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

1 files changed, 11 insertions, 4 deletions

diff --git a/pkg/tcpip/link/rawfile/rawfile_unsafe.go b/pkg/tcpip/link/rawfile/rawfile_unsafe.go
index 5d36ebe57..fe2779125 100644
--- a/pkg/tcpip/link/rawfile/rawfile_unsafe.go
+++ b/pkg/tcpip/link/rawfile/rawfile_unsafe.go
@@ -65,9 +65,9 @@ func NonBlockingWrite(fd int, buf []byte) *tcpip.Error {
 	return nil
 }
 
-// NonBlockingWrite2 writes up to two byte slices to a file descriptor in a
+// NonBlockingWrite3 writes up to three byte slices to a file descriptor in a
 // single syscall. It fails if partial data is written.
-func NonBlockingWrite2(fd int, b1, b2 []byte) *tcpip.Error {
+func NonBlockingWrite3(fd int, b1, b2, b3 []byte) *tcpip.Error {
 	// If the is no second buffer, issue a regular write.
 	if len(b2) == 0 {
 		return NonBlockingWrite(fd, b1)
@@ -75,7 +75,7 @@ func NonBlockingWrite2(fd int, b1, b2 []byte) *tcpip.Error {
 
 	// We have two buffers. Build the iovec that represents them and issue
 	// a writev syscall.
-	iovec := [...]syscall.Iovec{
+	iovec := [3]syscall.Iovec{
 		{
 			Base: &b1[0],
 			Len:  uint64(len(b1)),
@@ -85,8 +85,15 @@ func NonBlockingWrite2(fd int, b1, b2 []byte) *tcpip.Error {
 			Len:  uint64(len(b2)),
 		},
 	}
+	iovecLen := uintptr(2)
 
-	_, _, e := syscall.RawSyscall(syscall.SYS_WRITEV, uintptr(fd), uintptr(unsafe.Pointer(&iovec[0])), uintptr(len(iovec)))
+	if len(b3) > 0 {
+		iovecLen++
+		iovec[2].Base = &b3[0]
+		iovec[2].Len = uint64(len(b3))
+	}
+
+	_, _, e := syscall.RawSyscall(syscall.SYS_WRITEV, uintptr(fd), uintptr(unsafe.Pointer(&iovec[0])), iovecLen)
 	if e != 0 {
 		return TranslateErrno(e)
 	}