feature/taildrop: add integration test

Taildrop has never had an end-to-end test since it was introduced.

This adds a basic one.

It caught two recent refactoring bugs & one from 2022 (0f7da5c7dc0).

This is prep for moving the rest of Taildrop out of LocalBackend, so
we can do more refactorings with some confidence.

Updates #15812

Change-Id: I6182e49c5641238af0bfdd9fea1ef0420c112738
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
This commit is contained in:
Brad Fitzpatrick 2025-04-28 19:57:01 -07:00 committed by Brad Fitzpatrick
parent a0d7c81a27
commit e415f51351
6 changed files with 272 additions and 50 deletions

View File

@ -573,7 +573,7 @@ func getLocalBackend(ctx context.Context, logf logger.Logf, logID logid.PublicID
if ms, ok := sys.MagicSock.GetOK(); ok {
debugMux.HandleFunc("/debug/magicsock", ms.ServeHTTPDebug)
}
go runDebugServer(debugMux, args.debug)
go runDebugServer(logf, debugMux, args.debug)
}
ns, err := newNetstack(logf, sys)
@ -819,12 +819,20 @@ func servePrometheusMetrics(w http.ResponseWriter, r *http.Request) {
clientmetric.WritePrometheusExpositionFormat(w)
}
func runDebugServer(mux *http.ServeMux, addr string) {
func runDebugServer(logf logger.Logf, mux *http.ServeMux, addr string) {
ln, err := net.Listen("tcp", addr)
if err != nil {
log.Fatalf("debug server: %v", err)
}
if strings.HasSuffix(addr, ":0") {
// Log kernel-selected port number so integration tests
// can find it portably.
logf("DEBUG-ADDR=%v", ln.Addr())
}
srv := &http.Server{
Addr: addr,
Handler: mux,
}
if err := srv.ListenAndServe(); err != nil {
if err := srv.Serve(ln); err != nil {
log.Fatal(err)
}
}

View File

@ -0,0 +1,170 @@
// Copyright (c) Tailscale Inc & AUTHORS
// SPDX-License-Identifier: BSD-3-Clause
package taildrop_test
import (
"bytes"
"context"
"errors"
"fmt"
"io"
"testing"
"time"
"tailscale.com/client/local"
"tailscale.com/client/tailscale/apitype"
"tailscale.com/tailcfg"
"tailscale.com/tstest"
"tailscale.com/tstest/integration"
"tailscale.com/tstest/integration/testcontrol"
)
// TODO(bradfitz): add test where control doesn't send tailcfg.CapabilityFileSharing
// and verify that we get the "file sharing not enabled by Tailscale admin" error.
// TODO(bradfitz): add test between different users with the peercap to permit that?
func TestTaildropIntegration(t *testing.T) {
tstest.Parallel(t)
controlOpt := integration.ConfigureControl(func(s *testcontrol.Server) {
s.AllNodesSameUser = true // required for Taildrop
})
env := integration.NewTestEnv(t, controlOpt)
// Create two nodes:
n1 := integration.NewTestNode(t, env)
d1 := n1.StartDaemon()
n2 := integration.NewTestNode(t, env)
d2 := n2.StartDaemon()
n1.AwaitListening()
t.Logf("n1 is listening")
n2.AwaitListening()
t.Logf("n2 is listening")
