mirror of
https://source.quilibrium.com/quilibrium/ceremonyclient.git
synced 2024-11-16 05:05:17 +00:00
661 lines
13 KiB
Go
661 lines
13 KiB
Go
package mux
|
|
|
|
import (
|
|
"bytes"
|
|
"context"
|
|
crand "crypto/rand"
|
|
"fmt"
|
|
"io"
|
|
mrand "math/rand"
|
|
"net"
|
|
"reflect"
|
|
"runtime"
|
|
"runtime/debug"
|
|
"strings"
|
|
"sync"
|
|
"testing"
|
|
"time"
|
|
|
|
"github.com/libp2p/go-libp2p-testing/ci"
|
|
"github.com/libp2p/go-libp2p/core/network"
|
|
"github.com/libp2p/go-libp2p/core/peer"
|
|
|
|
"github.com/stretchr/testify/require"
|
|
)
|
|
|
|
var randomness []byte
|
|
var Subtests map[string]TransportTest
|
|
|
|
func init() {
|
|
// read 1MB of randomness
|
|
randomness = make([]byte, 1<<20)
|
|
if _, err := crand.Read(randomness); err != nil {
|
|
panic(err)
|
|
}
|
|
|
|
Subtests = make(map[string]TransportTest)
|
|
for _, f := range subtests {
|
|
Subtests[getFunctionName(f)] = f
|
|
}
|
|
}
|
|
|
|
func getFunctionName(i interface{}) string {
|
|
return runtime.FuncForPC(reflect.ValueOf(i).Pointer()).Name()
|
|
}
|
|
|
|
type peerScope struct {
|
|
mx sync.Mutex
|
|
memory int
|
|
}
|
|
|
|
func (p *peerScope) ReserveMemory(size int, _ uint8) error {
|
|
p.mx.Lock()
|
|
p.memory += size
|
|
p.mx.Unlock()
|
|
return nil
|
|
}
|
|
|
|
func (p *peerScope) ReleaseMemory(size int) {
|
|
p.mx.Lock()
|
|
defer p.mx.Unlock()
|
|
if p.memory < size {
|
|
panic(fmt.Sprintf("tried to release too much memory: %d (current: %d)", size, p.memory))
|
|
}
|
|
p.memory -= size
|
|
}
|
|
|
|
// Check checks that we don't have any more reserved memory.
|
|
func (p *peerScope) Check(t *testing.T) {
|
|
p.mx.Lock()
|
|
defer p.mx.Unlock()
|
|
require.Zero(t, p.memory, "expected all reserved memory to have been released")
|
|
}
|
|
|
|
type peerScopeSpan struct {
|
|
peerScope
|
|
}
|
|
|
|
func (p *peerScopeSpan) Done() {
|
|
p.mx.Lock()
|
|
defer p.mx.Unlock()
|
|
p.memory = 0
|
|
}
|
|
|
|
func (p *peerScope) Stat() network.ScopeStat { return network.ScopeStat{} }
|
|
func (p *peerScope) BeginSpan() (network.ResourceScopeSpan, error) { return &peerScopeSpan{}, nil }
|
|
func (p *peerScope) Peer() peer.ID { panic("implement me") }
|
|
|
|
var _ network.PeerScope = &peerScope{}
|
|
|
|
type Options struct {
|
|
tr network.Multiplexer
|
|
connNum int
|
|
streamNum int
|
|
msgNum int
|
|
msgMin int
|
|
msgMax int
|
|
}
|
|
|
|
func randBuf(size int) []byte {
|
|
n := len(randomness) - size
|
|
if size < 1 {
|
|
panic(fmt.Errorf("requested too large buffer (%d). max is %d", size, len(randomness)))
|
|
}
|
|
|
|
start := mrand.Intn(n)
|
|
return randomness[start : start+size]
|
|
}
|
|
|
|
func checkErr(t *testing.T, err error) {
|
|
if err != nil {
|
|
debug.PrintStack()
|
|
t.Fatal(err)
|
|
}
|
|
}
|
|
|
|
func echoStream(s network.MuxedStream) {
|
|
defer s.Close()
|
|
io.Copy(s, s) // echo everything
|
|
}
|
|
|
|
func GoServe(t *testing.T, tr network.Multiplexer, l net.Listener) (done func()) {
|
|
closed := make(chan struct{}, 1)
|
|
|
|
go func() {
|
|
for {
|
|
c1, err := l.Accept()
|
|
if err != nil {
|
|
select {
|
|
case <-closed:
|
|
return // closed naturally.
