Remove linux-compat loopback hacks from packet endpoint

Previously, gVisor did not represent loopback devices as an ethernet
device as Linux does. To maintain Linux API compatibility for packet
sockets, a workaround was used to add an ethernet header if a link
header was not already present in the packet buffer delivered to a
packet endpoint.

However, this workaround is a bug for non-ethernet based interfaces; not
all links use an ethernet header (e.g. pure L3/TUN interfaces).

As of 3b4bb947517d0d9010120aaa1c3989fd6abf278e, gVisor represents
loopback devices as an ethernet-based device so this workaround can
now be removed.

BUG: https://fxbug.dev/81592
Updates #6530, #6531.
PiperOrigin-RevId: 395819151

4 files changed, 209 insertions, 112 deletions

diff --git a/pkg/sentry/socket/socket.go b/pkg/sentry/socket/socket.go
index 83b9d9389..841d5bd55 100644
--- a/pkg/sentry/socket/socket.go
+++ b/pkg/sentry/socket/socket.go
@@ -743,6 +743,8 @@ func AddressAndFamily(addr []byte) (tcpip.FullAddress, uint16, *syserr.Error) {
 			return tcpip.FullAddress{}, family, syserr.ErrInvalidArgument
 		}
 		a.UnmarshalUnsafe(addr[:sockAddrLinkSize])
+		// TODO(https://gvisor.dev/issue/6530): Do not assume all interfaces have
+		// an ethernet address.
 		if a.Family != linux.AF_PACKET || a.HardwareAddrLen != header.EthernetAddressSize {
 			return tcpip.FullAddress{}, family, syserr.ErrInvalidArgument
 		}
diff --git a/pkg/tcpip/link/channel/channel.go b/pkg/tcpip/link/channel/channel.go
index d02eea93c..658557d62 100644
--- a/pkg/tcpip/link/channel/channel.go
+++ b/pkg/tcpip/link/channel/channel.go
@@ -28,7 +28,9 @@ import (
 
 // PacketInfo holds all the information about an outbound packet.
 type PacketInfo struct {
-	Pkt   *stack.PacketBuffer
+	Pkt *stack.PacketBuffer
+
+	// TODO(https://gvisor.dev/issue/6537): Remove these fields.
 	Proto tcpip.NetworkProtocolNumber
 	Route stack.RouteInfo
 }
@@ -244,7 +246,10 @@ func (e *Endpoint) WritePacket(r stack.RouteInfo, protocol tcpip.NetworkProtocol
 		Route: r,
 	}
 
-	e.q.Write(p)
+	// Write returns false if the queue is full. A full queue is not an error
+	// from the perspective of a LinkEndpoint so we ignore Write's return
+	// value and always return nil from this method.
+	_ = e.q.Write(p)
 
 	return nil
 }
@@ -292,4 +297,16 @@ func (*Endpoint) AddHeader(tcpip.LinkAddress, tcpip.LinkAddress, tcpip.NetworkPr
 }
 
 // WriteRawPacket implements stack.LinkEndpoint.
-func (*Endpoint) WriteRawPacket(*stack.PacketBuffer) tcpip.Error { return &tcpip.ErrNotSupported{} }
+func (e *Endpoint) WriteRawPacket(pkt *stack.PacketBuffer) tcpip.Error {
+	p := PacketInfo{
+		Pkt:   pkt,
+		Proto: pkt.NetworkProtocolNumber,
+	}
+
+	// Write returns false if the queue is full. A full queue is not an error
+	// from the perspective of a LinkEndpoint so we ignore Write's return
+	// value and always return nil from this method.
+	_ = e.q.Write(p)
+
+	return nil
+}
diff --git a/pkg/tcpip/transport/packet/endpoint.go b/pkg/tcpip/transport/packet/endpoint.go
index 89b4720aa..0554d2f4a 100644
--- a/pkg/tcpip/transport/packet/endpoint.go
+++ b/pkg/tcpip/transport/packet/endpoint.go
@@ -25,7 +25,6 @@
 package packet
 
 import (
-	"fmt"
 	"io"
 	"time"
 
