mirror of
https://github.com/ceph/ceph-csi.git
synced 2024-11-23 06:40:23 +00:00
457 lines
11 KiB
Go
457 lines
11 KiB
Go
// Copyright 2016 The Go Authors. All rights reserved.
|
|
// Use of this source code is governed by a BSD-style
|
|
// license that can be found in the LICENSE file.
|
|
|
|
// Package nettest provides utilities for network testing.
|
|
package nettest
|
|
|
|
import (
|
|
"bytes"
|
|
"encoding/binary"
|
|
"io"
|
|
"io/ioutil"
|
|
"math/rand"
|
|
"net"
|
|
"runtime"
|
|
"sync"
|
|
"testing"
|
|
"time"
|
|
)
|
|
|
|
var (
|
|
aLongTimeAgo = time.Unix(233431200, 0)
|
|
neverTimeout = time.Time{}
|
|
)
|
|
|
|
// MakePipe creates a connection between two endpoints and returns the pair
|
|
// as c1 and c2, such that anything written to c1 is read by c2 and vice-versa.
|
|
// The stop function closes all resources, including c1, c2, and the underlying
|
|
// net.Listener (if there is one), and should not be nil.
|
|
type MakePipe func() (c1, c2 net.Conn, stop func(), err error)
|
|
|
|
// TestConn tests that a net.Conn implementation properly satisfies the interface.
|
|
// The tests should not produce any false positives, but may experience
|
|
// false negatives. Thus, some issues may only be detected when the test is
|
|
// run multiple times. For maximal effectiveness, run the tests under the
|
|
// race detector.
|
|
func TestConn(t *testing.T, mp MakePipe) {
|
|
testConn(t, mp)
|
|
}
|
|
|
|
type connTester func(t *testing.T, c1, c2 net.Conn)
|
|
|
|
func timeoutWrapper(t *testing.T, mp MakePipe, f connTester) {
|
|
c1, c2, stop, err := mp()
|
|
if err != nil {
|
|
t.Fatalf("unable to make pipe: %v", err)
|
|
}
|
|
var once sync.Once
|
|
defer once.Do(func() { stop() })
|
|
timer := time.AfterFunc(time.Minute, func() {
|
|
once.Do(func() {
|
|
t.Error("test timed out; terminating pipe")
|
|
stop()
|
|
})
|
|
})
|
|
defer timer.Stop()
|
|
f(t, c1, c2)
|
|
}
|
|
|
|
// testBasicIO tests that the data sent on c1 is properly received on c2.
|
|
func testBasicIO(t *testing.T, c1, c2 net.Conn) {
|
|
want := make([]byte, 1<<20)
|
|
rand.New(rand.NewSource(0)).Read(want)
|
|
|
|
dataCh := make(chan []byte)
|
|
go func() {
|
|
rd := bytes.NewReader(want)
|
|
if err := chunkedCopy(c1, rd); err != nil {
|
|
t.Errorf("unexpected c1.Write error: %v", err)
|
|
}
|
|
if err := c1.Close(); err != nil {
|
|
t.Errorf("unexpected c1.Close error: %v", err)
|
|
}
|
|
}()
|
|
|
|
go func() {
|
|
wr := new(bytes.Buffer)
|
|
if err := chunkedCopy(wr, c2); err != nil {
|
|
t.Errorf("unexpected c2.Read error: %v", err)
|
|
}
|
|
if err := c2.Close(); err != nil {
|
|
t.Errorf("unexpected c2.Close error: %v", err)
|
|
}
|
|
dataCh <- wr.Bytes()
|
|
}()
|
|
|
|
if got := <-dataCh; !bytes.Equal(got, want) {
|
|
t.Errorf("transmitted data differs")
|
|
}
|
|
}
|
|
|
|
// testPingPong tests that the two endpoints can synchronously send data to
|
|
// each other in a typical request-response pattern.