|
|
default:
|
|
checkErr(t, err)
|
|
}
|
|
}
|
|
|
|
sc1, err := tr.NewConn(c1, true, nil)
|
|
checkErr(t, err)
|
|
go func() {
|
|
for {
|
|
str, err := sc1.AcceptStream()
|
|
if err != nil {
|
|
break
|
|
}
|
|
go echoStream(str)
|
|
}
|
|
}()
|
|
}
|
|
}()
|
|
|
|
return func() {
|
|
closed <- struct{}{}
|
|
}
|
|
}
|
|
|
|
func SubtestSimpleWrite(t *testing.T, tr network.Multiplexer) {
|
|
l, err := net.Listen("tcp", "localhost:0")
|
|
checkErr(t, err)
|
|
done := GoServe(t, tr, l)
|
|
defer done()
|
|
|
|
nc1, err := net.Dial("tcp", l.Addr().String())
|
|
checkErr(t, err)
|
|
defer nc1.Close()
|
|
|
|
scope := &peerScope{}
|
|
c1, err := tr.NewConn(nc1, false, scope)
|
|
checkErr(t, err)
|
|
defer func() {
|
|
c1.Close()
|
|
scope.Check(t)
|
|
}()
|
|
|
|
// serve the outgoing conn, because some muxers assume
|
|
// that we _always_ call serve. (this is an error?)
|
|
go c1.AcceptStream()
|
|
|
|
s1, err := c1.OpenStream(context.Background())
|
|
checkErr(t, err)
|
|
defer s1.Close()
|
|
|
|
buf1 := randBuf(4096)
|
|
_, err = s1.Write(buf1)
|
|
checkErr(t, err)
|
|
|
|
buf2 := make([]byte, len(buf1))
|
|
_, err = io.ReadFull(s1, buf2)
|
|
checkErr(t, err)
|
|
|
|
require.Equal(t, buf1, buf2)
|
|
}
|
|
|
|
func SubtestStress(t *testing.T, opt Options) {
|
|
msgsize := 1 << 11
|
|
errs := make(chan error) // dont block anything.
|
|
|
|
rateLimitN := 5000 // max of 5k funcs, because -race has 8k max.
|
|
rateLimitChan := make(chan struct{}, rateLimitN)
|
|
for i := 0; i < rateLimitN; i++ {
|
|
rateLimitChan <- struct{}{}
|
|
}
|
|
|
|
rateLimit := func(f func()) {
|
|
<-rateLimitChan
|
|
f()
|
|
rateLimitChan <- struct{}{}
|
|
}
|
|
|
|
writeStream := func(s network.MuxedStream, bufs chan<- []byte) {
|
|
for i := 0; i < opt.msgNum; i++ {
|
|
buf := randBuf(msgsize)
|
|
bufs <- buf
|
|
if _, err := s.Write(buf); err != nil {
|
|
errs <- fmt.Errorf("s.Write(buf): %s", err)
|
|
continue
|
|
}
|
|
}
|
|
}
|
|
|
|
readStream := func(s network.MuxedStream, bufs <-chan []byte) {
|
|
buf2 := make([]byte, msgsize)
|
|
for buf1 := range bufs {
|
|
if _, err := io.ReadFull(s, buf2); err != nil {
|
|
errs <- fmt.Errorf("io.ReadFull(s, buf2): %s", err)
|
|
continue
|
|
}
|
|
if !bytes.Equal(buf1, buf2) {
|
|
errs <- fmt.Errorf("buffers not equal (%x != %x)", buf1[:3], buf2[:3])
|
|
}
|
|
}
|
|
}
|
|
|
|
openStreamAndRW := func(c network.MuxedConn) {
|
|
s, err := c.OpenStream(context.Background())
|
|
if err != nil {
|
|
errs <- fmt.Errorf("failed to create NewStream: %s", err)
|
|
return
|
|
}
|
|
|
|
bufs := make(chan []byte, opt.msgNum)
|
|
go func() {
|
|
writeStream(s, bufs)
|
|
close(bufs)
|
|
}()
|
|
|
|
readStream(s, bufs)
|
|
s.Close()
|
|
}
|
|
|
|
openConnAndRW := func() {
|
|
l, err := net.Listen("tcp", "localhost:0")
|
|
checkErr(t, err)
|
|
done := GoServe(t, opt.tr, l)
|
|
defer done()
|
|
|
|
nla := l.Addr()
|
|
nc, err := net.Dial(nla.Network(), nla.String())
|
|
checkErr(t, err)
|
|
if err != nil {
|
|
t.Fatal(fmt.Errorf("net.Dial(%s, %s): %s", nla.Network(), nla.String(), err))
|
|
return
|
|
}
|
|
|
|
scope := &peerScope{}
|
|
c, err := opt.tr.NewConn(nc, false, scope)
|
|
if err != nil {
|
|
t.Fatal(fmt.Errorf("a.AddConn(%s <--> %s): %s", nc.LocalAddr(), nc.RemoteAddr(), err))
|
|
return
|
|
}
|
|
|
|
// serve the outgoing conn, because some muxers assume
|
|
// that we _always_ call serve. (this is an error?)