@@ -424,76 +423,39 @@ func (ep *endpoint) HandlePacket(nicID tcpip.NICID, localAddr tcpip.LinkAddress,
 
 	wasEmpty := ep.rcvBufSize == 0
 
-	// Push new packet into receive list and increment the buffer size.
-	var packet packet
+	rcvdPkt := packet{
+		packetInfo: tcpip.LinkPacketInfo{
+			Protocol: netProto,
+			PktType:  pkt.PktType,
+		},
+		senderAddr: tcpip.FullAddress{
+			NIC: nicID,
+		},
+		receivedAt: ep.stack.Clock().Now(),
+	}
+
 	if !pkt.LinkHeader().View().IsEmpty() {
-		// Get info directly from the ethernet header.
 		hdr := header.Ethernet(pkt.LinkHeader().View())
-		packet.senderAddr = tcpip.FullAddress{
-			NIC:  nicID,
-			Addr: tcpip.Address(hdr.SourceAddress()),
-		}
-		packet.packetInfo.Protocol = netProto
-		packet.packetInfo.PktType = pkt.PktType
-	} else {
-		// Guess the would-be ethernet header.
-		packet.senderAddr = tcpip.FullAddress{
-			NIC:  nicID,
-			Addr: tcpip.Address(localAddr),
-		}
-		packet.packetInfo.Protocol = netProto
-		packet.packetInfo.PktType = pkt.PktType
+		rcvdPkt.senderAddr.Addr = tcpip.Address(hdr.SourceAddress())
 	}
 
 	if ep.cooked {
-		// Cooked packets can simply be queued.
-		switch pkt.PktType {
-		case tcpip.PacketHost:
-			packet.data = pkt.Data().ExtractVV()
-		case tcpip.PacketOutgoing:
-			// Strip Link Header.
-			var combinedVV buffer.VectorisedView
-			if v := pkt.NetworkHeader().View(); !v.IsEmpty() {
-				combinedVV.AppendView(v)
-			}
-			if v := pkt.TransportHeader().View(); !v.IsEmpty() {
-				combinedVV.AppendView(v)
-			}
-			combinedVV.Append(pkt.Data().ExtractVV())
-			packet.data = combinedVV
-		default:
-			panic(fmt.Sprintf("unexpected PktType in pkt: %+v", pkt))
+		// Cooked packet endpoints don't include the link-headers in received
+		// packets.
+		if v := pkt.NetworkHeader().View(); !v.IsEmpty() {
+			rcvdPkt.data.AppendView(v)
 		}
-	} else {
-		// Raw packets need their ethernet headers prepended before
-		// queueing.
-		var linkHeader buffer.View
-		if pkt.PktType != tcpip.PacketOutgoing {
-			if pkt.LinkHeader().View().IsEmpty() {
-				// We weren't provided with an actual ethernet header,
-				// so fake one.
-				ethFields := header.EthernetFields{
-					SrcAddr: tcpip.LinkAddress([]byte{0x00, 0x00, 0x00, 0x00, 0x00, 0x00}),
-					DstAddr: localAddr,
-					Type:    netProto,
-				}
-				fakeHeader := make(header.Ethernet, header.EthernetMinimumSize)
-				fakeHeader.Encode(&ethFields)
-				linkHeader = buffer.View(fakeHeader)
-			} else {
-				linkHeader = append(buffer.View(nil), pkt.LinkHeader().View()...)
-			}
-			combinedVV := linkHeader.ToVectorisedView()
-			combinedVV.Append(pkt.Data().ExtractVV())
-			packet.data = combinedVV
-		} else {
-			packet.data = buffer.NewVectorisedView(pkt.Size(), pkt.Views())
+		if v := pkt.TransportHeader().View(); !v.IsEmpty() {
+			rcvdPkt.data.AppendView(v)
 		}
+		rcvdPkt.data.Append(pkt.Data().ExtractVV())
+	} else {
+		// Raw packet endpoints include link-headers in received packets.
+		rcvdPkt.data = buffer.NewVectorisedView(pkt.Size(), pkt.Views())
 	}
-	packet.receivedAt = ep.stack.Clock().Now()
 