|
|
func testPingPong(t *testing.T, c1, c2 net.Conn) {
|
|
var wg sync.WaitGroup
|
|
defer wg.Wait()
|
|
|
|
pingPonger := func(c net.Conn) {
|
|
defer wg.Done()
|
|
buf := make([]byte, 8)
|
|
var prev uint64
|
|
for {
|
|
if _, err := io.ReadFull(c, buf); err != nil {
|
|
if err == io.EOF {
|
|
break
|
|
}
|
|
t.Errorf("unexpected Read error: %v", err)
|
|
}
|
|
|
|
v := binary.LittleEndian.Uint64(buf)
|
|
binary.LittleEndian.PutUint64(buf, v+1)
|
|
if prev != 0 && prev+2 != v {
|
|
t.Errorf("mismatching value: got %d, want %d", v, prev+2)
|
|
}
|
|
prev = v
|
|
if v == 1000 {
|
|
break
|
|
}
|
|
|
|
if _, err := c.Write(buf); err != nil {
|
|
t.Errorf("unexpected Write error: %v", err)
|
|
break
|
|
}
|
|
}
|
|
if err := c.Close(); err != nil {
|
|
t.Errorf("unexpected Close error: %v", err)
|
|
}
|
|
}
|
|
|
|
wg.Add(2)
|
|
go pingPonger(c1)
|
|
go pingPonger(c2)
|
|
|
|
// Start off the chain reaction.
|
|
if _, err := c1.Write(make([]byte, 8)); err != nil {
|
|
t.Errorf("unexpected c1.Write error: %v", err)
|
|
}
|
|
}
|
|
|
|
// testRacyRead tests that it is safe to mutate the input Read buffer
|
|
// immediately after cancelation has occurred.
|
|
func testRacyRead(t *testing.T, c1, c2 net.Conn) {
|
|
go chunkedCopy(c2, rand.New(rand.NewSource(0)))
|
|
|
|
var wg sync.WaitGroup
|
|
defer wg.Wait()
|
|
|
|
c1.SetReadDeadline(time.Now().Add(time.Millisecond))
|
|
for i := 0; i < 10; i++ {
|
|
wg.Add(1)
|
|
go func() {
|
|
defer wg.Done()
|
|
|
|
b1 := make([]byte, 1024)
|
|
b2 := make([]byte, 1024)
|
|
for j := 0; j < 100; j++ {
|
|
_, err := c1.Read(b1)
|
|
copy(b1, b2) // Mutate b1 to trigger potential race
|
|
if err != nil {
|
|
checkForTimeoutError(t, err)
|
|
c1.SetReadDeadline(time.Now().Add(time.Millisecond))
|
|
}
|
|
}
|
|
}()
|
|
}
|
|
}
|
|
|
|
// testRacyWrite tests that it is safe to mutate the input Write buffer
|
|
// immediately after cancelation has occurred.
|
|
func testRacyWrite(t *testing.T, c1, c2 net.Conn) {
|
|
go chunkedCopy(ioutil.Discard, c2)
|
|
|
|
var wg sync.WaitGroup
|
|
defer wg.Wait()
|
|
|
|
c1.SetWriteDeadline(time.Now().Add(time.Millisecond))
|
|
for i := 0; i < 10; i++ {
|
|
wg.Add(1)
|
|
go func() {
|
|
defer wg.Done()
|
|
|
|
b1 := make([]byte, 1024)
|
|
b2 := make([]byte, 1024)
|
|
for j := 0; j < 100; j++ {
|
|
_, err := c1.Write(b1)
|
|
copy(b1, b2) // Mutate b1 to trigger potential race
|
|
if err != nil {
|
|
checkForTimeoutError(t, err)
|
|
c1.SetWriteDeadline(time.Now().Add(time.Millisecond))
|
|
}
|
|
}
|
|
}()
|
|
}
|
|
}
|
|
|
|
// testReadTimeout tests that Read timeouts do not affect Write.
|
|
func testReadTimeout(t *testing.T, c1, c2 net.Conn) {
|
|
go chunkedCopy(ioutil.Discard, c2)
|
|
|
|
c1.SetReadDeadline(aLongTimeAgo)
|
|
_, err := c1.Read(make([]byte, 1024))
|
|
checkForTimeoutError(t, err)
|
|
if _, err := c1.Write(make([]byte, 1024)); err != nil {
|
|
t.Errorf("unexpected Write error: %v", err)
|
|
}
|
|
}
|
|
|
|
// testWriteTimeout tests that Write timeouts do not affect Read.
|
|
func testWriteTimeout(t *testing.T, c1, c2 net.Conn) {
|
|
go chunkedCopy(c2, rand.New(rand.NewSource(0)))
|
|
|
|
c1.SetWriteDeadline(aLongTimeAgo)
|
|
_, err := c1.Write(make([]byte, 1024))
|
|
checkForTimeoutError(t, err)
|
|
if _, err := c1.Read(make([]byte, 1024)); err != nil {
|
|
t.Errorf("unexpected Read error: %v", err)
|
|
}
|
|
}
|
|
|
|
// testPastTimeout tests that a deadline set in the past immediately times out
|
|
// Read and Write requests.