|
|
go func() {
|
|
for {
|
|
str, err := c.AcceptStream()
|
|
if err != nil {
|
|
break
|
|
}
|
|
go echoStream(str)
|
|
}
|
|
}()
|
|
|
|
var wg sync.WaitGroup
|
|
for i := 0; i < opt.streamNum; i++ {
|
|
wg.Add(1)
|
|
go rateLimit(func() {
|
|
defer wg.Done()
|
|
openStreamAndRW(c)
|
|
})
|
|
}
|
|
wg.Wait()
|
|
c.Close()
|
|
scope.Check(t)
|
|
}
|
|
|
|
openConnsAndRW := func() {
|
|
var wg sync.WaitGroup
|
|
for i := 0; i < opt.connNum; i++ {
|
|
wg.Add(1)
|
|
go rateLimit(func() {
|
|
defer wg.Done()
|
|
openConnAndRW()
|
|
})
|
|
}
|
|
wg.Wait()
|
|
}
|
|
|
|
go func() {
|
|
openConnsAndRW()
|
|
close(errs) // done
|
|
}()
|
|
|
|
for err := range errs {
|
|
t.Error(err)
|
|
}
|
|
|
|
}
|
|
|
|
func tcpPipe(t *testing.T) (net.Conn, net.Conn) {
|
|
list, err := net.Listen("tcp", "0.0.0.0:0")
|
|
if err != nil {
|
|
t.Fatal(err)
|
|
}
|
|
|
|
con1, err := net.Dial("tcp", list.Addr().String())
|
|
if err != nil {
|
|
t.Fatal(err)
|
|
}
|
|
|
|
con2, err := list.Accept()
|
|
if err != nil {
|
|
t.Fatal(err)
|
|
}
|
|
|
|
return con1, con2
|
|
}
|
|
|
|
func SubtestStreamOpenStress(t *testing.T, tr network.Multiplexer) {
|
|
wg := new(sync.WaitGroup)
|
|
|
|
a, b := tcpPipe(t)
|
|
defer a.Close()
|
|
defer b.Close()
|
|
|
|
defer wg.Wait()
|
|
|
|
wg.Add(1)
|
|
count := 10000
|
|
workers := 5
|
|
go func() {
|
|
defer wg.Done()
|
|
muxa, err := tr.NewConn(a, true, nil)
|
|
if err != nil {
|
|
t.Error(err)
|
|
return
|
|
}
|
|
stress := func() {
|
|
defer wg.Done()
|
|
for i := 0; i < count; i++ {
|
|
s, err := muxa.OpenStream(context.Background())
|
|
if err != nil {
|
|
t.Error(err)
|
|
return
|
|
}
|
|
err = s.CloseWrite()
|
|
if err != nil {
|
|
t.Error(err)
|
|
}
|
|
n, err := s.Read([]byte{0})
|
|
if n != 0 {
|
|
t.Error("expected to read no bytes")
|
|
}
|
|
if err != io.EOF {
|
|
t.Errorf("expected an EOF, got %s", err)
|
|
}
|
|
}
|
|
}
|
|
|
|
for i := 0; i < workers; i++ {
|
|
wg.Add(1)
|
|
go stress()
|
|
}
|
|
}()
|
|
|
|
scope := &peerScope{}
|
|
muxb, err := tr.NewConn(b, false, scope)
|
|
if err != nil {
|
|
t.Fatal(err)
|
|
}
|
|
defer func() {
|
|
muxb.Close()
|
|
scope.Check(t)
|
|
}()
|
|
|
|
time.Sleep(time.Millisecond * 50)
|
|
|
|
wg.Add(1)
|
|
recv := make(chan struct{}, count*workers)
|
|
go func() {
|
|
defer wg.Done()
|
|
for i := 0; i < count*workers; i++ {
|
|
str, err := muxb.AcceptStream()
|
|
if err != nil {
|
|
break
|
|
}
|
|
wg.Add(1)
|
|
go func() {
|
|
defer wg.Done()
|
|
str.Close()
|
|
select {
|
|
case recv <- struct{}{}:
|
|
default:
|
|
t.Error("too many stream")
|
|
}
|
|
}()
|
|
}
|
|
}()
|
|
|
|
timeout := time.Second * 10
|
|
if ci.IsRunning() {
|
|
timeout *= 10
|
|
}
|
|
|
|
limit := time.After(timeout)
|
|
for i := 0; i < count*workers; i++ {
|
|
select {
|
|
case <-recv:
|
|
case <-limit:
|
|
t.Fatal("timed out receiving streams")
|
|
}
|
|
}
|
|
|
|
wg.Wait()
|
|
}
|
|
|
|