-	ep.rcvList.PushBack(&packet)
-	ep.rcvBufSize += packet.data.Size()
+	ep.rcvList.PushBack(&rcvdPkt)
+	ep.rcvBufSize += rcvdPkt.data.Size()
 
 	ep.rcvMu.Unlock()
 	ep.stats.PacketsReceived.Increment()
diff --git a/test/syscalls/linux/tuntap.cc b/test/syscalls/linux/tuntap.cc
index 7c9c5c870..956208bee 100644
--- a/test/syscalls/linux/tuntap.cc
+++ b/test/syscalls/linux/tuntap.cc
@@ -73,29 +73,14 @@ PosixErrorOr<Link> GetLinkByName(const std::string& name) {
   return PosixError(ENOENT, "interface not found");
 }
 
-struct pihdr {
-  uint16_t pi_flags;
-  uint16_t pi_protocol;
-} __attribute__((packed));
-
-struct ping_pkt {
-  pihdr pi;
-  struct ethhdr eth;
-  struct iphdr ip;
-  struct icmphdr icmp;
+struct ping_ip_pkt {
+  iphdr ip;
+  icmphdr icmp;
   char payload[64];
 } __attribute__((packed));
 
-ping_pkt CreatePingPacket(const uint8_t srcmac[ETH_ALEN], const in_addr_t srcip,
-                          const uint8_t dstmac[ETH_ALEN],
-                          const in_addr_t dstip) {
-  ping_pkt pkt = {};
-
-  pkt.pi.pi_protocol = htons(ETH_P_IP);
-
-  memcpy(pkt.eth.h_dest, dstmac, sizeof(pkt.eth.h_dest));
-  memcpy(pkt.eth.h_source, srcmac, sizeof(pkt.eth.h_source));
-  pkt.eth.h_proto = htons(ETH_P_IP);
+ping_ip_pkt CreatePingIPPacket(const in_addr_t srcip, const in_addr_t dstip) {
+  ping_ip_pkt pkt = {};
 
   pkt.ip.ihl = 5;
   pkt.ip.version = 4;
@@ -122,6 +107,33 @@ ping_pkt CreatePingPacket(const uint8_t srcmac[ETH_ALEN], const in_addr_t srcip,
   return pkt;
 }
 
+struct pihdr {
+  uint16_t pi_flags;
+  uint16_t pi_protocol;
+} __attribute__((packed));
+
+struct ping_pkt {
+  pihdr pi;
+  ethhdr eth;
+  ping_ip_pkt ip_pkt;
+} __attribute__((packed));
+
+ping_pkt CreatePingPacket(const uint8_t srcmac[ETH_ALEN], const in_addr_t srcip,
+                          const uint8_t dstmac[ETH_ALEN],
+                          const in_addr_t dstip) {
+  ping_pkt pkt = {};
+
+  pkt.pi.pi_protocol = htons(ETH_P_IP);
+
+  memcpy(pkt.eth.h_dest, dstmac, sizeof(pkt.eth.h_dest));
+  memcpy(pkt.eth.h_source, srcmac, sizeof(pkt.eth.h_source));
+  pkt.eth.h_proto = htons(ETH_P_IP);
+
+  pkt.ip_pkt = CreatePingIPPacket(srcip, dstip);
+
+  return pkt;
+}
+
 struct arp_pkt {
   pihdr pi;
   struct ethhdr eth;
@@ -274,13 +286,26 @@ TEST_F(TuntapTest, WriteToDownDevice) {
   EXPECT_THAT(write(fd.get(), buf, sizeof(buf)), SyscallFailsWithErrno(EIO));
 }
 
-PosixErrorOr<FileDescriptor> OpenAndAttachTap(const std::string& dev_name,
-                                              const in_addr_t dev_addr) {
+struct TunTapInterface {
+  FileDescriptor fd;
+  Link link;
+};
+
+PosixErrorOr<TunTapInterface> OpenAndAttachTunTap(const std::string& dev_name,
+                                                  const in_addr_t dev_addr,
+                                                  bool tap, bool no_pi) {
   // Interface creation.
   ASSIGN_OR_RETURN_ERRNO(FileDescriptor fd, Open(kDevNetTun, O_RDWR));
 