|
|
func testPastTimeout(t *testing.T, c1, c2 net.Conn) {
|
|
go chunkedCopy(c2, c2)
|
|
|
|
testRoundtrip(t, c1)
|
|
|
|
c1.SetDeadline(aLongTimeAgo)
|
|
n, err := c1.Write(make([]byte, 1024))
|
|
if n != 0 {
|
|
t.Errorf("unexpected Write count: got %d, want 0", n)
|
|
}
|
|
checkForTimeoutError(t, err)
|
|
n, err = c1.Read(make([]byte, 1024))
|
|
if n != 0 {
|
|
t.Errorf("unexpected Read count: got %d, want 0", n)
|
|
}
|
|
checkForTimeoutError(t, err)
|
|
|
|
testRoundtrip(t, c1)
|
|
}
|
|
|
|
// testPresentTimeout tests that a deadline set while there are pending
|
|
// Read and Write operations immediately times out those operations.
|
|
func testPresentTimeout(t *testing.T, c1, c2 net.Conn) {
|
|
var wg sync.WaitGroup
|
|
defer wg.Wait()
|
|
wg.Add(3)
|
|
|
|
deadlineSet := make(chan bool, 1)
|
|
go func() {
|
|
defer wg.Done()
|
|
time.Sleep(100 * time.Millisecond)
|
|
deadlineSet <- true
|
|
c1.SetReadDeadline(aLongTimeAgo)
|
|
c1.SetWriteDeadline(aLongTimeAgo)
|
|
}()
|
|
go func() {
|
|
defer wg.Done()
|
|
n, err := c1.Read(make([]byte, 1024))
|
|
if n != 0 {
|
|
t.Errorf("unexpected Read count: got %d, want 0", n)
|
|
}
|
|
checkForTimeoutError(t, err)
|
|
if len(deadlineSet) == 0 {
|
|
t.Error("Read timed out before deadline is set")
|
|
}
|
|
}()
|
|
go func() {
|
|
defer wg.Done()
|
|
var err error
|
|
for err == nil {
|
|
_, err = c1.Write(make([]byte, 1024))
|
|
}
|
|
checkForTimeoutError(t, err)
|
|
if len(deadlineSet) == 0 {
|
|
t.Error("Write timed out before deadline is set")
|
|
}
|
|
}()
|
|
}
|
|
|
|
// testFutureTimeout tests that a future deadline will eventually time out
|
|
// Read and Write operations.
|
|
func testFutureTimeout(t *testing.T, c1, c2 net.Conn) {
|
|
var wg sync.WaitGroup
|
|
wg.Add(2)
|
|
|
|
c1.SetDeadline(time.Now().Add(100 * time.Millisecond))
|
|
go func() {
|
|
defer wg.Done()
|
|
_, err := c1.Read(make([]byte, 1024))
|
|
checkForTimeoutError(t, err)
|
|
}()
|
|
go func() {
|
|
defer wg.Done()
|
|
var err error
|
|
for err == nil {
|
|
_, err = c1.Write(make([]byte, 1024))
|
|
}
|
|
checkForTimeoutError(t, err)
|
|
}()
|
|
wg.Wait()
|
|
|
|
go chunkedCopy(c2, c2)
|
|
resyncConn(t, c1)
|
|
testRoundtrip(t, c1)
|
|
}
|
|
|
|
// testCloseTimeout tests that calling Close immediately times out pending
|
|
// Read and Write operations.
|
|
func testCloseTimeout(t *testing.T, c1, c2 net.Conn) {
|
|
go chunkedCopy(c2, c2)
|
|
|
|
var wg sync.WaitGroup
|
|
defer wg.Wait()
|
|
wg.Add(3)
|
|
|
|
// Test for cancelation upon connection closure.
|
|
c1.SetDeadline(neverTimeout)
|
|
go func() {
|
|
defer wg.Done()
|
|
time.Sleep(100 * time.Millisecond)
|
|
c1.Close()
|
|
}()
|
|
go func() {
|
|
defer wg.Done()
|
|
var err error
|
|
buf := make([]byte, 1024)
|
|
for err == nil {
|
|
_, err = c1.Read(buf)
|
|
}
|
|
}()
|
|
go func() {
|
|
defer wg.Done()
|
|
var err error
|
|
buf := make([]byte, 1024)
|
|
for err == nil {
|
|
_, err = c1.Write(buf)
|
|
}
|
|
}()
|
|
}
|
|
|
|
// testConcurrentMethods tests that the methods of net.Conn can safely
|
|
// be called concurrently.
