Merge release-20200622.1-33-g8dbeac53c (automated)

6 files changed, 121 insertions, 72 deletions

diff --git a/pkg/abi/linux/linux_abi_autogen_unsafe.go b/pkg/abi/linux/linux_abi_autogen_unsafe.go
index 868b27001..ac656f14c 100644
--- a/pkg/abi/linux/linux_abi_autogen_unsafe.go
+++ b/pkg/abi/linux/linux_abi_autogen_unsafe.go
@@ -138,7 +138,7 @@ func (s *Statx) MarshalUnsafe(dst []byte) {
 
 // UnmarshalUnsafe implements marshal.Marshallable.UnmarshalUnsafe.
 func (s *Statx) UnmarshalUnsafe(src []byte) {
-    if s.Ctime.Packed() && s.Mtime.Packed() && s.Atime.Packed() && s.Btime.Packed() {
+    if s.Atime.Packed() && s.Btime.Packed() && s.Ctime.Packed() && s.Mtime.Packed() {
         safecopy.CopyOut(unsafe.Pointer(s), src)
     } else {
         s.UnmarshalBytes(src)
diff --git a/pkg/sentry/socket/netstack/netstack.go b/pkg/sentry/socket/netstack/netstack.go
index c0b63a803..e7d2c83d7 100644
--- a/pkg/sentry/socket/netstack/netstack.go
+++ b/pkg/sentry/socket/netstack/netstack.go
@@ -1169,6 +1169,17 @@ func getSockOptSocket(t *kernel.Task, s socket.SocketOps, ep commonEndpoint, fam
 
 		return int32(v), nil
 
+	case linux.SO_NO_CHECK:
+		if outLen < sizeOfInt32 {
+			return nil, syserr.ErrInvalidArgument
+		}
+
+		v, err := ep.GetSockOptBool(tcpip.NoChecksumOption)
+		if err != nil {
+			return nil, syserr.TranslateNetstackError(err)
+		}
+		return boolToInt32(v), nil
+
 	default:
 		socket.GetSockOptEmitUnimplementedEvent(t, name)
 	}
@@ -1720,6 +1731,14 @@ func setSockOptSocket(t *kernel.Task, s socket.SocketOps, ep commonEndpoint, nam
 
 		return syserr.TranslateNetstackError(ep.SetSockOpt(tcpip.OutOfBandInlineOption(v)))
 
+	case linux.SO_NO_CHECK:
+		if len(optVal) < sizeOfInt32 {
+			return syserr.ErrInvalidArgument
+		}
+
+		v := usermem.ByteOrder.Uint32(optVal)
+		return syserr.TranslateNetstackError(ep.SetSockOptBool(tcpip.NoChecksumOption, v != 0))
+
 	case linux.SO_LINGER:
 		if len(optVal) < linux.SizeOfLinger {
 			return syserr.ErrInvalidArgument
diff --git a/pkg/sentry/socket/socket.go b/pkg/sentry/socket/socket.go
index 6580bd6e9..fcd7f9d7f 100644
--- a/pkg/sentry/socket/socket.go
+++ b/pkg/sentry/socket/socket.go
@@ -407,7 +407,6 @@ func emitUnimplementedEvent(t *kernel.Task, name int) {
 		linux.SO_MARK,
 		linux.SO_MAX_PACING_RATE,
 		linux.SO_NOFCS,
-		linux.SO_NO_CHECK,
 		linux.SO_OOBINLINE,
 		linux.SO_PASSCRED,
 		linux.SO_PASSSEC,
diff --git a/pkg/tcpip/tcpip.go b/pkg/tcpip/tcpip.go
index 4d45dcc42..2be1c107a 100644
--- a/pkg/tcpip/tcpip.go
+++ b/pkg/tcpip/tcpip.go
@@ -585,59 +585,68 @@ type WriteOptions struct {
 type SockOptBool int
 
 const (
-	// BroadcastOption is used by SetSockOpt/GetSockOpt to specify whether
-	// datagram sockets are allowed to send packets to a broadcast address.
+	// BroadcastOption is used by SetSockOptBool/GetSockOptBool to specify
+	// whether datagram sockets are allowed to send packets to a broadcast
+	// address.
 	BroadcastOption SockOptBool = iota
 