func SubtestStreamReset(t *testing.T, tr network.Multiplexer) {
|
|
wg := new(sync.WaitGroup)
|
|
defer wg.Wait()
|
|
|
|
a, b := tcpPipe(t)
|
|
defer a.Close()
|
|
defer b.Close()
|
|
|
|
wg.Add(1)
|
|
scopea := &peerScope{}
|
|
muxa, err := tr.NewConn(a, true, scopea)
|
|
if err != nil {
|
|
t.Error(err)
|
|
return
|
|
}
|
|
defer func() {
|
|
muxa.Close()
|
|
scopea.Check(t)
|
|
}()
|
|
|
|
go func() {
|
|
defer wg.Done()
|
|
s, err := muxa.OpenStream(context.Background())
|
|
if err != nil {
|
|
t.Error(err)
|
|
return
|
|
}
|
|
time.Sleep(time.Millisecond * 50)
|
|
|
|
_, err = s.Write([]byte("foo"))
|
|
if err != network.ErrReset {
|
|
t.Error("should have been stream reset")
|
|
}
|
|
s.Close()
|
|
}()
|
|
|
|
scopeb := &peerScope{}
|
|
muxb, err := tr.NewConn(b, false, scopeb)
|
|
if err != nil {
|
|
t.Fatal(err)
|
|
}
|
|
defer func() {
|
|
muxb.Close()
|
|
scopeb.Check(t)
|
|
}()
|
|
|
|
str, err := muxb.AcceptStream()
|
|
checkErr(t, err)
|
|
str.Reset()
|
|
|
|
wg.Wait()
|
|
}
|
|
|
|
// check that Close also closes the underlying net.Conn
|
|
func SubtestWriteAfterClose(t *testing.T, tr network.Multiplexer) {
|
|
a, b := tcpPipe(t)
|
|
|
|
scopea := &peerScope{}
|
|
muxa, err := tr.NewConn(a, true, scopea)
|
|
checkErr(t, err)
|
|
|
|
scopeb := &peerScope{}
|
|
muxb, err := tr.NewConn(b, false, scopeb)
|
|
checkErr(t, err)
|
|
|
|
checkErr(t, muxa.Close())
|
|
scopea.Check(t)
|
|
checkErr(t, muxb.Close())
|
|
scopeb.Check(t)
|
|
|
|
// make sure the underlying net.Conn was closed
|
|
if _, err := a.Write([]byte("foobar")); err == nil || !strings.Contains(err.Error(), "use of closed network connection") {
|
|
t.Fatal("write should have failed")
|
|
}
|
|
if _, err := b.Write([]byte("foobar")); err == nil || !strings.Contains(err.Error(), "use of closed network connection") {
|
|
t.Fatal("write should have failed")
|
|
}
|
|
}
|
|
|
|
func SubtestStreamLeftOpen(t *testing.T, tr network.Multiplexer) {
|
|
a, b := tcpPipe(t)
|
|
|
|
const numStreams = 10
|
|
const dataLen = 50 * 1024
|
|
|
|
scopea := &peerScope{}
|
|
muxa, err := tr.NewConn(a, true, scopea)
|
|
checkErr(t, err)
|
|
|
|
scopeb := &peerScope{}
|
|
muxb, err := tr.NewConn(b, false, scopeb)
|
|
checkErr(t, err)
|
|
|
|
var wg sync.WaitGroup
|
|
wg.Add(1 + numStreams)
|
|
go func() {
|
|
defer wg.Done()
|
|
for i := 0; i < numStreams; i++ {
|
|
stra, err := muxa.OpenStream(context.Background())
|
|
checkErr(t, err)
|
|
go func() {
|
|
defer wg.Done()
|
|
_, err = stra.Write(randBuf(dataLen))
|
|
checkErr(t, err)
|
|
// do NOT close or reset the stream
|
|
}()
|
|
}
|
|
}()
|
|
|
|
wg.Add(1 + numStreams)
|
|
go func() {
|
|
defer wg.Done()
|
|
for i := 0; i < numStreams; i++ {
|
|
str, err := muxb.AcceptStream()
|
|
checkErr(t, err)
|
|
go func() {
|
|
defer wg.Done()
|
|
_, err = io.ReadFull(str, make([]byte, dataLen))
|
|
checkErr(t, err)
|
|
}()
|
|
}
|
|
}()
|
|
|
|
// Now we have a bunch of open streams.
