Add support for TIME_WAIT timeout.

This change adds explicit support for honoring the 2MSL timeout
for sockets in TIME_WAIT state. It also adds support for the
TCP_LINGER2 option that allows modification of the FIN_WAIT2
state timeout duration for a given socket.

It also adds an option to modify the Stack wide TIME_WAIT timeout
but this is only for testing. On Linux this is fixed at 60s.

Further, we also now correctly process RST's in CLOSE_WAIT and
close the socket similar to linux without moving it to error
state.

We also now handle SYN in ESTABLISHED state as per
RFC5961#section-4.1. Earlier we would just drop these SYNs.
Which can result in some tests that pass on linux to fail on
gVisor.

Netstack now honors TIME_WAIT correctly as well as handles the
following cases correctly.

- TCP RSTs in TIME_WAIT are ignored.
- A duplicate TCP FIN during TIME_WAIT extends the TIME_WAIT
  and a dup ACK is sent in response to the FIN as the dup FIN
  indicates potential loss of the original final ACK.
- An out of order segment during TIME_WAIT generates a dup ACK.
- A new SYN w/ a sequence number > the highest sequence number
  in the previous connection closes the TIME_WAIT early and
  opens a new connection.

Further to make the SYN case work correctly the ISN (Initial
Sequence Number) generation for Netstack has been updated to
be as per RFC. Its not a pure random number anymore and follows
the recommendation in https://tools.ietf.org/html/rfc6528#page-3.

The current hash used is not a cryptographically secure hash
function. A separate change will update the hash function used
to Siphash similar to what is used in Linux.

PiperOrigin-RevId: 279106406

1 files changed, 336 insertions, 0 deletions

diff --git a/test/syscalls/linux/socket_inet_loopback.cc b/test/syscalls/linux/socket_inet_loopback.cc
index ab375aaaf..2eeee352e 100644
--- a/test/syscalls/linux/socket_inet_loopback.cc
+++ b/test/syscalls/linux/socket_inet_loopback.cc
@@ -13,6 +13,7 @@
 // limitations under the License.
 
 #include <arpa/inet.h>
+#include <linux/tcp.h>
 #include <netinet/in.h>
 #include <poll.h>
 #include <string.h>
@@ -31,6 +32,7 @@
 #include "gtest/gtest.h"
 #include "absl/memory/memory.h"
 #include "absl/strings/str_cat.h"
+#include "absl/time/clock.h"
 #include "absl/time/time.h"
 #include "test/syscalls/linux/socket_test_util.h"
 #include "test/util/file_descriptor.h"
@@ -267,6 +269,340 @@ TEST_P(SocketInetLoopbackTest, TCPbacklog) {
   }
 }
 