n1.MustUp()
t.Logf("n1 is up")
n2.MustUp()
t.Logf("n2 is up")
n1.AwaitRunning()
t.Logf("n1 is running")
n2.AwaitRunning()
t.Logf("n2 is running")
var peerStableID tailcfg.StableNodeID
if err := tstest.WaitFor(5*time.Second, func() error {
st := n1.MustStatus()
if len(st.Peer) == 0 {
return errors.New("no peers")
}
if len(st.Peer) > 1 {
return fmt.Errorf("got %d peers; want 1", len(st.Peer))
}
peer := st.Peer[st.Peers()[0]]
peerStableID = peer.ID
if peer.ID == st.Self.ID {
return errors.New("peer is self")
}
if len(st.TailscaleIPs) == 0 {
return errors.New("no Tailscale IPs")
}
return nil
}); err != nil {
t.Fatal(err)
}
const timeout = 30 * time.Second
ctx, cancel := context.WithTimeout(context.Background(), timeout)
defer cancel()
c1 := n1.LocalClient()
c2 := n2.LocalClient()
wantNoWaitingFiles := func(c *local.Client) {
t.Helper()
files, err := c.WaitingFiles(ctx)
if err != nil {
t.Fatalf("WaitingFiles: %v", err)
}
if len(files) != 0 {
t.Fatalf("WaitingFiles: got %d files; want 0", len(files))
}
}
// Verify c2 has no files.
wantNoWaitingFiles(c2)
gotFile := make(chan bool, 1)
go func() {
v, err := c2.AwaitWaitingFiles(t.Context(), timeout)
if err != nil {
return
}
if len(v) != 0 {
gotFile <- true
}
}()
fileContents := []byte("hello world this is a file")
n2ID := n2.MustStatus().Self.ID
t.Logf("n2 self.ID = %q; n1's peer[0].ID = %q", n2ID, peerStableID)
t.Logf("Doing PushFile ...")
err := c1.PushFile(ctx, n2.MustStatus().Self.ID, int64(len(fileContents)), "test.txt", bytes.NewReader(fileContents))
if err != nil {
t.Fatalf("PushFile from n1->n2: %v", err)
}
t.Logf("PushFile done")
select {
case <-gotFile:
t.Logf("n2 saw AwaitWaitingFiles wake up")
case <-ctx.Done():
t.Fatalf("n2 timeout waiting for AwaitWaitingFiles")
}
files, err := c2.WaitingFiles(ctx)
if err != nil {
t.Fatalf("c2.WaitingFiles: %v", err)
}
if len(files) != 1 {
t.Fatalf("c2.WaitingFiles: got %d files; want 1", len(files))
}
got := files[0]
want := apitype.WaitingFile{
Name: "test.txt",
Size: int64(len(fileContents)),
}
if got != want {
t.Fatalf("c2.WaitingFiles: got %+v; want %+v", got, want)
}
// Download the file.
rc, size, err := c2.GetWaitingFile(ctx, got.Name)
if err != nil {
t.Fatalf("c2.GetWaitingFile: %v", err)
}
if size != int64(len(fileContents)) {
t.Fatalf("c2.GetWaitingFile: got size %d; want %d", size, len(fileContents))
}
gotBytes, err := io.ReadAll(rc)
if err != nil {
t.Fatalf("c2.GetWaitingFile: %v", err)
}
if !bytes.Equal(gotBytes, fileContents) {
t.Fatalf("c2.GetWaitingFile: got %q; want %q", gotBytes, fileContents)
}
// Now delete it.
if err := c2.DeleteWaitingFile(ctx, got.Name); err != nil {
t.Fatalf("c2.DeleteWaitingFile: %v", err)
}
wantNoWaitingFiles(c2)
d1.MustCleanShutdown(t)
d2.MustCleanShutdown(t)
}