-	// CorkOption is used by SetSockOpt/GetSockOpt to specify if data should be
-	// held until segments are full by the TCP transport protocol.
+	// CorkOption is used by SetSockOptBool/GetSockOptBool to specify if
+	// data should be held until segments are full by the TCP transport
+	// protocol.
 	CorkOption
 
-	// DelayOption is used by SetSockOpt/GetSockOpt to specify if data
-	// should be sent out immediately by the transport protocol. For TCP,
-	// it determines if the Nagle algorithm is on or off.
+	// DelayOption is used by SetSockOptBool/GetSockOptBool to specify if
+	// data should be sent out immediately by the transport protocol. For
+	// TCP, it determines if the Nagle algorithm is on or off.
 	DelayOption
 
-	// KeepaliveEnabledOption is used by SetSockOpt/GetSockOpt to specify whether
-	// TCP keepalive is enabled for this socket.
+	// KeepaliveEnabledOption is used by SetSockOptBool/GetSockOptBool to
+	// specify whether TCP keepalive is enabled for this socket.
 	KeepaliveEnabledOption
 
-	// MulticastLoopOption is used by SetSockOpt/GetSockOpt to specify whether
-	// multicast packets sent over a non-loopback interface will be looped back.
+	// MulticastLoopOption is used by SetSockOptBool/GetSockOptBool to
+	// specify whether multicast packets sent over a non-loopback interface
+	// will be looped back.
 	MulticastLoopOption
 
-	// PasscredOption is used by SetSockOpt/GetSockOpt to specify whether
-	// SCM_CREDENTIALS socket control messages are enabled.
+	// NoChecksumOption is used by SetSockOptBool/GetSockOptBool to specify
+	// whether UDP checksum is disabled for this socket.
+	NoChecksumOption
+
+	// PasscredOption is used by SetSockOptBool/GetSockOptBool to specify
+	// whether SCM_CREDENTIALS socket control messages are enabled.
 	//
 	// Only supported on Unix sockets.
 	PasscredOption
 
-	// QuickAckOption is stubbed out in SetSockOpt/GetSockOpt.
+	// QuickAckOption is stubbed out in SetSockOptBool/GetSockOptBool.
 	QuickAckOption
 
-	// ReceiveTClassOption is used by SetSockOpt/GetSockOpt to specify if the
-	// IPV6_TCLASS ancillary message is passed with incoming packets.
+	// ReceiveTClassOption is used by SetSockOptBool/GetSockOptBool to
+	// specify if the IPV6_TCLASS ancillary message is passed with incoming
+	// packets.
 	ReceiveTClassOption
 
-	// ReceiveTOSOption is used by SetSockOpt/GetSockOpt to specify if the TOS
-	// ancillary message is passed with incoming packets.
+	// ReceiveTOSOption is used by SetSockOptBool/GetSockOptBool to specify
+	// if the TOS ancillary message is passed with incoming packets.
 	ReceiveTOSOption
 
-	// ReceiveIPPacketInfoOption is used by {G,S}etSockOptBool to specify
-	// if more inforamtion is provided with incoming packets such
-	// as interface index and address.
+	// ReceiveIPPacketInfoOption is used by SetSockOptBool/GetSockOptBool to
+	// specify if more inforamtion is provided with incoming packets such as
+	// interface index and address.
 	ReceiveIPPacketInfoOption
 
-	// ReuseAddressOption is used by SetSockOpt/GetSockOpt to specify whether Bind()
-	// should allow reuse of local address.
+	// ReuseAddressOption is used by SetSockOptBool/GetSockOptBool to
+	// specify whether Bind() should allow reuse of local address.
 	ReuseAddressOption
 
-	// ReusePortOption is used by SetSockOpt/GetSockOpt to permit multiple sockets
-	// to be bound to an identical socket address.
+	// ReusePortOption is used by SetSockOptBool/GetSockOptBool to permit
+	// multiple sockets to be bound to an identical socket address.
 	ReusePortOption
 
