From ea5f6ed6ecab7f8b2648836117f62629b3c2cbb8 Mon Sep 17 00:00:00 2001
From: Nicolas Lacasse <nlacasse@google.com>
Date: Thu, 11 Oct 2018 16:05:44 -0700
Subject: Make Wait() return the sandbox exit status if the sandbox has exited.

It's possible for Start() and Wait() calls to race, if the sandboxed
application is short-lived. If the application finishes before (or during) the
Wait RPC, then Wait will fail.  In practice this looks like "connection
refused" or "EOF" errors when waiting for an RPC response.

This race is especially bad in tests, where we often run "true" inside a
sandbox.

This CL does a best-effort fix, by returning the sandbox exit status as the
container exit status.  In most cases, these are the same.

This fixes the remaining flakes in runsc/container:container_test.

PiperOrigin-RevId: 216777793
Change-Id: I9dfc6e6ec885b106a736055bc7a75b2008dfff7a
---
 runsc/container/container_test.go | 47 +++++++++++++++++++++++++++++++++++++++
 1 file changed, 47 insertions(+)

(limited to 'runsc/container/container_test.go')

diff --git a/runsc/container/container_test.go b/runsc/container/container_test.go
index 7ea99d06b..94572667e 100644
--- a/runsc/container/container_test.go
+++ b/runsc/container/container_test.go
@@ -1725,6 +1725,53 @@ func TestUserLog(t *testing.T) {
 	}
 }
 
+func TestWaitOnExitedSandbox(t *testing.T) {
+	for _, conf := range configs(all...) {
+		t.Logf("Running test with conf: %+v", conf)
+
+		// Run a shell that exits immediately with a non-zero code.
+		const wantExit = 17
+		cmd := fmt.Sprintf("exit %d", wantExit)
+		spec := testutil.NewSpecWithArgs("/bin/sh", "-c", cmd)
+		rootDir, bundleDir, err := testutil.SetupContainer(spec, conf)
+		if err != nil {
+			t.Fatalf("error setting up container: %v", err)
+		}
+		defer os.RemoveAll(rootDir)
+		defer os.RemoveAll(bundleDir)
+
+		// Create and Start the container.
+		c, err := Create(testutil.UniqueContainerID(), spec, conf, bundleDir, "", "", "")
+		if err != nil {
+			t.Fatalf("error creating container: %v", err)
+		}
+		defer c.Destroy()
+		if err := c.Start(conf); err != nil {
+			t.Fatalf("error starting container: %v", err)
+		}
+
+		// Wait for the sandbox to stop running.
+		if err := testutil.Poll(func() error {
+			if c.Sandbox.IsRunning() {
+				return nil
+			}
+			return fmt.Errorf("sandbox still running")
+		}, 10*time.Second); err != nil {
+			t.Fatalf("error waiting for sandbox to exit: %v", err)
+		}
+
+		// Now call Wait.
+		ws, err := c.Wait()
+		if err != nil {
+			t.Fatalf("error waiting on container: %v", err)
+		}
+
+		if got := ws.ExitStatus(); got != wantExit {
+			t.Errorf("got exit status %d, want %d", got, wantExit)
+		}
+	}
+}
+
 // executeSync synchronously executes a new process.
 func (cont *Container) executeSync(args *control.ExecArgs) (syscall.WaitStatus, error) {
 	pid, err := cont.Execute(args)
-- 
cgit v1.2.3