   struct ifreq ifr_set = {};
-  ifr_set.ifr_flags = IFF_TAP;
+  if (tap) {
+    ifr_set.ifr_flags |= IFF_TAP;
+  } else {
+    ifr_set.ifr_flags |= IFF_TUN;
+  }
+  if (no_pi) {
+    ifr_set.ifr_flags |= IFF_NO_PI;
+  }
   strncpy(ifr_set.ifr_name, dev_name.c_str(), IFNAMSIZ);
   if (ioctl(fd.get(), TUNSETIFF, &ifr_set) < 0) {
     return PosixError(errno);
@@ -296,13 +321,15 @@ PosixErrorOr<FileDescriptor> OpenAndAttachTap(const std::string& dev_name,
   if (!IsRunningOnGvisor()) {
     // FIXME(b/110961832): gVisor doesn't support setting MAC address on
     // interfaces yet.
-    RETURN_IF_ERRNO(LinkSetMacAddr(link.index, kMacA, sizeof(kMacA)));
+    if (tap) {
+      RETURN_IF_ERRNO(LinkSetMacAddr(link.index, kMacA, sizeof(kMacA)));
+    }
 
     // FIXME(b/110961832): gVisor always creates enabled/up'd interfaces.
     RETURN_IF_ERRNO(LinkChangeFlags(link.index, IFF_UP, IFF_UP));
   }
 
-  return fd;
+  return TunTapInterface{.fd = std::move(fd), .link = std::move(link)};
 }
 
 // This test sets up a TAP device and pings kernel by sending ICMP echo request.
@@ -322,8 +349,9 @@ PosixErrorOr<FileDescriptor> OpenAndAttachTap(const std::string& dev_name,
 TEST_F(TuntapTest, PingKernel) {
   SKIP_IF(!ASSERT_NO_ERRNO_AND_VALUE(HaveCapability(CAP_NET_ADMIN)));
 
-  FileDescriptor fd =
-      ASSERT_NO_ERRNO_AND_VALUE(OpenAndAttachTap(kTapName, kTapIPAddr));
+  const auto& [fd, link] = ASSERT_NO_ERRNO_AND_VALUE(OpenAndAttachTunTap(
+      kTapName, kTapIPAddr, true /* tap */, false /* no_pi */));
+
   ping_pkt ping_req =
       CreatePingPacket(kMacB, kTapPeerIPAddr, kMacA, kTapIPAddr);
   std::string arp_rep =
@@ -365,10 +393,10 @@ TEST_F(TuntapTest, PingKernel) {
 
     // Process ping response packet.
     if (n >= sizeof(ping_pkt) && r.pi.pi_protocol == ping_req.pi.pi_protocol &&
-        r.ping.ip.protocol == ping_req.ip.protocol &&
-        !memcmp(&r.ping.ip.saddr, &ping_req.ip.daddr, kIPLen) &&
-        !memcmp(&r.ping.ip.daddr, &ping_req.ip.saddr, kIPLen) &&
-        r.ping.icmp.type == 0 && r.ping.icmp.code == 0) {
+        r.ping.ip_pkt.ip.protocol == ping_req.ip_pkt.ip.protocol &&
+        !memcmp(&r.ping.ip_pkt.ip.saddr, &ping_req.ip_pkt.ip.daddr, kIPLen) &&
+        !memcmp(&r.ping.ip_pkt.ip.daddr, &ping_req.ip_pkt.ip.saddr, kIPLen) &&
+        r.ping.ip_pkt.icmp.type == 0 && r.ping.ip_pkt.icmp.code == 0) {
       // Ends and passes the test.
       break;
     }
@@ -378,8 +406,8 @@ TEST_F(TuntapTest, PingKernel) {
 TEST_F(TuntapTest, SendUdpTriggersArpResolution) {
   SKIP_IF(!ASSERT_NO_ERRNO_AND_VALUE(HaveCapability(CAP_NET_ADMIN)));
 
-  FileDescriptor fd =
-      ASSERT_NO_ERRNO_AND_VALUE(OpenAndAttachTap(kTapName, kTapIPAddr));
+  const auto& [fd, link] = ASSERT_NO_ERRNO_AND_VALUE(OpenAndAttachTunTap(
+      kTapName, kTapIPAddr, true /* tap */, false /* no_pi */));
 
   // Send a UDP packet to remote.
   int sock = socket(AF_INET, SOCK_DGRAM, IPPROTO_IP);
@@ -433,19 +461,17 @@ TEST_F(TuntapTest, TUNNoPacketInfo) {
   EXPECT_NO_ERRNO(LinkAddLocalAddr(link.index, AF_INET, 24, &dev_ipv4_addr,
                                    sizeof(dev_ipv4_addr)));
 
