosed
}
resp := &tailcfg.MapResponse{
PeersChangedPatch: []*tailcfg.PeerChange{{
NodeID: id,
LastSeen: &lastSeen,
Online: &online,
DiscoKey: &key,
}},
resp := responseWithSource{
response: &tailcfg.MapResponse{
PeersChangedPatch: []*tailcfg.PeerChange{{
NodeID: id,
Key: &key,
LastSeen: &lastSeen,
Online: &online,
DiscoKey: &discoKey,
}},
},
viaTSMP: true,
}
ms.changeQueue <- resp
ms.cqmu.Unlock()
@@ -221,7 +230,12 @@ func (ms *mapSession) HandleNonKeepAliveMapResponse(ctx context.Context, resp *t
return ErrChangeQueueClosed
}
ms.changeQueue <- resp
change := responseWithSource{
response: resp,
viaTSMP: false,
}
ms.changeQueue <- change
ms.cqmu.Unlock()
return nil
}
@@ -234,7 +248,7 @@ func (ms *mapSession) HandleNonKeepAliveMapResponse(ctx context.Context, resp *t
//
// 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 {
func (ms *mapSession) handleNonKeepAliveMapResponse(ctx context.Context, resp *tailcfg.MapResponse, viaTSMP bool) error {
if debug := resp.Debug; debug != nil {
if err := ms.onDebug(ctx, debug); err != nil {
return err
@@ -284,6 +298,13 @@ func (ms *mapSession) handleNonKeepAliveMapResponse(ctx context.Context, resp *t
ms.updateStateFromResponse(resp)
// If source was learned via TSMP, the updated disco key need to be marked in
// userspaceEngine as an update that should not reconfigure the wireguard
// connection.
if viaTSMP {
ms.tryMarkDiscoAsLearnedFromTSMP(resp)
}
if ms.tryHandleIncrementally(resp) {
ms.occasionallyPrintSummary(ms.lastNetmapSummary)
return nil
@@ -312,6 +333,21 @@ func (ms *mapSession) handleNonKeepAliveMapResponse(ctx context.Context, resp *t
return nil
}
func (ms *mapSession) tryMarkDiscoAsLearnedFromTSMP(res *tailcfg.MapResponse) {
dun, ok := ms.netmapUpdater.(patchDiscoKeyer)
if !ok {
return
}
// In reality we should never really have more than one change here over TSMP.
for _, change := range res.PeersChangedPatch {
if change == nil || change.DiscoKey == nil || change.Key == nil {
continue
}
dun.PatchDiscoKey(*change.Key, *change.DiscoKey)
}
}
// upgradeNode upgrades Node fields from the server into the modern forms
// not using deprecated fields.
func upgradeNode(n *tailcfg.Node) {