4 files changed, 30 insertions, 18 deletions
diff --git a/pkg/sentry/mm/mm.go b/pkg/sentry/mm/mm.go
index 3e85964e4..43567b92c 100644
--- a/pkg/sentry/mm/mm.go
+++ b/pkg/sentry/mm/mm.go
@@ -235,6 +235,14 @@ type MemoryManager struct {
 
 	// vdsoSigReturnAddr is the address of 'vdso_sigreturn'.
 	vdsoSigReturnAddr uint64
+
+	// membarrierPrivateEnabled is non-zero if EnableMembarrierPrivate has
+	// previously been called. Since, as of this writing,
+	// MEMBARRIER_CMD_PRIVATE_EXPEDITED is implemented as a global memory
+	// barrier, membarrierPrivateEnabled has no other effect.
+	//
+	// membarrierPrivateEnabled is accessed using atomic memory operations.
+	membarrierPrivateEnabled uint32
 }
 
 // vma represents a virtual memory area.
@@ -242,7 +250,7 @@ type MemoryManager struct {
 // +stateify savable
 type vma struct {
 	// mappable is the virtual memory object mapped by this vma. If mappable is
-	// nil, the vma represents a private anonymous mapping.
+	// nil, the vma represents an anonymous mapping.
 	mappable memmap.Mappable
 
 	// off is the offset into mappable at which this vma begins. If mappable is
diff --git a/pkg/sentry/mm/mm_test.go b/pkg/sentry/mm/mm_test.go
index fdc308542..acac3d357 100644
--- a/pkg/sentry/mm/mm_test.go
+++ b/pkg/sentry/mm/mm_test.go
@@ -51,7 +51,8 @@ func TestUsageASUpdates(t *testing.T) {
 	defer mm.DecUsers(ctx)
 
 	addr, err := mm.MMap(ctx, memmap.MMapOpts{
-		Length: 2 * usermem.PageSize,
+		Length:  2 * usermem.PageSize,
+		Private: true,
 	})
 	if err != nil {
 		t.Fatalf("MMap got err %v want nil", err)
diff --git a/pkg/sentry/mm/special_mappable.go b/pkg/sentry/mm/special_mappable.go
index f4c93baeb..2dbe5b751 100644
--- a/pkg/sentry/mm/special_mappable.go
+++ b/pkg/sentry/mm/special_mappable.go
@@ -136,9 +136,12 @@ func (m *SpecialMappable) Length() uint64 {
 // NewSharedAnonMappable returns a SpecialMappable that implements the
 // semantics of mmap(MAP_SHARED|MAP_ANONYMOUS) and mappings of /dev/zero.
 //
-// TODO(jamieliu): The use of SpecialMappable is a lazy code reuse hack. Linux
-// uses an ephemeral file created by mm/shmem.c:shmem_zero_setup(); we should
-// do the same to get non-zero device and inode IDs.
+// TODO(gvisor.dev/issue/1624): Linux uses an ephemeral file created by
+// mm/shmem.c:shmem_zero_setup(), and VFS2 does something analogous. VFS1 uses
+// a SpecialMappable instead, incorrectly getting device and inode IDs of zero
+// and causing memory for shared anonymous mappings to be allocated up-front
+// instead of on first touch; this is to avoid exacerbating the fs.MountSource
+// leak (b/143656263). Delete this function along with VFS1.
 func NewSharedAnonMappable(length uint64, mfp pgalloc.MemoryFileProvider) (*SpecialMappable, error) {
 	if length == 0 {
 		return nil, syserror.EINVAL
diff --git a/pkg/sentry/mm/syscalls.go b/pkg/sentry/mm/syscalls.go
index 4c9a575e7..0a66b1cdd 100644
--- a/pkg/sentry/mm/syscalls.go
+++ b/pkg/sentry/mm/syscalls.go
@@ -17,6 +17,7 @@ package mm
 import (
 	"fmt"
 	mrand "math/rand"
+	"sync/atomic"
 
 	"gvisor.dev/gvisor/pkg/abi/linux"
 	"gvisor.dev/gvisor/pkg/context"
@@ -24,7 +25,6 @@ import (
 	"gvisor.dev/gvisor/pkg/sentry/kernel/futex"
 	"gvisor.dev/gvisor/pkg/sentry/limits"
 	"gvisor.dev/gvisor/pkg/sentry/memmap"
-	"gvisor.dev/gvisor/pkg/sentry/pgalloc"
 	"gvisor.dev/gvisor/pkg/syserror"
 	"gvisor.dev/gvisor/pkg/usermem"
 )
@@ -93,18 +93,6 @@ func (mm *MemoryManager) MMap(ctx context.Context, opts memmap.MMapOpts) (userme
 		}
 	} else {
 		opts.Offset = 0
-		if !opts.Private {
-			if opts.MappingIdentity != nil {
-				return 0, syserror.EINVAL
-			}
-			m, err := NewSharedAnonMappable(opts.Length, pgalloc.MemoryFileProviderFromContext(ctx))
-			if err != nil {
-				return 0, err
-			}
-			defer m.DecRef(ctx)
-			opts.MappingIdentity = m
-			opts.Mappable = m
-		}
 	}
 
 	if opts.Addr.RoundDown() != opts.Addr {
@@ -1287,3 +1275,15 @@ func (mm *MemoryManager) VirtualDataSize() uint64 {
 	defer mm.mappingMu.RUnlock()
 	return mm.dataAS
 }
+
+// EnableMembarrierPrivate causes future calls to IsMembarrierPrivateEnabled to
+// return true.
+func (mm *MemoryManager) EnableMembarrierPrivate() {
+	atomic.StoreUint32(&mm.membarrierPrivateEnabled, 1)
+}
+
+// IsMembarrierPrivateEnabled returns true if mm.EnableMembarrierPrivate() has
+// previously been called.
+func (mm *MemoryManager) IsMembarrierPrivateEnabled() bool {
+	return atomic.LoadUint32(&mm.membarrierPrivateEnabled) != 0
+}