Fix warm fork guest MAC reconfiguration

lib/instances/restore.go

@@ -379,7 +379,7 @@ func (m *manager) restoreFromSnapshot(
 }
 
 func reconfigureGuestNetwork(ctx context.Context, stored *StoredMetadata, alloc *network.Allocation) error {
-	prefix, err := netmaskToPrefix(alloc.Netmask)
+	cmd, err := guestNetworkReconfigureCommand(alloc)
 	if err != nil {
 		return err
 	}
@@ -389,11 +389,6 @@ func reconfigureGuestNetwork(ctx context.Context, stored *StoredMetadata, alloc
 		return fmt.Errorf("create vsock dialer: %w", err)
 	}
 
-	cmd := fmt.Sprintf(
-		"ip -4 addr flush dev eth0 scope global && ip addr add %s/%d dev eth0 && ip link set dev eth0 up && ip route replace default via %s dev eth0",
-		alloc.IP, prefix, alloc.Gateway,
-	)
-
 	var stdout, stderr bytes.Buffer
 	exit, err := guest.ExecIntoInstance(ctx, dialer, guest.ExecOptions{
 		Command:      []string{"sh", "-c", cmd},
@@ -411,6 +406,49 @@ func reconfigureGuestNetwork(ctx context.Context, stored *StoredMetadata, alloc
 	return nil
 }
 
+func guestNetworkReconfigureCommand(alloc *network.Allocation) (string, error) {
+	if alloc == nil {
+		return "", fmt.Errorf("missing network allocation")
+	}
+	ip := strings.TrimSpace(alloc.IP)
+	if ip == "" {
+		return "", fmt.Errorf("missing network allocation IP")
+	}
+	mac := strings.ToLower(strings.TrimSpace(alloc.MAC))
+	if mac == "" {
+		return "", fmt.Errorf("missing network allocation MAC")
+	}
+	if _, err := net.ParseMAC(mac); err != nil {
+		return "", fmt.Errorf("invalid network allocation MAC %q: %w", alloc.MAC, err)
+	}
+	gateway := strings.TrimSpace(alloc.Gateway)
+	if gateway == "" {
+		return "", fmt.Errorf("missing network allocation gateway")
+	}
+	prefix, err := netmaskToPrefix(alloc.Netmask)
+	if err != nil {
+		return "", err
+	}
+
+	return fmt.Sprintf(
+		// Bring eth0 down so Linux permits changing the interface MAC.
+		"ip link set dev eth0 down && "+
+			// Replace the snapshotted MAC with the MAC allocated for this fork.
+			"ip link set dev eth0 address %s && "+
+			// Remove the snapshotted IPv4 address from the source/starter guest.
+			"ip -4 addr flush dev eth0 scope global && "+
+			// Add the IPv4 address allocated for this fork.
+			"ip addr add %s/%d dev eth0 && "+
+			// Bring the interface back up after applying the new identity.
+			"ip link set dev eth0 up && "+
+			// Ensure outbound traffic uses the fork's allocated gateway.
+			"ip route replace default via %s dev eth0 && "+
+			// Drop snapshotted ARP/neighbor entries so peers are rediscovered.
+			"(ip neigh flush dev eth0 || true)",
+		mac, ip, prefix, gateway,
+	), nil
+}
+
 func networkConfigFromAllocation(alloc *network.Allocation) *network.NetworkConfig {
 	if alloc == nil {
 		return nil

lib/instances/restore_egress_test.go

@@ -30,6 +30,38 @@ func TestNetworkConfigFromAllocation_PreservesDNS(t *testing.T) {
 	assert.Equal(t, alloc.TAPDevice, cfg.TAPDevice)
 }
 
+func TestGuestNetworkReconfigureCommand_AppliesAllocatedMAC(t *testing.T) {
+	t.Parallel()
+
+	alloc := &network.Allocation{
+		IP:      "10.102.146.62",
+		MAC:     "02:00:00:85:17:c8",
+		Gateway: "10.102.0.1",
+		Netmask: "255.255.0.0",
+	}
+
+	cmd, err := guestNetworkReconfigureCommand(alloc)
+	require.NoError(t, err)
+	assert.Contains(t, cmd, "ip link set dev eth0 down")
+	assert.Contains(t, cmd, "ip link set dev eth0 address 02:00:00:85:17:c8")
+	assert.Contains(t, cmd, "ip addr add 10.102.146.62/16 dev eth0")
+	assert.Contains(t, cmd, "ip route replace default via 10.102.0.1 dev eth0")
+	assert.Contains(t, cmd, "(ip neigh flush dev eth0 || true)")
+	assert.NotContains(t, cmd, "cat /sys/class/net/eth0/address")
+}
+
+func TestGuestNetworkReconfigureCommand_RequiresAllocatedMAC(t *testing.T) {
+	t.Parallel()
+
+	_, err := guestNetworkReconfigureCommand(&network.Allocation{
+		IP:      "10.102.146.62",
+		Gateway: "10.102.0.1",
+		Netmask: "255.255.0.0",
+	})
+	require.Error(t, err)
+	assert.Contains(t, err.Error(), "missing network allocation MAC")
+}
+
 func TestRequiresRestoreConfigDiskRefresh(t *testing.T) {
 	t.Parallel()
 

