wgengine/magicsock: fix bpf fragmentation jump offsets (#16204)

Fragmented datagrams would be processed instead of being dumped right
away. In reality, thse datagrams would be dropped anyway later so there
should functionally not be any change. Additionally, the feature is off
by default.

Closes #16203

Signed-off-by: Claus Lensbøl <claus@tailscale.com>

wgengine/magicsock/magicsock_linux.go

@@ -66,10 +66,10 @@ var (
 		// fragmented, and we don't want to handle reassembly.
 		bpf.LoadAbsolute{Off: 6, Size: 2},
 		// More Fragments bit set means this is part of a fragmented packet.
-		bpf.JumpIf{Cond: bpf.JumpBitsSet, Val: 0x2000, SkipTrue: 7, SkipFalse: 0},
+		bpf.JumpIf{Cond: bpf.JumpBitsSet, Val: 0x2000, SkipTrue: 8, SkipFalse: 0},
 		// Non-zero fragment offset with MF=0 means this is the last
 		// fragment of packet.
-		bpf.JumpIf{Cond: bpf.JumpBitsSet, Val: 0x1fff, SkipTrue: 6, SkipFalse: 0},
+		bpf.JumpIf{Cond: bpf.JumpBitsSet, Val: 0x1fff, SkipTrue: 7, SkipFalse: 0},
 
 		// Load IP header length into X register.
 		bpf.LoadMemShift{Off: 0},

wgengine/magicsock/magicsock_linux_test.go

@@ -9,6 +9,7 @@ import (
 	"net/netip"
 	"testing"
 
+	"golang.org/x/net/bpf"
 	"golang.org/x/sys/cpu"
 	"golang.org/x/sys/unix"
 	"tailscale.com/disco"
@@ -146,3 +147,78 @@ func TestEthernetProto(t *testing.T) {
 		}
 	}
 }
+
+func TestBpfDiscardV4(t *testing.T) {
+	// Good packet as a reference for what should not be rejected
+	udp4Packet := []byte{
+		// IPv4 header
+		0x45, 0x00, 0x00, 0x26, 0x00, 0x00, 0x00, 0x00,
+		0x40, 0x11, 0x00, 0x00,
+		0x7f, 0x00, 0x00, 0x01, // source ip
+		0x7f, 0x00, 0x00, 0x02, // dest ip
+
+		// UDP header
+		0x30, 0x39, // src port
+		0xd4, 0x31, // dest port
+		0x00, 0x12, // length; 8 bytes header + 10 bytes payload = 18 bytes
+		0x00, 0x00, // checksum; unused
+
+		// Payload: disco magic plus 32 bytes for key and 24 bytes for nonce
+		0x54, 0x53, 0xf0, 0x9f, 0x92, 0xac, 0x00, 0x01, 0x02, 0x03,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0x00,
+	}
+
+	vm, err := bpf.NewVM(magicsockFilterV4)
+	if err != nil {
+		t.Fatalf("failed creating BPF VM: %v", err)
+	}
+
+	tests := []struct {
+		name    string
+		replace map[int]byte
+		accept  bool
+	}{
+		{
+			name:    "base accepted datagram",
+			replace: map[int]byte{},
+			accept:  true,
+		},
+		{
+			name: "more fragments",
+			replace: map[int]byte{
+				6: 0x20,
+			},
+			accept: false,
+		},
+		{
+			name: "some fragment",
+			replace: map[int]byte{
+				7: 0x01,
+			},
+			accept: false,
+		},
+	}
+
+	udp4PacketChanged := make([]byte, len(udp4Packet))
+	copy(udp4PacketChanged, udp4Packet)
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			for k, v := range tt.replace {
+				udp4PacketChanged[k] = v
+			}
+			ret, err := vm.Run(udp4PacketChanged)
+			if err != nil {
+				t.Fatalf("BPF VM error: %v", err)
+			}
+
+			if (ret != 0) != tt.accept {
+				t.Errorf("expected accept=%v, got ret=%v", tt.accept, ret)
+			}
+		})
+	}
+}