|
|
func testConcurrentMethods(t *testing.T, c1, c2 net.Conn) {
|
|
if runtime.GOOS == "plan9" {
|
|
t.Skip("skipping on plan9; see https://golang.org/issue/20489")
|
|
}
|
|
go chunkedCopy(c2, c2)
|
|
|
|
// The results of the calls may be nonsensical, but this should
|
|
// not trigger a race detector warning.
|
|
var wg sync.WaitGroup
|
|
for i := 0; i < 100; i++ {
|
|
wg.Add(7)
|
|
go func() {
|
|
defer wg.Done()
|
|
c1.Read(make([]byte, 1024))
|
|
}()
|
|
go func() {
|
|
defer wg.Done()
|
|
c1.Write(make([]byte, 1024))
|
|
}()
|
|
go func() {
|
|
defer wg.Done()
|
|
c1.SetDeadline(time.Now().Add(10 * time.Millisecond))
|
|
}()
|
|
go func() {
|
|
defer wg.Done()
|
|
c1.SetReadDeadline(aLongTimeAgo)
|
|
}()
|
|
go func() {
|
|
defer wg.Done()
|
|
c1.SetWriteDeadline(aLongTimeAgo)
|
|
}()
|
|
go func() {
|
|
defer wg.Done()
|
|
c1.LocalAddr()
|
|
}()
|
|
go func() {
|
|
defer wg.Done()
|
|
c1.RemoteAddr()
|
|
}()
|
|
}
|
|
wg.Wait() // At worst, the deadline is set 10ms into the future
|
|
|
|
resyncConn(t, c1)
|
|
testRoundtrip(t, c1)
|
|
}
|
|
|
|
// checkForTimeoutError checks that the error satisfies the Error interface
|
|
// and that Timeout returns true.
|
|
func checkForTimeoutError(t *testing.T, err error) {
|
|
if nerr, ok := err.(net.Error); ok {
|
|
if !nerr.Timeout() {
|
|
t.Errorf("err.Timeout() = false, want true")
|
|
}
|
|
} else {
|
|
t.Errorf("got %T, want net.Error", err)
|
|
}
|
|
}
|
|
|
|
// testRoundtrip writes something into c and reads it back.
|
|
// It assumes that everything written into c is echoed back to itself.
|
|
func testRoundtrip(t *testing.T, c net.Conn) {
|
|
if err := c.SetDeadline(neverTimeout); err != nil {
|
|
t.Errorf("roundtrip SetDeadline error: %v", err)
|
|
}
|
|
|
|
const s = "Hello, world!"
|
|
buf := []byte(s)
|
|
if _, err := c.Write(buf); err != nil {
|
|
t.Errorf("roundtrip Write error: %v", err)
|
|
}
|
|
if _, err := io.ReadFull(c, buf); err != nil {
|
|
t.Errorf("roundtrip Read error: %v", err)
|
|
}
|
|
if string(buf) != s {
|
|
t.Errorf("roundtrip data mismatch: got %q, want %q", buf, s)
|
|
}
|
|
}
|
|
|
|
// resyncConn resynchronizes the connection into a sane state.
|
|
// It assumes that everything written into c is echoed back to itself.
|
|
// It assumes that 0xff is not currently on the wire or in the read buffer.
|
|
func resyncConn(t *testing.T, c net.Conn) {
|
|
c.SetDeadline(neverTimeout)
|
|
errCh := make(chan error)
|
|
go func() {
|
|
_, err := c.Write([]byte{0xff})
|
|
errCh <- err
|
|
}()
|
|
buf := make([]byte, 1024)
|
|
for {
|
|
n, err := c.Read(buf)
|
|
if n > 0 && bytes.IndexByte(buf[:n], 0xff) == n-1 {
|
|
break
|
|
}
|
|
if err != nil {
|
|
t.Errorf("unexpected Read error: %v", err)
|
|
break
|
|
}
|
|
}
|
|
if err := <-errCh; err != nil {
|
|
t.Errorf("unexpected Write error: %v", err)
|
|
}
|
|
}
|
|
|
|
// chunkedCopy copies from r to w in fixed-width chunks to avoid
|
|
// causing a Write that exceeds the maximum packet size for packet-based
|
|
// connections like "unixpacket".
|
|
// We assume that the maximum packet size is at least 1024.
|
|
func chunkedCopy(w io.Writer, r io.Reader) error {
|
|
b := make([]byte, 1024)
|
|
_, err := io.CopyBuffer(struct{ io.Writer }{w}, struct{ io.Reader }{r}, b)
|
|
return err
|
|
}
|