lib/network/README.md

@@ -194,7 +194,7 @@ sudo setcap 'cap_net_admin,cap_net_bind_service=+eip' /path/to/hypeman
 1. Get default network details
 2. Check name uniqueness globally
 3. Allocate next available IP (starting from .2, after gateway at .1)
-4. Generate MAC (02:00:00:... format - locally administered)
+4. Generate unused MAC (02:00:00:... format - locally administered)
 5. Generate TAP name (tap-{first8chars-of-instance-id})
 6. Create TAP device and attach to bridge
 

lib/network/allocate.go

@@ -49,10 +49,10 @@ func (m *manager) CreateAllocation(ctx context.Context, req AllocateRequest) (*N
 		return nil, fmt.Errorf("allocate IP: %w", err)
 	}
 
-	// 3. Generate MAC (02:00:00:... format - locally administered)
-	mac, err := generateMAC()
+	// 3. Generate unused MAC (02:00:00:... format - locally administered)
+	mac, err := m.allocateUniqueMAC(ctx)
 	if err != nil {
-		return nil, fmt.Errorf("generate MAC: %w", err)
+		return nil, fmt.Errorf("allocate MAC: %w", err)
 	}
 
 	// 4. Generate TAP name (tap-{first8chars-of-id})
@@ -312,6 +312,51 @@ func incrementIP(ip net.IP, n int) net.IP {
 	return result
 }
 
+const (
+	macAllocationRandomAttempts = 5
+	macSuffixSpace              = 1 << 24
+)
+
+// allocateUniqueMAC picks an unused locally administered MAC address.
+func (m *manager) allocateUniqueMAC(ctx context.Context) (string, error) {
+	allocations, err := m.ListAllocations(ctx)
+	if err != nil {
+		return "", fmt.Errorf("list allocations: %w", err)
+	}
+
+	usedMACs := make(map[string]bool)
+	for _, alloc := range allocations {
+		mac := strings.ToLower(strings.TrimSpace(alloc.MAC))
+		if mac != "" {
+			usedMACs[mac] = true
+		}
+	}
+
+	return allocateUniqueMACFromSet(usedMACs, generateMAC)
+}
+
+func allocateUniqueMACFromSet(usedMACs map[string]bool, generate func() (string, error)) (string, error) {
+	for attempt := 0; attempt < macAllocationRandomAttempts; attempt++ {
+		mac, err := generate()
+		if err != nil {
+			return "", err
+		}
+		mac = strings.ToLower(strings.TrimSpace(mac))
+		if !usedMACs[mac] {
+			return mac, nil
+		}
+	}
+
+	for suffix := 0; suffix < macSuffixSpace; suffix++ {
+		mac := formatMACSuffix(suffix)
+		if !usedMACs[mac] {
+			return mac, nil
+		}
+	}
+
+	return "", fmt.Errorf("no available MAC addresses after %d random attempts and full scan", macAllocationRandomAttempts)
+}
+
 // generateMAC generates a random MAC address with local administration bit set
 func generateMAC() (string, error) {
 	// Generate 6 random bytes
@@ -331,6 +376,11 @@ func generateMAC() (string, error) {
 		buf[0], buf[1], buf[2], buf[3], buf[4], buf[5]), nil
 }
 
+func formatMACSuffix(suffix int) string {
+	return fmt.Sprintf("02:00:00:%02x:%02x:%02x",
+		byte(suffix>>16), byte(suffix>>8), byte(suffix))
+}
+
 // saveClassID persists the tc class ID for an instance so it survives restarts.
 func (m *manager) saveClassID(instanceID, classID string) error {
 	return os.WriteFile(m.paths.InstanceDir(instanceID)+"/classid", []byte(classID), 0644)

lib/network/manager_test.go

@@ -28,6 +28,44 @@ func TestGenerateMAC(t *testing.T) {
 	}
 }
 
+func TestAllocateUniqueMACFromSetRetriesCollisions(t *testing.T) {
+	used := map[string]bool{
+		"02:00:00:00:00:01": true,
+	}
+	candidates := []string{
+		"02:00:00:00:00:01",
+		"02:00:00:00:00:02",
+	}
+	calls := 0
+
+	mac, err := allocateUniqueMACFromSet(used, func() (string, error) {
+		candidate := candidates[calls]
+		calls++
+		return candidate, nil
+	})
+
+	require.NoError(t, err)
+	assert.Equal(t, "02:00:00:00:00:02", mac)
+	assert.Equal(t, 2, calls)
+}
+
+func TestAllocateUniqueMACFromSetFallsBackToSequentialScan(t *testing.T) {
+	used := map[string]bool{
+		"02:00:00:00:00:00": true,
+		"02:00:00:00:00:01": true,
+	}
+	calls := 0
+
+	mac, err := allocateUniqueMACFromSet(used, func() (string, error) {
+		calls++
+		return "02:00:00:00:00:01", nil
+	})
+
+	require.NoError(t, err)
+	assert.Equal(t, "02:00:00:00:00:02", mac)
+	assert.Equal(t, macAllocationRandomAttempts, calls)
+}
+
 func TestGenerateTAPName(t *testing.T) {
 	tests := []struct {
 		name       string