+// TCPFinWait2Test creates a pair of connected sockets then closes one end to
+// trigger FIN_WAIT2 state for the closed endpoint. Then it binds the same local
+// IP/port on a new socket and tries to connect. The connect should fail w/
+// an EADDRINUSE. Then we wait till the FIN_WAIT2 timeout is over and try the
+// connect again with a new socket and this time it should succeed.
+//
+// TCP timers are not S/R today, this can cause this test to be flaky when run
+// under random S/R due to timer being reset on a restore.
+TEST_P(SocketInetLoopbackTest, TCPFinWait2Test_NoRandomSave) {
+  auto const& param = GetParam();
+  TestAddress const& listener = param.listener;
+  TestAddress const& connector = param.connector;
+
+  // Create the listening socket.
+  const FileDescriptor listen_fd = ASSERT_NO_ERRNO_AND_VALUE(
+      Socket(listener.family(), SOCK_STREAM, IPPROTO_TCP));
+  sockaddr_storage listen_addr = listener.addr;
+  ASSERT_THAT(bind(listen_fd.get(), reinterpret_cast<sockaddr*>(&listen_addr),
+                   listener.addr_len),
+              SyscallSucceeds());
+  ASSERT_THAT(listen(listen_fd.get(), SOMAXCONN), SyscallSucceeds());
+
+  // Get the port bound by the listening socket.
+  socklen_t addrlen = listener.addr_len;
+  ASSERT_THAT(getsockname(listen_fd.get(),
+                          reinterpret_cast<sockaddr*>(&listen_addr), &addrlen),
+              SyscallSucceeds());
+
+  uint16_t const port =
+      ASSERT_NO_ERRNO_AND_VALUE(AddrPort(listener.family(), listen_addr));
+
+  // Connect to the listening socket.
+  FileDescriptor conn_fd = ASSERT_NO_ERRNO_AND_VALUE(
+      Socket(connector.family(), SOCK_STREAM, IPPROTO_TCP));
+
+  // Lower FIN_WAIT2 state to 5 seconds for test.
+  constexpr int kTCPLingerTimeout = 5;
+  EXPECT_THAT(setsockopt(conn_fd.get(), IPPROTO_TCP, TCP_LINGER2,
+                         &kTCPLingerTimeout, sizeof(kTCPLingerTimeout)),
+              SyscallSucceedsWithValue(0));
+
+  sockaddr_storage conn_addr = connector.addr;
+  ASSERT_NO_ERRNO(SetAddrPort(connector.family(), &conn_addr, port));
+  ASSERT_THAT(RetryEINTR(connect)(conn_fd.get(),
+                                  reinterpret_cast<sockaddr*>(&conn_addr),
+                                  connector.addr_len),
+              SyscallSucceeds());
+
+  // Accept the connection.
+  auto accepted =
+      ASSERT_NO_ERRNO_AND_VALUE(Accept(listen_fd.get(), nullptr, nullptr));
+
+  // Get the address/port bound by the connecting socket.
+  sockaddr_storage conn_bound_addr;
+  socklen_t conn_addrlen = connector.addr_len;
+  ASSERT_THAT(
+      getsockname(conn_fd.get(), reinterpret_cast<sockaddr*>(&conn_bound_addr),
+                  &conn_addrlen),
+      SyscallSucceeds());
+
+  // close the connecting FD to trigger FIN_WAIT2  on the connected fd.
+  conn_fd.reset();
+
+  // Now bind and connect a new socket.
+  const FileDescriptor conn_fd2 = ASSERT_NO_ERRNO_AND_VALUE(
+      Socket(connector.family(), SOCK_STREAM, IPPROTO_TCP));
+
+  // Disable cooperative saves after this point. As a save between the first
+  // bind/connect and the second one can cause the linger timeout timer to
+  // be restarted causing the final bind/connect to fail.
+  DisableSave ds;
+
+  // TODO(gvisor.dev/issue/1030): Portmanager does not track all 5 tuple
+  //   reservations which causes the bind() to succeed on gVisor but connect
+  //   correctly fails.
+  if (IsRunningOnGvisor()) {
+    ASSERT_THAT(
+        bind(conn_fd2.get(), reinterpret_cast<sockaddr*>(&conn_bound_addr),
+             conn_addrlen),
+        SyscallSucceeds());
+    ASSERT_THAT(RetryEINTR(connect)(conn_fd2.get(),
