What causes "INetworkConnection.GetAdapterId: %s" in slackhq/nebula?

Re-enumerate network connections to get fresh iNetworkConnection pointers, then retry GetAdapterId

INetworkConnection.GetAdapterId: %s

Error message

INetworkConnection.GetAdapterId: %s

What it means

This error wraps a failed HRESULT from INetworkConnection::GetAdapterId, a COM call used during Windows network category detection to correlate a network connection with its adapter GUID. A failing HRESULT means the NLM COM object refused or failed the call (service unavailable, stale connection pointer, or permission problem).

Source

Thrown at overlay/network_category_windows.go:188

	GetConnectivity       uintptr
	GetConnectionId       uintptr
	GetAdapterId          uintptr
	GetDomainType         uintptr
}

type iNetworkConnection struct{ Vtbl *iNetworkConnectionVtbl }

func (c *iNetworkConnection) Release() {
	syscall.SyscallN(c.Vtbl.Release, uintptr(unsafe.Pointer(c)))
}

func (c *iNetworkConnection) GetAdapterId() (windows.GUID, error) {
	var g windows.GUID
	r1, _, _ := syscall.SyscallN(c.Vtbl.GetAdapterId,
		uintptr(unsafe.Pointer(c)), uintptr(unsafe.Pointer(&g)),
	)
	if hr := hresult(r1); hr.failed() {
		return windows.GUID{}, fmt.Errorf("INetworkConnection.GetAdapterId: %s", hr)
	}
	return g, nil
}

func (c *iNetworkConnection) GetNetwork() (*iNetwork, error) {
	var net *iNetwork
	r1, _, _ := syscall.SyscallN(c.Vtbl.GetNetwork,
		uintptr(unsafe.Pointer(c)), uintptr(unsafe.Pointer(&net)),
	)
	if hr := hresult(r1); hr.failed() {
		return nil, fmt.Errorf("INetworkConnection.GetNetwork: %s", hr)
	}
	return net, nil
}

type iNetworkVtbl struct {
	iDispatchVtbl
	GetName                    uintptr

View on GitHub (pinned to dd8f660c0a)

Solutions

Re-enumerate network connections to get fresh iNetworkConnection pointers, then retry GetAdapterId
Verify the Network List Service (netprofm) is healthy and running
Ensure the COM call runs in the same apartment/thread where the object was created
Inspect the hresult in the message for the exact COM error (e.g. RPC_E_DISCONNECTED) and handle stale-object cases by refreshing the enumeration

Example fix

// before
adapterId, err := conn.GetAdapterId()
if err != nil { return err }
// after: refresh stale connections once
adapterId, err := conn.GetAdapterId()
if err != nil {
    if refreshed, rerr := refreshConnections(); rerr == nil {
        adapterId, err = refreshed.GetAdapterId()
    }
    if err != nil { return err }
}

Defensive patterns

Strategy: retry

Validate before calling

// ensure the connection pointer is fresh before querying
if connsStaleAt.After(netProfileChangedAt) {
    conns, err = nlm.GetNetworkConnections() // re-enumerate
    if err != nil { return err }
}

Type guard

func isAdapterIdError(err error) bool {
    return err != nil && strings.HasPrefix(err.Error(), "INetworkConnection.GetAdapterId:")
}

Try / catch

var adapterId windows.GUID
var err error
for i := 0; i < 2; i++ {
    adapterId, err = conn.GetAdapterId()
    if err == nil { break }
    if conn, err = refreshConnection(); err != nil { break }
}
if err != nil { log.Warnf("adapter id lookup failed: %v", err) }

Prevention

Re-enumerate connections after adapter up/down events before querying
Keep NLM object usage on the creating thread
Check for RPC_E_DISCONNECTED-style hresults indicating stale pointers
Skip classification non-fatally when the adapter is mid-reconfiguration

When it happens

Trigger: Calling GetAdapterId() on an iNetworkConnection pointer obtained from the enumerator while classifying the tunnel adapter's network, when the COM vtbl call returns a failed HRESULT.

Common situations: The connection object became stale after a network change (adapter removed, VPN adapter recreated); NLM service instability; running in restricted service contexts where NLM queries are denied.

Related errors

AI-assisted analysis of slackhq/nebula@dd8f660c0a (2026-09-03). Data as JSON: /api/errors/b5996b134f27b12c. Report an issue: GitHub.