View File

@ -365,6 +365,7 @@ func serveFiles(h *localapi.Handler, w http.ResponseWriter, r *http.Request) {
return
}
ctx := r.Context()
var wfs []apitype.WaitingFile
if s := r.FormValue("waitsec"); s != "" && s != "0" {
d, err := strconv.Atoi(s)
if err != nil {
@ -375,11 +376,18 @@ func serveFiles(h *localapi.Handler, w http.ResponseWriter, r *http.Request) {
var cancel context.CancelFunc
ctx, cancel = context.WithDeadline(ctx, deadline)
defer cancel()
}
wfs, err := lb.AwaitWaitingFiles(ctx)
if err != nil && ctx.Err() == nil {
http.Error(w, err.Error(), http.StatusInternalServerError)
return
wfs, err = lb.AwaitWaitingFiles(ctx)
if err != nil && ctx.Err() == nil {
http.Error(w, err.Error(), http.StatusInternalServerError)
return
}
} else {
var err error
wfs, err = lb.WaitingFiles()
if err != nil {
http.Error(w, err.Error(), http.StatusInternalServerError)
return
}
}
w.Header().Set("Content-Type", "application/json")
json.NewEncoder(w).Encode(wfs)

View File

@ -33,6 +33,7 @@ import (
"time"
"go4.org/mem"
"tailscale.com/client/local"
"tailscale.com/derp"
"tailscale.com/derp/derphttp"
"tailscale.com/ipn"
@ -436,6 +437,7 @@ func NewTestEnv(t testing.TB, opts ...TestEnvOpt) *TestEnv {
derpMap := RunDERPAndSTUN(t, logger.Discard, "127.0.0.1")
logc := new(LogCatcher)
control := &testcontrol.Server{
Logf: logger.WithPrefix(t.Logf, "testcontrol: "),
DERPMap: derpMap,
}
control.HTTPTestServer = httptest.NewUnstartedServer(control)
@ -484,6 +486,7 @@ type TestNode struct {
mu sync.Mutex
onLogLine []func([]byte)
lc *local.Client
}
// NewTestNode allocates a temp directory for a new test node.
@ -500,14 +503,18 @@ func NewTestNode(t *testing.T, env *TestEnv) *TestNode {
env: env,
dir: dir,
sockFile: sockFile,
stateFile: filepath.Join(dir, "tailscale.state"),
stateFile: filepath.Join(dir, "tailscaled.state"), // matches what cmd/tailscaled uses
}
// Look for a data race. Once we see the start marker, start logging the rest.
// Look for a data race or panic.
// Once we see the start marker, start logging the rest.
var sawRace bool
var sawPanic bool
n.addLogLineHook(func(line []byte) {
lineB := mem.B(line)
if mem.Contains(lineB, mem.S("DEBUG-ADDR=")) {
t.Log(strings.TrimSpace(string(line)))
}
if mem.Contains(lineB, mem.S("WARNING: DATA RACE")) {
sawRace = true
}
@ -522,6 +529,20 @@ func NewTestNode(t *testing.T, env *TestEnv) *TestNode {
return n
}
func (n *TestNode) LocalClient() *local.Client {
n.mu.Lock()
defer n.mu.Unlock()
if n.lc == nil {
tr := &http.Transport{}
n.lc = &local.Client{
Socket: n.sockFile,
UseSocketOnly: true,
}
n.env.t.Cleanup(tr.CloseIdleConnections)
}
return n.lc
}
func (n *TestNode) diskPrefs() *ipn.Prefs {
t := n.env.t
t.Helper()
@ -668,9 +689,10 @@ func (n *TestNode) StartDaemonAsIPNGOOS(ipnGOOS string) *Daemon {
t := n.env.t
cmd := exec.Command(n.env.daemon)
cmd.Args = append(cmd.Args,
"--state="+n.stateFile,
"--statedir="+n.dir,
"--socket="+n.sockFile,
"--socks5-server=localhost:0",
"--debug=localhost:0",
)
if *verboseTailscaled {
cmd.Args = append(cmd.Args, "-verbose=2")

View File

@ -278,15 +278,20 @@ func TestOneNodeUpAuth(t *testing.T) {
t.Logf("Running up --login-server=%s ...", env.ControlURL())
cmd := n1.Tailscale("up", "--login-server="+env.ControlURL())
var authCountAtomic int32
var authCountAtomic atomic.Int32
cmd.Stdout = &authURLParserWriter{fn: func(urlStr string) error {
t.Logf("saw auth URL %q", urlStr)
if env.Control.CompleteAuth(urlStr) {
atomic.AddInt32(&authCountAtomic, 1)
if authCountAtomic.Add(1) > 1 {
err := errors.New("completed multple auth URLs")
t.Error(err)
return err
}
t.Logf("completed auth path %s", urlStr)
return nil
}
err := fmt.Errorf("Failed to complete auth path to %q", urlStr)
t.Log(err)
t.Error(err)
return err
}}
cmd.Stderr = cmd.Stdout
@ -297,7 +302,7 @@ func TestOneNodeUpAuth(t *testing.T) {
n1.AwaitRunning()
if n := atomic.LoadInt32(&authCountAtomic); n != 1 {
if n := authCountAtomic.Load(); n != 1 {
t.Errorf("Auth URLs completed = %d; want 1", n)
}