|
|
// Make sure that their memory is returned when we close the connection.
|
|
wg.Wait()
|
|
|
|
muxa.Close()
|
|
scopea.Check(t)
|
|
muxb.Close()
|
|
scopeb.Check(t)
|
|
}
|
|
|
|
func SubtestStress1Conn1Stream1Msg(t *testing.T, tr network.Multiplexer) {
|
|
SubtestStress(t, Options{
|
|
tr: tr,
|
|
connNum: 1,
|
|
streamNum: 1,
|
|
msgNum: 1,
|
|
msgMax: 100,
|
|
msgMin: 100,
|
|
})
|
|
}
|
|
|
|
func SubtestStress1Conn1Stream100Msg(t *testing.T, tr network.Multiplexer) {
|
|
SubtestStress(t, Options{
|
|
tr: tr,
|
|
connNum: 1,
|
|
streamNum: 1,
|
|
msgNum: 100,
|
|
msgMax: 100,
|
|
msgMin: 100,
|
|
})
|
|
}
|
|
|
|
func SubtestStress1Conn100Stream100Msg(t *testing.T, tr network.Multiplexer) {
|
|
SubtestStress(t, Options{
|
|
tr: tr,
|
|
connNum: 1,
|
|
streamNum: 100,
|
|
msgNum: 100,
|
|
msgMax: 100,
|
|
msgMin: 100,
|
|
})
|
|
}
|
|
|
|
func SubtestStress10Conn10Stream50Msg(t *testing.T, tr network.Multiplexer) {
|
|
SubtestStress(t, Options{
|
|
tr: tr,
|
|
connNum: 10,
|
|
streamNum: 10,
|
|
msgNum: 50,
|
|
msgMax: 100,
|
|
msgMin: 100,
|
|
})
|
|
}
|
|
|
|
func SubtestStress1Conn1000Stream10Msg(t *testing.T, tr network.Multiplexer) {
|
|
SubtestStress(t, Options{
|
|
tr: tr,
|
|
connNum: 1,
|
|
streamNum: 1000,
|
|
msgNum: 10,
|
|
msgMax: 100,
|
|
msgMin: 100,
|
|
})
|
|
}
|
|
|
|
func SubtestStress1Conn100Stream100Msg10MB(t *testing.T, tr network.Multiplexer) {
|
|
SubtestStress(t, Options{
|
|
tr: tr,
|
|
connNum: 1,
|
|
streamNum: 100,
|
|
msgNum: 100,
|
|
msgMax: 10000,
|
|
msgMin: 1000,
|
|
})
|
|
}
|
|
|
|
// Subtests are all the subtests run by SubtestAll
|
|
var subtests = []TransportTest{
|
|
SubtestSimpleWrite,
|
|
SubtestWriteAfterClose,
|
|
SubtestStress1Conn1Stream1Msg,
|
|
SubtestStress1Conn1Stream100Msg,
|
|
SubtestStress1Conn100Stream100Msg,
|
|
SubtestStress10Conn10Stream50Msg,
|
|
SubtestStress1Conn1000Stream10Msg,
|
|
SubtestStress1Conn100Stream100Msg10MB,
|
|
SubtestStreamOpenStress,
|
|
SubtestStreamReset,
|
|
SubtestStreamLeftOpen,
|
|
}
|
|
|
|
// SubtestAll runs all the stream multiplexer tests against the target
|
|
// transport.
|
|
func SubtestAll(t *testing.T, tr network.Multiplexer) {
|
|
for name, f := range Subtests {
|
|
t.Run(name, func(t *testing.T) {
|
|
f(t, tr)
|
|
})
|
|
}
|
|
}
|
|
|
|
// TransportTest is a stream multiplex transport test case
|
|
type TransportTest func(t *testing.T, tr network.Multiplexer)
|