netstack/fdbased: add generic segmentation offload (GSO) support

The linux packet socket can handle GSO packets, so we can segment packets to
64K instead of the MTU which is usually 1500.

Here are numbers for the nginx-1m test:
runsc:		579330.01 [Kbytes/sec] received
runsc-gso:	1794121.66 [Kbytes/sec] received
runc:		2122139.06 [Kbytes/sec] received

and for tcp_benchmark:

$ tcp_benchmark  --duration 15   --ideal
[  4]  0.0-15.0 sec  86647 MBytes  48456 Mbits/sec

$ tcp_benchmark --client --duration 15   --ideal
[  4]  0.0-15.0 sec  2173 MBytes  1214 Mbits/sec

$ tcp_benchmark --client --duration 15   --ideal --gso 65536
[  4]  0.0-15.0 sec  19357 MBytes  10825 Mbits/sec

PiperOrigin-RevId: 240809103
Change-Id: I2637f104db28b5d4c64e1e766c610162a195775a

2 files changed, 13 insertions, 5 deletions

diff --git a/pkg/tcpip/network/ipv6/icmp.go b/pkg/tcpip/network/ipv6/icmp.go
index cfc05d9e1..36d98caef 100644
--- a/pkg/tcpip/network/ipv6/icmp.go
+++ b/pkg/tcpip/network/ipv6/icmp.go
@@ -118,7 +118,7 @@ func (e *endpoint) handleICMP(r *stack.Route, netHeader buffer.View, vv buffer.V
 		defer r.Release()
 		r.LocalAddress = targetAddr
 		pkt.SetChecksum(icmpChecksum(pkt, r.LocalAddress, r.RemoteAddress, buffer.VectorisedView{}))
-		r.WritePacket(hdr, buffer.VectorisedView{}, header.ICMPv6ProtocolNumber, r.DefaultTTL())
+		r.WritePacket(nil /* gso */, hdr, buffer.VectorisedView{}, header.ICMPv6ProtocolNumber, r.DefaultTTL())
 
 		e.linkAddrCache.AddLinkAddress(e.nicid, r.RemoteAddress, r.RemoteLinkAddress)
 
@@ -143,7 +143,7 @@ func (e *endpoint) handleICMP(r *stack.Route, netHeader buffer.View, vv buffer.V
 		copy(pkt, h)
 		pkt.SetType(header.ICMPv6EchoReply)
 		pkt.SetChecksum(icmpChecksum(pkt, r.LocalAddress, r.RemoteAddress, vv))
-		r.WritePacket(hdr, vv, header.ICMPv6ProtocolNumber, r.DefaultTTL())
+		r.WritePacket(nil /* gso */, hdr, vv, header.ICMPv6ProtocolNumber, r.DefaultTTL())
 
 	case header.ICMPv6EchoReply:
 		if len(v) < header.ICMPv6EchoMinimumSize {
@@ -202,7 +202,7 @@ func (*protocol) LinkAddressRequest(addr, localAddr tcpip.Address, linkEP stack.
 		DstAddr:       r.RemoteAddress,
 	})
 
-	return linkEP.WritePacket(r, hdr, buffer.VectorisedView{}, ProtocolNumber)
+	return linkEP.WritePacket(r, nil /* gso */, hdr, buffer.VectorisedView{}, ProtocolNumber)
 }
 
 // ResolveStaticAddress implements stack.LinkAddressResolver.
diff --git a/pkg/tcpip/network/ipv6/ipv6.go b/pkg/tcpip/network/ipv6/ipv6.go
index df3b64c98..9a743ea80 100644
--- a/pkg/tcpip/network/ipv6/ipv6.go
+++ b/pkg/tcpip/network/ipv6/ipv6.go
@@ -83,8 +83,16 @@ func (e *endpoint) MaxHeaderLength() uint16 {
 	return e.linkEP.MaxHeaderLength() + header.IPv6MinimumSize
 }
 
+// GSOMaxSize returns the maximum GSO packet size.
+func (e *endpoint) GSOMaxSize() uint32 {
+	if gso, ok := e.linkEP.(stack.GSOEndpoint); ok {
+		return gso.GSOMaxSize()
+	}
+	return 0
+}
+
 // WritePacket writes a packet to the given destination address and protocol.
-func (e *endpoint) WritePacket(r *stack.Route, hdr buffer.Prependable, payload buffer.VectorisedView, protocol tcpip.TransportProtocolNumber, ttl uint8, loop stack.PacketLooping) *tcpip.Error {
+func (e *endpoint) WritePacket(r *stack.Route, gso *stack.GSO, hdr buffer.Prependable, payload buffer.VectorisedView, protocol tcpip.TransportProtocolNumber, ttl uint8, loop stack.PacketLooping) *tcpip.Error {
 	length := uint16(hdr.UsedLength() + payload.Size())
 	ip := header.IPv6(hdr.Prepend(header.IPv6MinimumSize))
 	ip.Encode(&header.IPv6Fields{
@@ -107,7 +115,7 @@ func (e *endpoint) WritePacket(r *stack.Route, hdr buffer.Prependable, payload b
 	}
 
 	r.Stats().IP.PacketsSent.Increment()
-	return e.linkEP.WritePacket(r, hdr, payload, ProtocolNumber)
+	return e.linkEP.WritePacket(r, gso, hdr, payload, ProtocolNumber)
 }
 
 // HandlePacket is called by the link layer when new ipv6 packets arrive for