Add option for autounmount

Even though filesystem notifications aren't implemented for FUSE, I decided to
try my hand at implementing the autounmount feature (#128). I based it on the
EncFS autounmount code, which records filesystem accesses and checks every X
seconds whether it's idled long enough to unmount.

I've tested the feature locally, but I haven't added any tests for this flag.
I also haven't worked with Go before. So please let me know if there's
anything that should be done differently.

One particular concern: I worked from the assumption that the open files table
is unique per-filesystem. If that's not true, I'll need to add an open file
count and associated lock to the Filesystem type instead.

#265

cli_args.go

@@ -7,6 +7,7 @@ import (
 	"os"
 	"strconv"
 	"strings"
+	"time"
 
 	"github.com/hanwen/go-fuse/fuse"
 	"github.com/rfjakob/gocryptfs/internal/configfile"
@@ -33,6 +34,8 @@ type argContainer struct {
 	// Configuration file name override
 	config             string
 	notifypid, scryptn int
+	// Idle time before autounmount
+	idle time.Duration
 	// Helper variables that are NOT cli options all start with an underscore
 	// _configCustom is true when the user sets a custom config file name.
 	_configCustom bool
@@ -187,6 +190,11 @@ func parseCliOpts() (args argContainer) {
 		"successful mount - used internally for daemonization")
 	flagSet.IntVar(&args.scryptn, "scryptn", configfile.ScryptDefaultLogN, "scrypt cost parameter logN. Possible values: 10-28. "+
 		"A lower value speeds up mounting and reduces its memory needs, but makes the password susceptible to brute-force attacks")
+
+	flagSet.DurationVar(&args.idle, "i", 0, "Alias for -idle")
+	flagSet.DurationVar(&args.idle, "idle", 0, "Auto-unmount after specified idle duration (ignored in reverse mode). "+
+		"Durations are specified like \"500s\" or \"2h45m\". 0 means stay mounted indefinitely.")
+
 	var dummyString string
 	flagSet.StringVar(&dummyString, "o", "", "For compatibility with mount(1), options can be also passed as a comma-separated list to -o on the end.")
 	// Actual parsing
@@ -247,6 +255,10 @@ func parseCliOpts() (args argContainer) {
 		tlog.Fatal.Printf("The options -extpass and -trezor cannot be used at the same time")
 		os.Exit(exitcodes.Usage)
 	}
+	if args.idle < 0 {
+		tlog.Fatal.Printf("Idle timeout cannot be less than 0")
+		os.Exit(exitcodes.Usage)
+	}
 	return args
 }
 

help.go

@@ -19,6 +19,7 @@ func helpShort() {
 Common Options (use -hh to show all):
   -aessiv            Use AES-SIV encryption (with -init)
   -allow_other       Allow other users to access the mount
+  -i, -idle          Unmount automatically after specified idle duration
   -config            Custom path to config file
   -ctlsock           Create control socket at location
   -extpass           Call external program to prompt for the password

internal/fusefrontend/fs.go

@@ -47,6 +47,11 @@ type FS struct {
 	// "gocryptfs -fsck" reads from the channel to also catch these transparently-
 	// mitigated corruptions.
 	MitigatedCorruptions chan string
+	// Track accesses to the filesystem so that we can know when to autounmount.
+	// An access is considered to have happened on every call to encryptPath,
+	// which is called as part of every filesystem operation.
+	// (This flag uses a uint32 so that it can be reset with CompareAndSwapUint32.)
+	AccessedSinceLastCheck uint32
 }
 
 var _ pathfs.FileSystem = &FS{} // Verify that interface is implemented.

internal/fusefrontend/names.go

@@ -60,9 +60,15 @@ func (fs *FS) openBackingDir(relPath string) (dirfd int, cName string, err error
 
 // encryptPath - encrypt relative plaintext path
 func (fs *FS) encryptPath(plainPath string) (string, error) {
+	if plainPath != "" { // Empty path gets encrypted all the time without actual file accesses.
+		fs.AccessedSinceLastCheck = 1
+	} else { // Empty string gets encrypted as empty string
+		return plainPath, nil
+	}
 	if fs.args.PlaintextNames {
 		return plainPath, nil
 	}
+
 	fs.dirIVLock.RLock()
 	cPath, err := fs.nameTransform.EncryptPathDirIV(plainPath, fs.args.Cipherdir)
 	tlog.Debug.Printf("encryptPath '%s' -> '%s' (err: %v)", plainPath, cPath, err)

internal/openfiletable/open_file_table.go

@@ -112,3 +112,11 @@ func (c *countingMutex) Lock() {
 func WriteOpCount() uint64 {
 	return atomic.LoadUint64(&t.writeOpCount)
 }
+
+// CountOpenFiles returns how many entries are currently in the table
+// in a threadsafe manner.
+func CountOpenFiles() int {
+	t.Lock()
+	defer t.Unlock()
+	return len(t.entries)
+}

mount.go

@@ -5,6 +5,7 @@ import (
 	"fmt"
 	"log"
 	"log/syslog"
+	"math"
 	"net"
 	"os"
 	"os/exec"
@@ -14,6 +15,7 @@ import (
 	"runtime"
 	"runtime/debug"
 	"strings"
+	"sync/atomic"
 	"syscall"
 	"time"
 