+                                    reinterpret_cast<sockaddr*>(&conn_addr),
+                                    conn_addrlen),
+                SyscallFailsWithErrno(EADDRINUSE));
+  } else {
+    ASSERT_THAT(
+        bind(conn_fd2.get(), reinterpret_cast<sockaddr*>(&conn_bound_addr),
+             conn_addrlen),
+        SyscallFailsWithErrno(EADDRINUSE));
+  }
+
+  // Sleep for a little over the linger timeout to reduce flakiness in
+  // save/restore tests.
+  absl::SleepFor(absl::Seconds(kTCPLingerTimeout + 1));
+
+  ds.reset();
+
+  if (!IsRunningOnGvisor()) {
+    ASSERT_THAT(
+        bind(conn_fd2.get(), reinterpret_cast<sockaddr*>(&conn_bound_addr),
+             conn_addrlen),
+        SyscallSucceeds());
+  }
+  ASSERT_THAT(RetryEINTR(connect)(conn_fd2.get(),
+                                  reinterpret_cast<sockaddr*>(&conn_addr),
+                                  conn_addrlen),
+              SyscallSucceeds());
+}
+
+// TCPLinger2TimeoutAfterClose creates a pair of connected sockets
+// then closes one end to trigger FIN_WAIT2 state for the closed endpont.
+// It then sleeps for the TCP_LINGER2 timeout and verifies that bind/
+// connecting the same address succeeds.
+//
+// TCP timers are not S/R today, this can cause this test to be flaky when run
+// under random S/R due to timer being reset on a restore.
+TEST_P(SocketInetLoopbackTest, TCPLinger2TimeoutAfterClose_NoRandomSave) {
+  auto const& param = GetParam();
+  TestAddress const& listener = param.listener;
+  TestAddress const& connector = param.connector;
+
+  // Create the listening socket.
+  const FileDescriptor listen_fd = ASSERT_NO_ERRNO_AND_VALUE(
+      Socket(listener.family(), SOCK_STREAM, IPPROTO_TCP));
+  sockaddr_storage listen_addr = listener.addr;
+  ASSERT_THAT(bind(listen_fd.get(), reinterpret_cast<sockaddr*>(&listen_addr),
+                   listener.addr_len),
+              SyscallSucceeds());
+  ASSERT_THAT(listen(listen_fd.get(), SOMAXCONN), SyscallSucceeds());
+
+  // Get the port bound by the listening socket.
+  socklen_t addrlen = listener.addr_len;
+  ASSERT_THAT(getsockname(listen_fd.get(),
+                          reinterpret_cast<sockaddr*>(&listen_addr), &addrlen),
+              SyscallSucceeds());
+
+  uint16_t const port =
+      ASSERT_NO_ERRNO_AND_VALUE(AddrPort(listener.family(), listen_addr));
+
+  // Connect to the listening socket.
+  FileDescriptor conn_fd = ASSERT_NO_ERRNO_AND_VALUE(
+      Socket(connector.family(), SOCK_STREAM, IPPROTO_TCP));
+
+  sockaddr_storage conn_addr = connector.addr;
+  ASSERT_NO_ERRNO(SetAddrPort(connector.family(), &conn_addr, port));
+  ASSERT_THAT(RetryEINTR(connect)(conn_fd.get(),
+                                  reinterpret_cast<sockaddr*>(&conn_addr),
+                                  connector.addr_len),
+              SyscallSucceeds());
+
+  // Accept the connection.
+  auto accepted =
+      ASSERT_NO_ERRNO_AND_VALUE(Accept(listen_fd.get(), nullptr, nullptr));
+
+  // Get the address/port bound by the connecting socket.
+  sockaddr_storage conn_bound_addr;
+  socklen_t conn_addrlen = connector.addr_len;
+  ASSERT_THAT(
+      getsockname(conn_fd.get(), reinterpret_cast<sockaddr*>(&conn_bound_addr),
+                  &conn_addrlen),
+      SyscallSucceeds());
+
+  constexpr int kTCPLingerTimeout = 5;
+  EXPECT_THAT(setsockopt(conn_fd.get(), IPPROTO_TCP, TCP_LINGER2,
+                         &kTCPLingerTimeout, sizeof(kTCPLingerTimeout)),
+              SyscallSucceedsWithValue(0));
+
+  // close the connecting FD to trigger FIN_WAIT2  on the connected fd.
+  conn_fd.reset();