View File

@ -55,6 +55,10 @@ type Server struct {
MagicDNSDomain string
HandleC2N http.Handler // if non-nil, used for /some-c2n-path/ in tests
// AllNodesSameUser, if true, makes all created nodes
// belong to the same user.
AllNodesSameUser bool
// ExplicitBaseURL or HTTPTestServer must be set.
ExplicitBaseURL string // e.g. "http://127.0.0.1:1234" with no trailing URL
HTTPTestServer *httptest.Server // if non-nil, used to get BaseURL
@ -96,9 +100,9 @@ type Server struct {
logins map[key.NodePublic]*tailcfg.Login
updates map[tailcfg.NodeID]chan updateType
authPath map[string]*AuthPath
nodeKeyAuthed map[key.NodePublic]bool // key => true once authenticated
msgToSend map[key.NodePublic]any // value is *tailcfg.PingRequest or entire *tailcfg.MapResponse
allExpired bool // All nodes will be told their node key is expired.
nodeKeyAuthed set.Set[key.NodePublic]
msgToSend map[key.NodePublic]any // value is *tailcfg.PingRequest or entire *tailcfg.MapResponse
allExpired bool // All nodes will be told their node key is expired.
}
// BaseURL returns the server's base URL, without trailing slash.
@ -522,6 +526,10 @@ func (s *Server) getUser(nodeKey key.NodePublic) (*tailcfg.User, *tailcfg.Login)
return u, s.logins[nodeKey]
}
id := tailcfg.UserID(len(s.users) + 1)
if s.AllNodesSameUser {
id = 123
}
s.logf("Created user %v for node %s", id, nodeKey)
loginName := fmt.Sprintf("user-%d@%s", id, domain)
displayName := fmt.Sprintf("User %d", id)
login := &tailcfg.Login{
@ -582,10 +590,8 @@ func (s *Server) CompleteAuth(authPathOrURL string) bool {
if ap.nodeKey.IsZero() {
panic("zero AuthPath.NodeKey")
}
if s.nodeKeyAuthed == nil {
s.nodeKeyAuthed = map[key.NodePublic]bool{}
}
s.nodeKeyAuthed[ap.nodeKey] = true
s.nodeKeyAuthed.Make()
s.nodeKeyAuthed.Add(ap.nodeKey)
ap.CompleteSuccessfully()
return true
}
@ -645,36 +651,40 @@ func (s *Server) serveRegister(w http.ResponseWriter, r *http.Request, mkey key.
if s.nodes == nil {
s.nodes = map[key.NodePublic]*tailcfg.Node{}
}
_, ok := s.nodes[nk]
machineAuthorized := true // TODO: add Server.RequireMachineAuth
if !ok {
v4Prefix := netip.PrefixFrom(netaddr.IPv4(100, 64, uint8(tailcfg.NodeID(user.ID)>>8), uint8(tailcfg.NodeID(user.ID))), 32)
v6Prefix := netip.PrefixFrom(tsaddr.Tailscale4To6(v4Prefix.Addr()), 128)
nodeID := len(s.nodes) + 1
v4Prefix := netip.PrefixFrom(netaddr.IPv4(100, 64, uint8(nodeID>>8), uint8(nodeID)), 32)
v6Prefix := netip.PrefixFrom(tsaddr.Tailscale4To6(v4Prefix.Addr()), 128)
allowedIPs := []netip.Prefix{
v4Prefix,
v6Prefix,
}
s.nodes[nk] = &tailcfg.Node{
ID: tailcfg.NodeID(user.ID),
StableID: tailcfg.StableNodeID(fmt.Sprintf("TESTCTRL%08x", int(user.ID))),
User: user.ID,
Machine: mkey,
Key: req.NodeKey,
MachineAuthorized: machineAuthorized,
Addresses: allowedIPs,
AllowedIPs: allowedIPs,
Hostinfo: req.Hostinfo.View(),
Name: req.Hostinfo.Hostname,
Capabilities: []tailcfg.NodeCapability{
tailcfg.CapabilityHTTPS,
tailcfg.NodeAttrFunnel,
tailcfg.CapabilityFunnelPorts + "?ports=8080,443",
},
allowedIPs := []netip.Prefix{
v4Prefix,
v6Prefix,
}
node := &tailcfg.Node{
ID: tailcfg.NodeID(nodeID),
StableID: tailcfg.StableNodeID(fmt.Sprintf("TESTCTRL%08x", int(nodeID))),
User: user.ID,
Machine: mkey,
Key: req.NodeKey,
MachineAuthorized: machineAuthorized,
Addresses: allowedIPs,
AllowedIPs: allowedIPs,
Hostinfo: req.Hostinfo.View(),
Name: req.Hostinfo.Hostname,
Capabilities: []tailcfg.NodeCapability{
tailcfg.CapabilityHTTPS,
tailcfg.NodeAttrFunnel,
tailcfg.CapabilityFileSharing,
tailcfg.CapabilityFunnelPorts + "?ports=8080,443",
},
}
s.nodes[nk] = node
}
requireAuth := s.RequireAuth
if requireAuth && s.nodeKeyAuthed[nk] {
if requireAuth && s.nodeKeyAuthed.Contains(nk) {
requireAuth = false
}
allExpired := s.allExpired
@ -951,7 +961,6 @@ func (s *Server) MapResponse(req *tailcfg.MapRequest) (res *tailcfg.MapResponse,
node.CapMap = nodeCapMap
node.Capabilities = append(node.Capabilities, tailcfg.NodeAttrDisableUPnP)
user, _ := s.getUser(nk)
t := time.Date(2020, 8, 3, 0, 0, 0, 1, time.UTC)
dns := s.DNSConfig
if dns != nil && s.MagicDNSDomain != "" {
@ -1013,7 +1022,7 @@ func (s *Server) MapResponse(req *tailcfg.MapRequest) (res *tailcfg.MapResponse,
})
res.UserProfiles = s.allUserProfiles()
v4Prefix := netip.PrefixFrom(netaddr.IPv4(100, 64, uint8(tailcfg.NodeID(user.ID)>>8), uint8(tailcfg.NodeID(user.ID))), 32)
v4Prefix := netip.PrefixFrom(netaddr.IPv4(100, 64, uint8(node.ID>>8), uint8(node.ID)), 32)
v6Prefix := netip.PrefixFrom(tsaddr.Tailscale4To6(v4Prefix.Addr()), 128)
res.Node.Addresses = []netip.Prefix{