Merge release-20201208.0-89-g3ff7324df (automated)

1 files changed, 99 insertions, 11 deletions

diff --git a/pkg/sentry/socket/netstack/netstack.go b/pkg/sentry/socket/netstack/netstack.go
index 23d5cab9c..3f587638f 100644
--- a/pkg/sentry/socket/netstack/netstack.go
+++ b/pkg/sentry/socket/netstack/netstack.go
@@ -1042,10 +1042,7 @@ func getSockOptSocket(t *kernel.Task, s socket.SocketOps, ep commonEndpoint, fam
 		return &v, nil
 
 	case linux.SO_BINDTODEVICE:
-		var v tcpip.BindToDeviceOption
-		if err := ep.GetSockOpt(&v); err != nil {
-			return nil, syserr.TranslateNetstackError(err)
-		}
+		v := ep.SocketOptions().GetBindToDevice()
 		if v == 0 {
 			var b primitive.ByteSlice
 			return &b, nil
@@ -1405,6 +1402,13 @@ func getSockOptIPv6(t *kernel.Task, s socket.SocketOps, ep commonEndpoint, name
 
 		v := primitive.Int32(boolToInt32(ep.SocketOptions().GetReceiveTClass()))
 		return &v, nil
+	case linux.IPV6_RECVERR:
+		if outLen < sizeOfInt32 {
+			return nil, syserr.ErrInvalidArgument
+		}
+
+		v := primitive.Int32(boolToInt32(ep.SocketOptions().GetRecvError()))
+		return &v, nil
 
 	case linux.IPV6_RECVORIGDSTADDR:
 		if outLen < sizeOfInt32 {
@@ -1579,6 +1583,14 @@ func getSockOptIP(t *kernel.Task, s socket.SocketOps, ep commonEndpoint, name in
 		v := primitive.Int32(boolToInt32(ep.SocketOptions().GetReceiveTOS()))
 		return &v, nil
 
+	case linux.IP_RECVERR:
+		if outLen < sizeOfInt32 {
+			return nil, syserr.ErrInvalidArgument
+		}
+
+		v := primitive.Int32(boolToInt32(ep.SocketOptions().GetRecvError()))
+		return &v, nil
+
 	case linux.IP_PKTINFO:
 		if outLen < sizeOfInt32 {
 			return nil, syserr.ErrInvalidArgument
@@ -1789,8 +1801,7 @@ func setSockOptSocket(t *kernel.Task, s socket.SocketOps, ep commonEndpoint, nam
 		}
 		name := string(optVal[:n])
 		if name == "" {
-			v := tcpip.BindToDeviceOption(0)
-			return syserr.TranslateNetstackError(ep.SetSockOpt(&v))
+			return syserr.TranslateNetstackError(ep.SocketOptions().SetBindToDevice(0))
 		}
 		s := t.NetworkContext()
 		if s == nil {
@@ -1798,8 +1809,7 @@ func setSockOptSocket(t *kernel.Task, s socket.SocketOps, ep commonEndpoint, nam
 		}
 		for nicID, nic := range s.Interfaces() {
 			if nic.Name == name {
-				v := tcpip.BindToDeviceOption(nicID)
-				return syserr.TranslateNetstackError(ep.SetSockOpt(&v))
+				return syserr.TranslateNetstackError(ep.SocketOptions().SetBindToDevice(nicID))
 			}
 		}
 		return syserr.ErrUnknownDevice
@@ -2129,6 +2139,16 @@ func setSockOptIPv6(t *kernel.Task, s socket.SocketOps, ep commonEndpoint, name
 
 		ep.SocketOptions().SetReceiveTClass(v != 0)
 		return nil
+	case linux.IPV6_RECVERR:
+		if len(optVal) == 0 {
+			return nil
+		}
+		v, err := parseIntOrChar(optVal)
+		if err != nil {
+			return err
+		}
+		ep.SocketOptions().SetRecvError(v != 0)
+		return nil
 
 	case linux.IP6T_SO_SET_REPLACE:
 		if len(optVal) < linux.SizeOfIP6TReplace {
@@ -2317,6 +2337,17 @@ func setSockOptIP(t *kernel.Task, s socket.SocketOps, ep commonEndpoint, name in
 		ep.SocketOptions().SetReceiveTOS(v != 0)
 		return nil
 