-	// V6OnlyOption is used by {G,S}etSockOptBool to specify whether an IPv6
-	// socket is to be restricted to sending and receiving IPv6 packets only.
+	// V6OnlyOption is used by SetSockOptBool/GetSockOptBool to specify
+	// whether an IPv6 socket is to be restricted to sending and receiving
+	// IPv6 packets only.
 	V6OnlyOption
 )
 
@@ -645,25 +654,27 @@ const (
 type SockOptInt int
 
 const (
-	// KeepaliveCountOption is used by SetSockOpt/GetSockOpt to specify the number
-	// of un-ACKed TCP keepalives that will be sent before the connection is
-	// closed.
+	// KeepaliveCountOption is used by SetSockOptInt/GetSockOptInt to
+	// specify the number of un-ACKed TCP keepalives that will be sent
+	// before the connection is closed.
 	KeepaliveCountOption SockOptInt = iota
 
-	// IPv4TOSOption is used by SetSockOpt/GetSockOpt to specify TOS
+	// IPv4TOSOption is used by SetSockOptInt/GetSockOptInt to specify TOS
 	// for all subsequent outgoing IPv4 packets from the endpoint.
 	IPv4TOSOption
 
-	// IPv6TrafficClassOption is used by SetSockOpt/GetSockOpt to specify TOS
-	// for all subsequent outgoing IPv6 packets from the endpoint.
+	// IPv6TrafficClassOption is used by SetSockOptInt/GetSockOptInt to
+	// specify TOS for all subsequent outgoing IPv6 packets from the
+	// endpoint.
 	IPv6TrafficClassOption
 
-	// MaxSegOption is used by SetSockOpt/GetSockOpt to set/get the current
-	// Maximum Segment Size(MSS) value as specified using the TCP_MAXSEG option.
+	// MaxSegOption is used by SetSockOptInt/GetSockOptInt to set/get the
+	// current Maximum Segment Size(MSS) value as specified using the
+	// TCP_MAXSEG option.
 	MaxSegOption
 
-	// MulticastTTLOption is used by SetSockOpt/GetSockOpt to control the default
-	// TTL value for multicast messages. The default is 1.
+	// MulticastTTLOption is used by SetSockOptInt/GetSockOptInt to control
+	// the default TTL value for multicast messages. The default is 1.
 	MulticastTTLOption
 
 	// ReceiveQueueSizeOption is used in GetSockOptInt to specify that the
@@ -682,21 +693,22 @@ const (
 	// number of unread bytes in the output buffer should be returned.
 	SendQueueSizeOption
 
-	// TTLOption is used by SetSockOpt/GetSockOpt to control the default TTL/hop
-	// limit value for unicast messages. The default is protocol specific.
+	// TTLOption is used by SetSockOptInt/GetSockOptInt to control the
+	// default TTL/hop limit value for unicast messages. The default is
+	// protocol specific.
 	//
 	// A zero value indicates the default.
 	TTLOption
 
-	// TCPSynCountOption is used by SetSockOpt/GetSockOpt to specify the number of
-	// SYN retransmits that TCP should send before aborting the attempt to
-	// connect. It cannot exceed 255.
+	// TCPSynCountOption is used by SetSockOptInt/GetSockOptInt to specify
+	// the number of SYN retransmits that TCP should send before aborting
+	// the attempt to connect. It cannot exceed 255.
 	//
 	// NOTE: This option is currently only stubbed out and is no-op.
 	TCPSynCountOption
 
