ipn/ipnlocal: add wireguard session state metrics + publish on IPN bus

Updates #19989
Updates tailscale/corp#42874

Change-Id: I843ed95bc7b0f5cd38ba1467332c6b022901e254
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
This commit is contained in:
Brad Fitzpatrick
2026-06-15 11:41:18 -07:00
committed by Brad Fitzpatrick
parent ae743642d9
commit 6596d237a3
11 changed files with 602 additions and 10 deletions
+24
View File
@@ -21,6 +21,7 @@ import (
"github.com/gaissmai/bart"
"github.com/tailscale/wireguard-go/device"
"github.com/tailscale/wireguard-go/tun"
"go4.org/mem"
"tailscale.com/control/controlknobs"
"tailscale.com/drive"
"tailscale.com/envknob"
@@ -726,6 +727,29 @@ func (e *userspaceEngine) SetPeerByIPPacketFunc(fn func(netip.Addr) (_ key.NodeP
})
}
func (e *userspaceEngine) SetPeerSessionStateFunc(fn func(key.NodePublic, PeerWireGuardState)) {
e.wgdev.SetSessionStateFunc(func(pk device.NoisePublicKey, state device.PeerSessionState) {
if fn != nil {
fn(key.NodePublicFromRaw32(mem.B(pk[:])), peerWireGuardStateFromDevice(state))
}
})
}
func peerWireGuardStateFromDevice(state device.PeerSessionState) PeerWireGuardState {
switch state {
case device.PeerSessionNone:
return PeerWireGuardStateNone
case device.PeerSessionHandshake:
return PeerWireGuardStateHandshake
case device.PeerSessionEstablished:
return PeerWireGuardStateEstablished
case device.PeerSessionExpired:
return PeerWireGuardStateExpired
default:
panic(fmt.Sprintf("unexpected wireguard-go PeerSessionState %d", state))
}
}
// hasOverlap checks if there is a IPPrefix which is common amongst the two
// provided slices.
func hasOverlap(aips, rips views.Slice[netip.Prefix]) bool {
+33
View File
@@ -13,6 +13,7 @@ import (
"sync"
"testing"
"github.com/tailscale/wireguard-go/device"
"go4.org/mem"
"tailscale.com/cmd/testwrapper/flakytest"
"tailscale.com/control/controlknobs"
@@ -35,6 +36,38 @@ import (
"tailscale.com/wgengine/wgcfg"
)
func TestPeerWireGuardStateValuesMatchWireguardGo(t *testing.T) {
const unknownPeerSessionState device.PeerSessionState = 255
tests := []struct {
name string
wg device.PeerSessionState
want PeerWireGuardState
}{
{"none", device.PeerSessionNone, PeerWireGuardStateNone},
{"handshake", device.PeerSessionHandshake, PeerWireGuardStateHandshake},
{"established", device.PeerSessionEstablished, PeerWireGuardStateEstablished},
{"expired", device.PeerSessionExpired, PeerWireGuardStateExpired},
}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {
if got := peerWireGuardStateFromDevice(tt.wg); got != tt.want {
t.Fatalf("converted state = %v; want %v", got, tt.want)
}
if got, want := uint8(tt.wg), uint8(tt.want); got != want {
t.Fatalf("wireguard-go const = %v; want %v", got, want)
}
})
}
defer func() {
if recover() == nil {
t.Fatal("expected panic for unknown wireguard-go state")
}
}()
_ = peerWireGuardStateFromDevice(unknownPeerSessionState)
}
func nodeViews(v []*tailcfg.Node) []tailcfg.NodeView {
nv := make([]tailcfg.NodeView, len(v))
for i, n := range v {
+38
View File
@@ -41,6 +41,28 @@ type StatusCallback func(*Status, error)
// into network map updates.
type NetworkMapCallback func(*netmap.NetworkMap)
// PeerWireGuardState is the current WireGuard session state for a peer.
type PeerWireGuardState uint8
const (
// PeerWireGuardStateNone means there is no handshake in progress and no
// session key material retained for this peer.
PeerWireGuardStateNone PeerWireGuardState = 0
// PeerWireGuardStateHandshake means a handshake is in progress for this
// peer, but there is not currently a usable WireGuard session.
PeerWireGuardStateHandshake PeerWireGuardState = 1
// PeerWireGuardStateEstablished means the peer has a completed WireGuard
// session with usable session key material.
PeerWireGuardStateEstablished PeerWireGuardState = 2
// PeerWireGuardStateExpired means the peer's session key material is no
// longer considered usable, but final key cleanup or lazy peer removal may
// not have happened yet.
PeerWireGuardStateExpired PeerWireGuardState = 3
)
// ErrNoChanges is returned by Engine.Reconfig if no changes were made.
var ErrNoChanges = errors.New("no changes made to Engine config")
@@ -141,4 +163,20 @@ type Engine interface {
// SetPeerByIPPacketFunc installs a callback used by wireguard-go to
// look up which peer should handle an outbound packet by destination IP.
SetPeerByIPPacketFunc(func(netip.Addr) (_ key.NodePublic, ok bool))
// SetPeerSessionStateFunc installs a callback used to observe WireGuard
// peer session state transitions.
//
// Calls are serialized per Engine and delivered in transition order from
// wireguard-go, while wireguard-go is holding locks. The callback must be
// cheap and must not call back into wireguard-go.
//
// It does not replay current state. Callers that need a complete view should
// set it before peers are started or lazily created, and maintain any
// snapshots, sequence numbers, and pubsub state outside wireguard-go.
//
// In Tailscale, the usual implementation is
// ipnlocal.LocalBackend.onPeerWireGuardState, installed early in
// LocalBackend construction.
SetPeerSessionStateFunc(func(key.NodePublic, PeerWireGuardState))
}