+	case linux.IP_RECVERR:
+		if len(optVal) == 0 {
+			return nil
+		}
+		v, err := parseIntOrChar(optVal)
+		if err != nil {
+			return err
+		}
+		ep.SocketOptions().SetRecvError(v != 0)
+		return nil
+
 	case linux.IP_PKTINFO:
 		if len(optVal) == 0 {
 			return nil
@@ -2386,7 +2417,6 @@ func setSockOptIP(t *kernel.Task, s socket.SocketOps, ep commonEndpoint, name in
 		linux.IP_NODEFRAG,
 		linux.IP_OPTIONS,
 		linux.IP_PASSSEC,
-		linux.IP_RECVERR,
 		linux.IP_RECVFRAGSIZE,
 		linux.IP_RECVOPTS,
 		linux.IP_RECVTTL,
@@ -2462,7 +2492,6 @@ func emitUnimplementedEventIPv6(t *kernel.Task, name int) {
 		linux.IPV6_MULTICAST_IF,
 		linux.IPV6_MULTICAST_LOOP,
 		linux.IPV6_RECVDSTOPTS,
-		linux.IPV6_RECVERR,
 		linux.IPV6_RECVFRAGSIZE,
 		linux.IPV6_RECVHOPLIMIT,
 		linux.IPV6_RECVHOPOPTS,
@@ -2496,7 +2525,6 @@ func emitUnimplementedEventIP(t *kernel.Task, name int) {
 		linux.IP_PKTINFO,
 		linux.IP_PKTOPTIONS,
 		linux.IP_MTU_DISCOVER,
-		linux.IP_RECVERR,
 		linux.IP_RECVTTL,
 		linux.IP_RECVTOS,
 		linux.IP_MTU,
@@ -2772,6 +2800,8 @@ func (s *socketOpsCommon) controlMessages() socket.ControlMessages {
 		IP: socket.IPControlMessages{
 			HasTimestamp:       s.readCM.HasTimestamp && s.sockOptTimestamp,
 			Timestamp:          s.readCM.Timestamp,
+			HasInq:             s.readCM.HasInq,
+			Inq:                s.readCM.Inq,
 			HasTOS:             s.readCM.HasTOS,
 			TOS:                s.readCM.TOS,
 			HasTClass:          s.readCM.HasTClass,
@@ -2779,6 +2809,7 @@ func (s *socketOpsCommon) controlMessages() socket.ControlMessages {
 			HasIPPacketInfo:    s.readCM.HasIPPacketInfo,
 			PacketInfo:         s.readCM.PacketInfo,
 			OriginalDstAddress: s.readCM.OriginalDstAddress,
+			SockErr:            s.readCM.SockErr,
 		},
 	}
 }
@@ -2795,9 +2826,66 @@ func (s *socketOpsCommon) updateTimestamp() {
 	}
 }
 
+// dequeueErr is analogous to net/core/skbuff.c:sock_dequeue_err_skb().
+func (s *socketOpsCommon) dequeueErr() *tcpip.SockError {
+	so := s.Endpoint.SocketOptions()
+	err := so.DequeueErr()
+	if err == nil {
+		return nil
+	}
+
+	// Update socket error to reflect ICMP errors in queue.
+	if nextErr := so.PeekErr(); nextErr != nil && nextErr.ErrOrigin.IsICMPErr() {
+		so.SetLastError(nextErr.Err)
+	} else if err.ErrOrigin.IsICMPErr() {
+		so.SetLastError(nil)
+	}
+	return err
+}
+
+// addrFamilyFromNetProto returns the address family identifier for the given
+// network protocol.
+func addrFamilyFromNetProto(net tcpip.NetworkProtocolNumber) int {
+	switch net {
+	case header.IPv4ProtocolNumber:
+		return linux.AF_INET
+	case header.IPv6ProtocolNumber:
+		return linux.AF_INET6
+	default:
+		panic(fmt.Sprintf("invalid net proto for addr family inference: %d", net))
+	}
+}
+
+// recvErr handles MSG_ERRQUEUE for recvmsg(2).
+// This is analogous to net/ipv4/ip_sockglue.c:ip_recv_error().
+func (s *socketOpsCommon) recvErr(t *kernel.Task, dst usermem.IOSequence) (int, int, linux.SockAddr, uint32, socket.ControlMessages, *syserr.Error) {
+	sockErr := s.dequeueErr()
+	if sockErr == nil {
+		return 0, 0, nil, 0, socket.ControlMessages{}, syserr.ErrTryAgain
+	}
+
+	// The payload of the original packet that caused the error is passed as
+	// normal data via msg_iovec.  -- recvmsg(2)
+	msgFlags := linux.MSG_ERRQUEUE
+	if int(dst.NumBytes()) < len(sockErr.Payload) {
+		msgFlags |= linux.MSG_TRUNC
+	}
+	n, err := dst.CopyOut(t, sockErr.Payload)
+
+	// The original destination address of the datagram that caused the error is
+	// supplied via msg_name.  -- recvmsg(2)
+	dstAddr, dstAddrLen := socket.ConvertAddress(addrFamilyFromNetProto(sockErr.NetProto), sockErr.Dst)
+	cmgs := socket.ControlMessages{IP: socket.NewIPControlMessages(s.family, tcpip.ControlMessages{SockErr: sockErr})}
+	return n, msgFlags, dstAddr, dstAddrLen, cmgs, syserr.FromError(err)
+}
+
 // RecvMsg implements the linux syscall recvmsg(2) for sockets backed by
 // tcpip.Endpoint.
 func (s *socketOpsCommon) RecvMsg(t *kernel.Task, dst usermem.IOSequence, flags int, haveDeadline bool, deadline ktime.Time, senderRequested bool, controlDataLen uint64) (n int, msgFlags int, senderAddr linux.SockAddr, senderAddrLen uint32, controlMessages socket.ControlMessages, err *syserr.Error) {
+	if flags&linux.MSG_ERRQUEUE != 0 {
+		return s.recvErr(t, dst)
+	}
+
 	trunc := flags&linux.MSG_TRUNC != 0
 	peek := flags&linux.MSG_PEEK != 0
 	dontWait := flags&linux.MSG_DONTWAIT != 0