-	// TCPWindowClampOption is used by SetSockOpt/GetSockOpt to bound the size
-	// of the advertised window to this value.
+	// TCPWindowClampOption is used by SetSockOptInt/GetSockOptInt to bound
+	// the size of the advertised window to this value.
 	//
 	// NOTE: This option is currently only stubed out and is a no-op
 	TCPWindowClampOption
diff --git a/pkg/tcpip/transport/udp/endpoint.go b/pkg/tcpip/transport/udp/endpoint.go
index 8bdc1ee1f..cae29fbff 100644
--- a/pkg/tcpip/transport/udp/endpoint.go
+++ b/pkg/tcpip/transport/udp/endpoint.go
@@ -109,6 +109,7 @@ type endpoint struct {
 	portFlags      ports.Flags
 	bindToDevice   tcpip.NICID
 	broadcast      bool
+	noChecksum     bool
 
 	lastErrorMu sync.Mutex   `state:"nosave"`
 	lastError   *tcpip.Error `state:".(string)"`
@@ -529,7 +530,7 @@ func (e *endpoint) write(p tcpip.Payloader, opts tcpip.WriteOptions) (int64, <-c
 		useDefaultTTL = false
 	}
 
-	if err := sendUDP(route, buffer.View(v).ToVectorisedView(), e.ID.LocalPort, dstPort, ttl, useDefaultTTL, e.sendTOS, e.owner); err != nil {
+	if err := sendUDP(route, buffer.View(v).ToVectorisedView(), e.ID.LocalPort, dstPort, ttl, useDefaultTTL, e.sendTOS, e.owner, e.noChecksum); err != nil {
 		return 0, nil, err
 	}
 	return int64(len(v)), nil, nil
@@ -553,6 +554,11 @@ func (e *endpoint) SetSockOptBool(opt tcpip.SockOptBool, v bool) *tcpip.Error {
 		e.multicastLoop = v
 		e.mu.Unlock()
 
+	case tcpip.NoChecksumOption:
+		e.mu.Lock()
+		e.noChecksum = v
+		e.mu.Unlock()
+
 	case tcpip.ReceiveTOSOption:
 		e.mu.Lock()
 		e.receiveTOS = v
@@ -825,6 +831,12 @@ func (e *endpoint) GetSockOptBool(opt tcpip.SockOptBool) (bool, *tcpip.Error) {
 		e.mu.RUnlock()
 		return v, nil
 
+	case tcpip.NoChecksumOption:
+		e.mu.RLock()
+		v := e.noChecksum
+		e.mu.RUnlock()
+		return v, nil
+
 	case tcpip.ReceiveTOSOption:
 		e.mu.RLock()
 		v := e.receiveTOS
@@ -959,7 +971,7 @@ func (e *endpoint) GetSockOpt(opt interface{}) *tcpip.Error {
 
 // sendUDP sends a UDP segment via the provided network endpoint and under the
 // provided identity.
-func sendUDP(r *stack.Route, data buffer.VectorisedView, localPort, remotePort uint16, ttl uint8, useDefaultTTL bool, tos uint8, owner tcpip.PacketOwner) *tcpip.Error {
+func sendUDP(r *stack.Route, data buffer.VectorisedView, localPort, remotePort uint16, ttl uint8, useDefaultTTL bool, tos uint8, owner tcpip.PacketOwner, noChecksum bool) *tcpip.Error {
 	// Allocate a buffer for the UDP header.
 	hdr := buffer.NewPrependable(header.UDPMinimumSize + int(r.MaxHeaderLength()))
 
@@ -973,8 +985,12 @@ func sendUDP(r *stack.Route, data buffer.VectorisedView, localPort, remotePort u
 		Length:  length,
 	})
 