-  ping_pkt ping_req =
-      CreatePingPacket(kMacB, kTapPeerIPAddr, kMacA, kTapIPAddr);
-  size_t packet_size = sizeof(ping_req) - offsetof(ping_pkt, ip);
+  ping_ip_pkt ping_req = CreatePingIPPacket(kTapPeerIPAddr, kTapIPAddr);
 
   // Send ICMP query
-  EXPECT_THAT(write(fd.get(), &ping_req.ip, packet_size),
-              SyscallSucceedsWithValue(packet_size));
+  EXPECT_THAT(write(fd.get(), &ping_req, sizeof(ping_req)),
+              SyscallSucceedsWithValue(sizeof(ping_req)));
 
   // Receive loop to process inbound packets.
   while (1) {
-    ping_pkt ping_resp = {};
-    EXPECT_THAT(read(fd.get(), &ping_resp.ip, packet_size),
-                SyscallSucceedsWithValue(packet_size));
+    ping_ip_pkt ping_resp = {};
+    EXPECT_THAT(read(fd.get(), &ping_resp, sizeof(ping_req)),
+                SyscallSucceedsWithValue(sizeof(ping_req)));
 
     // Process ping response packet.
     if (!memcmp(&ping_resp.ip.saddr, &ping_req.ip.daddr, kIPLen) &&
@@ -465,8 +491,8 @@ TEST_F(TuntapTest, TCPBlockingConnectFailsArpResolution) {
   FileDescriptor sender =
       ASSERT_NO_ERRNO_AND_VALUE(Socket(AF_INET, SOCK_STREAM, IPPROTO_TCP));
 
-  FileDescriptor fd =
-      ASSERT_NO_ERRNO_AND_VALUE(OpenAndAttachTap(kTapName, kTapIPAddr));
+  const auto tuntap = ASSERT_NO_ERRNO_AND_VALUE(OpenAndAttachTunTap(
+      kTapName, kTapIPAddr, true /* tap */, false /* no_pi */));
 
   sockaddr_in connect_addr = {
       .sin_family = AF_INET,
@@ -487,8 +513,8 @@ TEST_F(TuntapTest, TCPNonBlockingConnectFailsArpResolution) {
   FileDescriptor sender = ASSERT_NO_ERRNO_AND_VALUE(
       Socket(AF_INET, SOCK_STREAM | SOCK_NONBLOCK, IPPROTO_TCP));
 
-  FileDescriptor fd =
-      ASSERT_NO_ERRNO_AND_VALUE(OpenAndAttachTap(kTapName, kTapIPAddr));
+  const auto tuntap = ASSERT_NO_ERRNO_AND_VALUE(OpenAndAttachTunTap(
+      kTapName, kTapIPAddr, true /* tap */, false /* no_pi */));
 
   sockaddr_in connect_addr = {
       .sin_family = AF_INET,
@@ -518,8 +544,8 @@ TEST_F(TuntapTest, TCPNonBlockingConnectFailsArpResolution) {
 TEST_F(TuntapTest, WriteHangBug155928773) {
   SKIP_IF(!ASSERT_NO_ERRNO_AND_VALUE(HaveCapability(CAP_NET_ADMIN)));
 
-  FileDescriptor fd =
-      ASSERT_NO_ERRNO_AND_VALUE(OpenAndAttachTap(kTapName, kTapIPAddr));
+  const auto tuntap = ASSERT_NO_ERRNO_AND_VALUE(OpenAndAttachTunTap(
+      kTapName, kTapIPAddr, true /* tap */, false /* no_pi */));
 
   int sock = socket(AF_INET, SOCK_DGRAM, 0);
   ASSERT_THAT(sock, SyscallSucceeds());
@@ -534,5 +560,95 @@ TEST_F(TuntapTest, WriteHangBug155928773) {
   write(sock, "hello", 5);
 }
 
