try to make certificate addition/removal reloadable in some cases (#1468)

* try to make certificate addition/removal reloadable in some cases

* very spicy change to respond to handshakes with cert versions we cannot match with a cert that we can indeed match

* even spicier change to rehandshake if we detect our cert is lower-version than our peer, and we have a newer-version cert available

* make tryRehandshake easier to understand
This commit is contained in:
Jack Doan
2025-11-03 19:38:44 -06:00
committed by GitHub
parent 770147264d
commit 01909f4715
7 changed files with 528 additions and 82 deletions

View File

@@ -354,7 +354,6 @@ func (cm *connectionManager) makeTrafficDecision(localIndex uint32, now time.Tim
if mainHostInfo {
decision = tryRehandshake
} else {
if cm.shouldSwapPrimary(hostinfo) {
decision = swapPrimary
@@ -461,6 +460,10 @@ func (cm *connectionManager) shouldSwapPrimary(current *HostInfo) bool {
}
crt := cm.intf.pki.getCertState().getCertificate(current.ConnectionState.myCert.Version())
if crt == nil {
//my cert was reloaded away. We should definitely swap from this tunnel
return true
}
// If this tunnel is using the latest certificate then we should swap it to primary for a bit and see if things
// settle down.
return bytes.Equal(current.ConnectionState.myCert.Signature(), crt.Signature())
@@ -475,31 +478,34 @@ func (cm *connectionManager) swapPrimary(current, primary *HostInfo) {
cm.hostMap.Unlock()
}
// isInvalidCertificate will check if we should destroy a tunnel if pki.disconnect_invalid is true and
// the certificate is no longer valid. Block listed certificates will skip the pki.disconnect_invalid
// check and return true.
// isInvalidCertificate decides if we should destroy a tunnel.
// returns true if pki.disconnect_invalid is true and the certificate is no longer valid.
// Blocklisted certificates will skip the pki.disconnect_invalid check and return true.
func (cm *connectionManager) isInvalidCertificate(now time.Time, hostinfo *HostInfo) bool {
remoteCert := hostinfo.GetCert()
if remoteCert == nil {
return false
return false //don't tear down tunnels for handshakes in progress
}
caPool := cm.intf.pki.GetCAPool()
err := caPool.VerifyCachedCertificate(now, remoteCert)
if err == nil {
return false
}
if !cm.intf.disconnectInvalid.Load() && err != cert.ErrBlockListed {
return false //cert is still valid! yay!
} else if err == cert.ErrBlockListed { //avoiding errors.Is for speed
// Block listed certificates should always be disconnected
hostinfo.logger(cm.l).WithError(err).
WithField("fingerprint", remoteCert.Fingerprint).
Info("Remote certificate is blocked, tearing down the tunnel")
return true
} else if cm.intf.disconnectInvalid.Load() {
hostinfo.logger(cm.l).WithError(err).
WithField("fingerprint", remoteCert.Fingerprint).
Info("Remote certificate is no longer valid, tearing down the tunnel")
return true
} else {
//if we reach here, the cert is no longer valid, but we're configured to keep tunnels from now-invalid certs open
return false
}
hostinfo.logger(cm.l).WithError(err).
WithField("fingerprint", remoteCert.Fingerprint).
Info("Remote certificate is no longer valid, tearing down the tunnel")
return true
}
func (cm *connectionManager) sendPunch(hostinfo *HostInfo) {
@@ -530,15 +536,45 @@ func (cm *connectionManager) sendPunch(hostinfo *HostInfo) {
func (cm *connectionManager) tryRehandshake(hostinfo *HostInfo) {
cs := cm.intf.pki.getCertState()
curCrt := hostinfo.ConnectionState.myCert
myCrt := cs.getCertificate(curCrt.Version())
if curCrt.Version() >= cs.initiatingVersion && bytes.Equal(curCrt.Signature(), myCrt.Signature()) == true {
// The current tunnel is using the latest certificate and version, no need to rehandshake.
curCrtVersion := curCrt.Version()
myCrt := cs.getCertificate(curCrtVersion)
if myCrt == nil {
cm.l.WithField("vpnAddrs", hostinfo.vpnAddrs).
WithField("version", curCrtVersion).
WithField("reason", "local certificate removed").
Info("Re-handshaking with remote")
cm.intf.handshakeManager.StartHandshake(hostinfo.vpnAddrs[0], nil)
return
}
peerCrt := hostinfo.ConnectionState.peerCert
if peerCrt != nil && curCrtVersion < peerCrt.Certificate.Version() {
// if our certificate version is less than theirs, and we have a matching version available, rehandshake?
if cs.getCertificate(peerCrt.Certificate.Version()) != nil {
cm.l.WithField("vpnAddrs", hostinfo.vpnAddrs).
WithField("version", curCrtVersion).
WithField("peerVersion", peerCrt.Certificate.Version()).
WithField("reason", "local certificate version lower than peer, attempting to correct").
Info("Re-handshaking with remote")
cm.intf.handshakeManager.StartHandshake(hostinfo.vpnAddrs[0], func(hh *HandshakeHostInfo) {
hh.initiatingVersionOverride = peerCrt.Certificate.Version()
})
return
}
}
if !bytes.Equal(curCrt.Signature(), myCrt.Signature()) {
cm.l.WithField("vpnAddrs", hostinfo.vpnAddrs).
WithField("reason", "local certificate is not current").
Info("Re-handshaking with remote")
cm.l.WithField("vpnAddrs", hostinfo.vpnAddrs).
WithField("reason", "local certificate is not current").
Info("Re-handshaking with remote")
cm.intf.handshakeManager.StartHandshake(hostinfo.vpnAddrs[0], nil)
return
}
if curCrtVersion < cs.initiatingVersion {
cm.l.WithField("vpnAddrs", hostinfo.vpnAddrs).
WithField("reason", "current cert version < pki.initiatingVersion").
Info("Re-handshaking with remote")
cm.intf.handshakeManager.StartHandshake(hostinfo.vpnAddrs[0], nil)
cm.intf.handshakeManager.StartHandshake(hostinfo.vpnAddrs[0], nil)
return
}
}