+
+  absl::SleepFor(absl::Seconds(kTCPLingerTimeout + 1));
+
+  // Now bind and connect a new socket and verify that we can immediately
+  // rebind the address bound by the conn_fd as it never entered TIME_WAIT.
+  const FileDescriptor conn_fd2 = ASSERT_NO_ERRNO_AND_VALUE(
+      Socket(connector.family(), SOCK_STREAM, IPPROTO_TCP));
+
+  ASSERT_THAT(bind(conn_fd2.get(),
+                   reinterpret_cast<sockaddr*>(&conn_bound_addr), conn_addrlen),
+              SyscallSucceeds());
+  ASSERT_THAT(RetryEINTR(connect)(conn_fd2.get(),
+                                  reinterpret_cast<sockaddr*>(&conn_addr),
+                                  conn_addrlen),
+              SyscallSucceeds());
+}
+
+// TCPResetAfterClose creates a pair of connected sockets then closes
+// one end to trigger FIN_WAIT2 state for the closed endpoint verifies
+// that we generate RSTs for any new data after the socket is fully
+// closed.
+TEST_P(SocketInetLoopbackTest, TCPResetAfterClose) {
+  auto const& param = GetParam();
+  TestAddress const& listener = param.listener;
+  TestAddress const& connector = param.connector;
+
+  // Create the listening socket.
+  const FileDescriptor listen_fd = ASSERT_NO_ERRNO_AND_VALUE(
+      Socket(listener.family(), SOCK_STREAM, IPPROTO_TCP));
+  sockaddr_storage listen_addr = listener.addr;
+  ASSERT_THAT(bind(listen_fd.get(), reinterpret_cast<sockaddr*>(&listen_addr),
+                   listener.addr_len),
+              SyscallSucceeds());
+  ASSERT_THAT(listen(listen_fd.get(), SOMAXCONN), SyscallSucceeds());
+
+  // Get the port bound by the listening socket.
+  socklen_t addrlen = listener.addr_len;
+  ASSERT_THAT(getsockname(listen_fd.get(),
+                          reinterpret_cast<sockaddr*>(&listen_addr), &addrlen),
+              SyscallSucceeds());
+
+  uint16_t const port =
+      ASSERT_NO_ERRNO_AND_VALUE(AddrPort(listener.family(), listen_addr));
+
+  // Connect to the listening socket.
+  FileDescriptor conn_fd = ASSERT_NO_ERRNO_AND_VALUE(
+      Socket(connector.family(), SOCK_STREAM, IPPROTO_TCP));
+
+  sockaddr_storage conn_addr = connector.addr;
+  ASSERT_NO_ERRNO(SetAddrPort(connector.family(), &conn_addr, port));
+  ASSERT_THAT(RetryEINTR(connect)(conn_fd.get(),
+                                  reinterpret_cast<sockaddr*>(&conn_addr),
+                                  connector.addr_len),
+              SyscallSucceeds());
+
+  // Accept the connection.
+  auto accepted =
+      ASSERT_NO_ERRNO_AND_VALUE(Accept(listen_fd.get(), nullptr, nullptr));
+
+  // close the connecting FD to trigger FIN_WAIT2  on the connected fd.
+  conn_fd.reset();
+
+  int data = 1234;
+
+  // Now send data which should trigger a RST as the other end should
+  // have timed out and closed the socket.
+  EXPECT_THAT(RetryEINTR(send)(accepted.get(), &data, sizeof(data), 0),
+              SyscallSucceeds());
+  // Sleep for a shortwhile to get a RST back.
+  absl::SleepFor(absl::Seconds(1));
+
+  // Try writing again and we should get an EPIPE back.
+  EXPECT_THAT(RetryEINTR(send)(accepted.get(), &data, sizeof(data), 0),
+              SyscallFailsWithErrno(EPIPE));
+
+  // Trying to read should return zero as the other end did send
+  // us a FIN. We do it twice to verify that the RST does not cause an
+  // ECONNRESET on the read after EOF has been read by applicaiton.
+  EXPECT_THAT(RetryEINTR(recv)(accepted.get(), &data, sizeof(data), 0),
+              SyscallSucceedsWithValue(0));
+  EXPECT_THAT(RetryEINTR(recv)(accepted.get(), &data, sizeof(data), 0),
+              SyscallSucceedsWithValue(0));
+}
+
