1 files changed, 111 insertions, 0 deletions
diff --git a/tests/cluster/cluster_test.go b/tests/cluster/cluster_test.go
new file mode 100644
index 0000000..2e969ce
--- /dev/null
+++ b/tests/cluster/cluster_test.go
@@ -0,0 +1,111 @@
+// package cluster_test finds out what happens if multiple
+// gocryptfs mounts write to one file concurrently
+// (usually, nothing good).
+//
+// This use case is relevant for HPC clusters.
+package cluster_test
+
+import (
+	"bytes"
+	"math/rand"
+	"os"
+	"sync"
+	"testing"
+
+	"github.com/rfjakob/gocryptfs/v2/tests/test_helpers"
+)
+
+// This test passes on XFS but fails on ext4 and tmpfs!!!
+//
+// Quoting https://lists.samba.org/archive/samba-technical/2019-March/133050.html
+//
+// > It turns out that xfs respects POSIX w.r.t "atomic read/write" and
+// > this is implemented by taking a file-wide shared lock on every
+// > buffered read.
+// > This behavior is unique to XFS on Linux and is not optional.
+// > Other Linux filesystems only guaranty page level atomicity for
+// > buffered read/write.
+//
+// See also:
+// * https://lore.kernel.org/linux-xfs/20190325001044.GA23020@dastard/
+//   Dave Chinner: XFS is the only linux filesystem that provides this behaviour.
+func TestClusterConcurrentRW(t *testing.T) {
+	if os.Getenv("ENABLE_CLUSTER_TEST") != "1" {
+		t.Skipf("This test is disabled by default because it fails unless on XFS.\n" +
+			"Run it like this: ENABLE_CLUSTER_TEST=1 go test\n" +
+			"Choose a backing directory by setting TMPDIR.")
+	}
+
+	const blocksize = 4096
+	const fileSize = 25 * blocksize // 100 kiB
+
+	cDir := test_helpers.InitFS(t)
+	mnt1 := cDir + ".mnt1"
+	mnt2 := cDir + ".mnt2"
+	test_helpers.MountOrFatal(t, cDir, mnt1, "-extpass=echo test", "-wpanic=0")
+	defer test_helpers.UnmountPanic(mnt1)
+	test_helpers.MountOrFatal(t, cDir, mnt2, "-extpass=echo test", "-wpanic=0")
+	defer test_helpers.UnmountPanic(mnt2)
+
+	f1, err := os.Create(mnt1 + "/foo")
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer f1.Close()
+	// Preallocate space
+	_, err = f1.WriteAt(make([]byte, fileSize), 0)
+	if err != nil {
+		t.Fatal(err)
+	}
+	f2, err := os.OpenFile(mnt2+"/foo", os.O_RDWR, 0)
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer f2.Close()
+
+	var wg sync.WaitGroup
+
+	const loops = 10000
+	writeThread := func(f *os.File) {
+		defer wg.Done()
+		buf := make([]byte, blocksize)
+		for i := 0; i < loops; i++ {
+			if t.Failed() {
+				return
+			}
+			off := rand.Int63n(fileSize / blocksize)
+			_, err := f.WriteAt(buf, off)
+			if err != nil {
+				t.Errorf("writeThread iteration %d: WriteAt failed: %v", i, err)
+				return
+			}
+		}
+	}
+	readThread := func(f *os.File) {
+		defer wg.Done()
+		zeroBlock := make([]byte, blocksize)
+		buf := make([]byte, blocksize)
+		for i := 0; i < loops; i++ {
+			if t.Failed() {
+				return
+			}
+			off := rand.Int63n(fileSize / blocksize)
+			_, err := f.ReadAt(buf, off)
+			if err != nil {
+				t.Errorf("readThread iteration %d: ReadAt failed: %v", i, err)
+				return
+			}
+			if !bytes.Equal(buf, zeroBlock) {
+				t.Errorf("readThread iteration %d: data mismatch", i)
+				return
+			}
+		}
+	}
+
+	wg.Add(4)
+	go writeThread(f1)
+	go writeThread(f2)
+	go readThread(f1)
+	go readThread(f2)
+	wg.Wait()
+}