net: limit number of concurrent cgo calls

The limit is 500. There is no way to change it.
This primarily affects name resolution.
If a million goroutines try to resolve DNS names,
only 500 will get to execute cgo calls at a time.
But in return the operating system will not crash.

Fixes #5625.

R=golang-dev, dan.kortschak, r, dvyukov
CC=bradfitz, golang-dev
https://golang.org/cl/13038043

Author: rsc

src/pkg/net/cgo_unix.go

@@ -32,6 +32,9 @@ func cgoLookupHost(name string) (addrs []string, err error, completed bool) {
 }
 
 func cgoLookupPort(net, service string) (port int, err error, completed bool) {
+	acquireThread()
+	defer releaseThread()
+
 	var res *C.struct_addrinfo
 	var hints C.struct_addrinfo
 
@@ -79,6 +82,9 @@ func cgoLookupPort(net, service string) (port int, err error, completed bool) {
 }
 
 func cgoLookupIPCNAME(name string) (addrs []IP, cname string, err error, completed bool) {
+	acquireThread()
+	defer releaseThread()
+
 	var res *C.struct_addrinfo
 	var hints C.struct_addrinfo
 

src/pkg/net/dialgoogle_test.go

@@ -54,6 +54,30 @@ var googleaddrsipv4 = []string{
 	"[0:0:0:0:0:ffff::%d.%d.%d.%d]:80",
 }
 
+func TestDNSThreadLimit(t *testing.T) {
+	if testing.Short() || !*testExternal {
+		t.Skip("skipping test to avoid external network")
+	}
+
+	const N = 10000
+	c := make(chan int, N)
+	for i := 0; i < N; i++ {
+		go func() {
+			LookupIP(fmt.Sprintf("%d.net-test.golang.org", i))
+			c <- 1
+		}()
+	}
+	// Don't bother waiting for the stragglers; stop at 0.9 N.
+	for i := 0; i < N*9/10; i++ {
+		if i%100 == 0 {
+			//println("TestDNSThreadLimit:", i)
+		}
+		<-c
+	}
+
+	// If we're still here, it worked.
+}
+
 func TestDialGoogleIPv4(t *testing.T) {
 	if testing.Short() || !*testExternal {
 		t.Skip("skipping test to avoid external network")

src/pkg/net/lookup_windows.go

@@ -34,6 +34,8 @@ func lookupProtocol(name string) (proto int, err error) {
 	}
 	ch := make(chan result)
 	go func() {
+		acquireThread()
+		defer releaseThread()
 		runtime.LockOSThread()
 		defer runtime.UnlockOSThread()
 		proto, err := getprotobyname(name)
@@ -56,6 +58,7 @@ func lookupHost(name string) (addrs []string, err error) {
 }
 
 func gethostbyname(name string) (addrs []IP, err error) {
+	// caller already acquired thread
 	h, err := syscall.GetHostByName(name)
 	if err != nil {
 		return nil, os.NewSyscallError("GetHostByName", err)
@@ -83,6 +86,8 @@ func oldLookupIP(name string) (addrs []IP, err error) {
 	}
 	ch := make(chan result)
 	go func() {
+		acquireThread()
+		defer releaseThread()
 		runtime.LockOSThread()
 		defer runtime.UnlockOSThread()
 		addrs, err := gethostbyname(name)
@@ -93,6 +98,8 @@ func oldLookupIP(name string) (addrs []IP, err error) {
 }
 
 func newLookupIP(name string) (addrs []IP, err error) {
+	acquireThread()
+	defer releaseThread()
 	hints := syscall.AddrinfoW{
 		Family:   syscall.AF_UNSPEC,
 		Socktype: syscall.SOCK_STREAM,
@@ -122,6 +129,8 @@ func newLookupIP(name string) (addrs []IP, err error) {
 }
 
 func getservbyname(network, service string) (port int, err error) {
+	acquireThread()
+	defer releaseThread()
 	switch network {
 	case "tcp4", "tcp6":
 		network = "tcp"
@@ -144,6 +153,8 @@ func oldLookupPort(network, service string) (port int, err error) {
 	}
 	ch := make(chan result)
 	go func() {
+		acquireThread()
+		defer releaseThread()
 		runtime.LockOSThread()
 		defer runtime.UnlockOSThread()
 		port, err := getservbyname(network, service)
@@ -154,6 +165,8 @@ func oldLookupPort(network, service string) (port int, err error) {
 }
 
 func newLookupPort(network, service string) (port int, err error) {
+	acquireThread()
+	defer releaseThread()
 	var stype int32
 	switch network {
 	case "tcp4", "tcp6":
@@ -188,6 +201,8 @@ func newLookupPort(network, service string) (port int, err error) {
 }
 
 func lookupCNAME(name string) (cname string, err error) {
+	acquireThread()
+	defer releaseThread()
 	var r *syscall.DNSRecord
 	e := syscall.DnsQuery(name, syscall.DNS_TYPE_CNAME, 0, nil, &r, nil)
 	if e != nil {
@@ -202,6 +217,8 @@ func lookupCNAME(name string) (cname string, err error) {
 }
 
 func lookupSRV(service, proto, name string) (cname string, addrs []*SRV, err error) {
+	acquireThread()
+	defer releaseThread()
 	var target string
 	if service == "" && proto == "" {
 		target = name
@@ -224,6 +241,8 @@ func lookupSRV(service, proto, name string) (cname string, addrs []*SRV, err err
 }
 
 func lookupMX(name string) (mx []*MX, err error) {
+	acquireThread()
+	defer releaseThread()
 	var r *syscall.DNSRecord
 	e := syscall.DnsQuery(name, syscall.DNS_TYPE_MX, 0, nil, &r, nil)
 	if e != nil {
@@ -240,6 +259,8 @@ func lookupMX(name string) (mx []*MX, err error) {
 }
 
 func lookupNS(name string) (ns []*NS, err error) {
+	acquireThread()
+	defer releaseThread()
 	var r *syscall.DNSRecord
 	e := syscall.DnsQuery(name, syscall.DNS_TYPE_NS, 0, nil, &r, nil)
 	if e != nil {
@@ -255,6 +276,8 @@ func lookupNS(name string) (ns []*NS, err error) {
 }
 
 func lookupTXT(name string) (txt []string, err error) {
+	acquireThread()
+	defer releaseThread()
 	var r *syscall.DNSRecord
 	e := syscall.DnsQuery(name, syscall.DNS_TYPE_TEXT, 0, nil, &r, nil)
 	if e != nil {
@@ -273,6 +296,8 @@ func lookupTXT(name string) (txt []string, err error) {
 }
 
 func lookupAddr(addr string) (name []string, err error) {
+	acquireThread()
+	defer releaseThread()
 	arpa, err := reverseaddr(addr)
 	if err != nil {
 		return nil, err

src/pkg/net/net.go

@@ -433,3 +433,19 @@ func (d *deadline) setTime(t time.Time) {
 		d.set(t.UnixNano())
 	}
 }
+
+// Limit the number of concurrent cgo-using goroutines, because
+// each will block an entire operating system thread. The usual culprit
+// is resolving many DNS names in separate goroutines but the DNS
+// server is not responding. Then the many lookups each use a different
+// thread, and the system or the program runs out of threads.
+
+var threadLimit = make(chan struct{}, 500)
+
+func acquireThread() {
+	threadLimit <- struct{}{}
+}
+
+func releaseThread() {
+	<-threadLimit
+}