+// Test that raw packet sockets do not need/include link headers when
+// sending/receiving packets to/from pure L3 (e.g. TUN) interfaces.
+TEST_F(TuntapTest, RawPacketSocket) {
+  SKIP_IF(!ASSERT_NO_ERRNO_AND_VALUE(HaveCapability(CAP_NET_ADMIN)));
+  SKIP_IF(!ASSERT_NO_ERRNO_AND_VALUE(HaveCapability(CAP_NET_RAW)));
+
+  auto [tun, link] = ASSERT_NO_ERRNO_AND_VALUE(OpenAndAttachTunTap(
+      kTunName, kTapIPAddr, false /* tap */, true /* no_pi */));
+  FileDescriptor packet_sock =
+      ASSERT_NO_ERRNO_AND_VALUE(Socket(AF_PACKET, SOCK_RAW, htons(ETH_P_IP)));
+
+  constexpr int kInfiniteTimeout = -1;
+
+  uint8_t hardware_address_length = 0;
+  if (IsRunningOnGvisor()) {
+    // TODO(https://gvisor.dev/issue/6530): Do not assume all interfaces have
+    // an ethernet address.
+    hardware_address_length = ETH_ALEN;
+  }
+
+  {
+    const ping_ip_pkt ping_req = CreatePingIPPacket(kTapPeerIPAddr, kTapIPAddr);
+    ASSERT_THAT(write(tun.get(), &ping_req, sizeof(ping_req)),
+                SyscallSucceedsWithValue(sizeof(ping_req)));
+    // Wait for the packet socket to become readable.
+    pollfd pfd = {
+        .fd = packet_sock.get(),
+        .events = POLLIN,
+    };
+    ASSERT_THAT(RetryEINTR(poll)(&pfd, 1, kInfiniteTimeout),
+                SyscallSucceedsWithValue(1));
+
+    char read_buf[sizeof(ping_req) + 1];
+    struct sockaddr_ll src;
+    socklen_t src_len = sizeof(src);
+    ASSERT_THAT(recvfrom(packet_sock.get(), read_buf, sizeof(read_buf), 0,
+                         reinterpret_cast<struct sockaddr*>(&src), &src_len),
+                SyscallSucceedsWithValue(sizeof(ping_req)));
+    EXPECT_EQ(memcmp(read_buf, &ping_req, sizeof(ping_req)), 0);
+    ASSERT_EQ(src_len, sizeof(src));
+    EXPECT_EQ(src.sll_family, AF_PACKET);
+    EXPECT_EQ(ntohs(src.sll_protocol), ETH_P_IP);
+    EXPECT_EQ(src.sll_ifindex, link.index);
+    EXPECT_EQ(src.sll_pkttype, PACKET_HOST);
+    EXPECT_EQ(src.sll_halen, hardware_address_length);
+    if (IsRunningOnGvisor()) {
+      // TODO(https://gvisor.dev/issue/6531): Check this field for the right
+      // hardware type.
+      EXPECT_EQ(src.sll_hatype, 0);
+    } else {
+      EXPECT_EQ(src.sll_hatype, ARPHRD_NONE);
+    }
+  }
+
+  {
+    const struct sockaddr_ll dest = {
+        .sll_family = AF_PACKET,
+        .sll_protocol = htons(ETH_P_IP),
+        .sll_ifindex = link.index,
+        .sll_halen = hardware_address_length,
+    };
+
+    const ping_ip_pkt ping_req = CreatePingIPPacket(kTapIPAddr, kTapPeerIPAddr);
+    ASSERT_THAT(
+        sendto(packet_sock.get(), &ping_req, sizeof(ping_req), 0,
+               reinterpret_cast<const struct sockaddr*>(&dest), sizeof(dest)),
+        SyscallSucceedsWithValue(sizeof(ping_req)));
+
+    // Loop until we receive the packet we expect - the kernel may send packets
+    // we do not care about.
+    while (true) {
+      // Wait for the TUN interface to become readable.
+      pollfd pfd = {
+          .fd = tun.get(),
+          .events = POLLIN,
+      };
+      ASSERT_THAT(RetryEINTR(poll)(&pfd, 1, kInfiniteTimeout),
+                  SyscallSucceedsWithValue(1));
+
+      char read_buf[sizeof(ping_req) + 1];
+      int n = read(tun.get(), &read_buf, sizeof(read_buf));
+      ASSERT_THAT(n, SyscallSucceeds());
+      if (n == sizeof(ping_req) &&
+          memcmp(read_buf, &ping_req, sizeof(ping_req)) == 0) {
+        break;
+      }
+    }
+  }
+}
+
 }  // namespace testing
 }  // namespace gvisor