+// This test is disabled under random save as the the restore run
+// results in the stack.Seed() being different which can cause
+// sequence number of final connect to be one that is considered
+// old and can cause the test to be flaky.
+TEST_P(SocketInetLoopbackTest, TCPTimeWaitTest_NoRandomSave) {
+  auto const& param = GetParam();
+  TestAddress const& listener = param.listener;
+  TestAddress const& connector = param.connector;
+
+  // Create the listening socket.
+  const FileDescriptor listen_fd = ASSERT_NO_ERRNO_AND_VALUE(
+      Socket(listener.family(), SOCK_STREAM, IPPROTO_TCP));
+  sockaddr_storage listen_addr = listener.addr;
+  ASSERT_THAT(bind(listen_fd.get(), reinterpret_cast<sockaddr*>(&listen_addr),
+                   listener.addr_len),
+              SyscallSucceeds());
+  ASSERT_THAT(listen(listen_fd.get(), SOMAXCONN), SyscallSucceeds());
+
+  // Get the port bound by the listening socket.
+  socklen_t addrlen = listener.addr_len;
+  ASSERT_THAT(getsockname(listen_fd.get(),
+                          reinterpret_cast<sockaddr*>(&listen_addr), &addrlen),
+              SyscallSucceeds());
+
+  uint16_t const port =
+      ASSERT_NO_ERRNO_AND_VALUE(AddrPort(listener.family(), listen_addr));
+
+  // Connect to the listening socket.
+  FileDescriptor conn_fd = ASSERT_NO_ERRNO_AND_VALUE(
+      Socket(connector.family(), SOCK_STREAM, IPPROTO_TCP));
+
+  // We disable saves after this point as a S/R causes the netstack seed
+  // to be regenerated which changes what ports/ISN is picked for a given
+  // tuple (src ip,src port, dst ip, dst port). This can cause the final
+  // SYN to use a sequence number that looks like one from the current
+  // connection in TIME_WAIT and will not be accepted causing the test
+  // to timeout.
+  //
+  // TODO(gvisor.dev/issue/940): S/R portSeed/portHint
+  DisableSave ds;
+  sockaddr_storage conn_addr = connector.addr;
+  ASSERT_NO_ERRNO(SetAddrPort(connector.family(), &conn_addr, port));
+  ASSERT_THAT(RetryEINTR(connect)(conn_fd.get(),
+                                  reinterpret_cast<sockaddr*>(&conn_addr),
+                                  connector.addr_len),
+              SyscallSucceeds());
+
+  // Accept the connection.
+  auto accepted =
+      ASSERT_NO_ERRNO_AND_VALUE(Accept(listen_fd.get(), nullptr, nullptr));
+
+  // Get the address/port bound by the connecting socket.
+  sockaddr_storage conn_bound_addr;
+  socklen_t conn_addrlen = connector.addr_len;
+  ASSERT_THAT(
+      getsockname(conn_fd.get(), reinterpret_cast<sockaddr*>(&conn_bound_addr),
+                  &conn_addrlen),
+      SyscallSucceeds());
+
+  // close the accept FD to trigger TIME_WAIT on the accepted socket which
+  // should cause the conn_fd to follow CLOSE_WAIT->LAST_ACK->CLOSED instead of
+  // TIME_WAIT.
+  accepted.reset();
+  absl::SleepFor(absl::Seconds(1));
+  conn_fd.reset();
+  absl::SleepFor(absl::Seconds(1));
+
+  // Now bind and connect a new socket and verify that we can immediately
+  // rebind the address bound by the conn_fd as it never entered TIME_WAIT.
+  const FileDescriptor conn_fd2 = ASSERT_NO_ERRNO_AND_VALUE(
+      Socket(connector.family(), SOCK_STREAM, IPPROTO_TCP));
+
+  ASSERT_THAT(bind(conn_fd2.get(),
+                   reinterpret_cast<sockaddr*>(&conn_bound_addr), conn_addrlen),
+              SyscallSucceeds());
+  ASSERT_THAT(RetryEINTR(connect)(conn_fd2.get(),
+                                  reinterpret_cast<sockaddr*>(&conn_addr),
+                                  conn_addrlen),
+              SyscallSucceeds());
+}
+
 INSTANTIATE_TEST_SUITE_P(
     All, SocketInetLoopbackTest,
     ::testing::Values(