@@ -29,6 +31,7 @@ import (
 	"github.com/rfjakob/gocryptfs/internal/fusefrontend"
 	"github.com/rfjakob/gocryptfs/internal/fusefrontend_reverse"
 	"github.com/rfjakob/gocryptfs/internal/nametransform"
+	"github.com/rfjakob/gocryptfs/internal/openfiletable"
 	"github.com/rfjakob/gocryptfs/internal/tlog"
 )
 
@@ -98,7 +101,7 @@ func doMount(args *argContainer) {
 	fs, wipeKeys := initFuseFrontend(args)
 	// Initialize go-fuse FUSE server
 	srv := initGoFuse(fs, args)
-	// Try to wipe secrect keys from memory after unmount
+	// Try to wipe secret keys from memory after unmount
 	defer wipeKeys()
 
 	tlog.Info.Println(tlog.ColorGreen + "Filesystem mounted and ready." + tlog.ColorReset)
@@ -137,10 +140,49 @@ func doMount(args *argContainer) {
 	// Return memory that was allocated for scrypt (64M by default!) and other
 	// stuff that is no longer needed to the OS
 	debug.FreeOSMemory()
+	// Set up autounmount, if requested.
+	if args.idle > 0 && !args.reverse {
+		// Not being in reverse mode means we always have a forward file system.
+		fwdFs := fs.(*fusefrontend.FS)
+		go idleMonitor(args.idle, fwdFs, srv, args.mountpoint)
+	}
 	// Jump into server loop. Returns when it gets an umount request from the kernel.
 	srv.Serve()
 }
 
+// Based on the EncFS idle monitor:
+// https://github.com/vgough/encfs/blob/1974b417af189a41ffae4c6feb011d2a0498e437/encfs/main.cpp#L851
+// idleMonitor is a function to be run as a thread that checks for
+// filesystem idleness and unmounts if we've been idle for long enough.
+const checksDuringTimeoutPeriod = 4
+
+func idleMonitor(idleTimeout time.Duration, fs *fusefrontend.FS, srv *fuse.Server, mountpoint string) {
+	sleepTimeBetweenChecks := contentenc.MinUint64(
+		uint64(idleTimeout/checksDuringTimeoutPeriod),
+		uint64(2*time.Minute))
+	timeoutCycles := int(math.Ceil(float64(idleTimeout) / float64(sleepTimeBetweenChecks)))
+	idleCount := 0
+	for {
+		// Atomically check whether the access flag is set and reset it to 0 if so.
+		recentAccess := atomic.CompareAndSwapUint32(&fs.AccessedSinceLastCheck, 1, 0)
+		// Any form of current or recent access resets the idle counter.
+		openFileCount := openfiletable.CountOpenFiles()
+		if recentAccess || openFileCount > 0 {
+			idleCount = 0
+		} else {
+			idleCount++
+		}
+		tlog.Debug.Printf(
+			"Checking for idle (recentAccess = %t, open = %d): %s",
+			recentAccess, openFileCount, time.Now().String())
+		if idleCount > 0 && idleCount%timeoutCycles == 0 {
+			tlog.Info.Printf("Filesystem idle; unmounting: %s", mountpoint)
+			unmount(srv, mountpoint)
+		}
+		time.Sleep(time.Duration(sleepTimeBetweenChecks))
+	}
+}
+
 // setOpenFileLimit tries to increase the open file limit to 4096 (the default hard
 // limit on Linux).
 func setOpenFileLimit() {
@@ -379,18 +421,22 @@ func handleSigint(srv *fuse.Server, mountpoint string) {
 	signal.Notify(ch, syscall.SIGTERM)
 	go func() {
 		<-ch
-		err := srv.Unmount()
-		if err != nil {
-			tlog.Warn.Printf("handleSigint: srv.Unmount returned %v", err)
-			if runtime.GOOS == "linux" {
-				// MacOSX does not support lazy unmount
-				tlog.Info.Printf("Trying lazy unmount")
-				cmd := exec.Command("fusermount", "-u", "-z", mountpoint)
-				cmd.Stdout = os.Stdout
-				cmd.Stderr = os.Stderr
-				cmd.Run()
-			}
-		}
+		unmount(srv, mountpoint)
 		os.Exit(exitcodes.SigInt)
 	}()
 }
+
+func unmount(srv *fuse.Server, mountpoint string) {
+	err := srv.Unmount()
+	if err != nil {
+		tlog.Warn.Printf("unmount: srv.Unmount returned %v", err)
+		if runtime.GOOS == "linux" {
+			// MacOSX does not support lazy unmount
+			tlog.Info.Printf("Trying lazy unmount")
+			cmd := exec.Command("fusermount", "-u", "-z", mountpoint)
+			cmd.Stdout = os.Stdout
+			cmd.Stderr = os.Stderr
+			cmd.Run()
+		}
+	}
+}