-	// Only calculate the checksum if offloading isn't supported.
-	if r.Capabilities()&stack.CapabilityTXChecksumOffload == 0 {
+	// Set the checksum field unless TX checksum offload is enabled.
+	// On IPv4, UDP checksum is optional, and a zero value indicates the
+	// transmitter skipped the checksum generation (RFC768).
+	// On IPv6, UDP checksum is not optional (RFC2460 Section 8.1).
+	if r.Capabilities()&stack.CapabilityTXChecksumOffload == 0 &&
+		(!noChecksum || r.NetProto == header.IPv6ProtocolNumber) {
 		xsum := r.PseudoHeaderChecksum(ProtocolNumber, length)
 		for _, v := range data.Views() {
 			xsum = header.Checksum(v, xsum)
diff --git a/pkg/tcpip/transport/udp/udp_state_autogen.go b/pkg/tcpip/transport/udp/udp_state_autogen.go
index 943022c38..50fd0f186 100644
--- a/pkg/tcpip/transport/udp/udp_state_autogen.go
+++ b/pkg/tcpip/transport/udp/udp_state_autogen.go
@@ -73,6 +73,7 @@ func (x *endpoint) StateFields() []string {
 		"portFlags",
 		"bindToDevice",
 		"broadcast",
+		"noChecksum",
 		"lastError",
 		"boundBindToDevice",
 		"boundPortFlags",
@@ -92,7 +93,7 @@ func (x *endpoint) StateSave(m state.Sink) {
 	var rcvBufSizeMax int = x.saveRcvBufSizeMax()
 	m.SaveValue(5, rcvBufSizeMax)
 	var lastError string = x.saveLastError()
-	m.SaveValue(21, lastError)
+	m.SaveValue(22, lastError)
 	m.Save(0, &x.TransportEndpointInfo)
 	m.Save(1, &x.waiterQueue)
 	m.Save(2, &x.uniqueID)
@@ -113,16 +114,17 @@ func (x *endpoint) StateSave(m state.Sink) {
 	m.Save(18, &x.portFlags)
 	m.Save(19, &x.bindToDevice)
 	m.Save(20, &x.broadcast)
-	m.Save(22, &x.boundBindToDevice)
-	m.Save(23, &x.boundPortFlags)
-	m.Save(24, &x.sendTOS)
-	m.Save(25, &x.receiveTOS)
-	m.Save(26, &x.receiveTClass)
-	m.Save(27, &x.receiveIPPacketInfo)
-	m.Save(28, &x.shutdownFlags)
-	m.Save(29, &x.multicastMemberships)
-	m.Save(30, &x.effectiveNetProtos)
-	m.Save(31, &x.owner)
+	m.Save(21, &x.noChecksum)
+	m.Save(23, &x.boundBindToDevice)
+	m.Save(24, &x.boundPortFlags)
+	m.Save(25, &x.sendTOS)
+	m.Save(26, &x.receiveTOS)
+	m.Save(27, &x.receiveTClass)
+	m.Save(28, &x.receiveIPPacketInfo)
+	m.Save(29, &x.shutdownFlags)
+	m.Save(30, &x.multicastMemberships)
+	m.Save(31, &x.effectiveNetProtos)
+	m.Save(32, &x.owner)
 }
 
 func (x *endpoint) StateLoad(m state.Source) {
@@ -146,18 +148,19 @@ func (x *endpoint) StateLoad(m state.Source) {
 	m.Load(18, &x.portFlags)
 	m.Load(19, &x.bindToDevice)
 	m.Load(20, &x.broadcast)
-	m.Load(22, &x.boundBindToDevice)
-	m.Load(23, &x.boundPortFlags)
-	m.Load(24, &x.sendTOS)
-	m.Load(25, &x.receiveTOS)
-	m.Load(26, &x.receiveTClass)
-	m.Load(27, &x.receiveIPPacketInfo)
-	m.Load(28, &x.shutdownFlags)
-	m.Load(29, &x.multicastMemberships)
-	m.Load(30, &x.effectiveNetProtos)
-	m.Load(31, &x.owner)
+	m.Load(21, &x.noChecksum)
+	m.Load(23, &x.boundBindToDevice)
+	m.Load(24, &x.boundPortFlags)
+	m.Load(25, &x.sendTOS)
+	m.Load(26, &x.receiveTOS)
+	m.Load(27, &x.receiveTClass)
+	m.Load(28, &x.receiveIPPacketInfo)
+	m.Load(29, &x.shutdownFlags)
+	m.Load(30, &x.multicastMemberships)
+	m.Load(31, &x.effectiveNetProtos)
+	m.Load(32, &x.owner)
 	m.LoadValue(5, new(int), func(y interface{}) { x.loadRcvBufSizeMax(y.(int)) })
-	m.LoadValue(21, new(string), func(y interface{}) { x.loadLastError(y.(string)) })
+	m.LoadValue(22, new(string), func(y interface{}) { x.loadLastError(y.(string)) })
 	m.AfterLoad(x.afterLoad)
 }