diff --git a/pkg/sentry/vfs/mount.go b/pkg/sentry/vfs/mount.go
index af571a7b1..02cf2c540 100644
--- a/pkg/sentry/vfs/mount.go
+++ b/pkg/sentry/vfs/mount.go
@@ -108,10 +108,6 @@ type Mount struct {
 	// Mount.EndWrite(). The MSB of writers is set if MS_RDONLY is in effect.
 	// writers is accessed using atomic memory operations.
 	writers atomicbitops.Int64
-
-	// pendingChildren is a list of new child mounts that have not yet been
-	// connected to this mount as the parent.
-	pendingChildren []*Mount
 }
 
 func newMount(vfs *VirtualFilesystem, fs *Filesystem, root *Dentry, mntns *MountNamespace, opts *MountOptions) *Mount {
@@ -308,39 +304,6 @@ func (vfs *VirtualFilesystem) cloneMount(mnt *Mount, root *Dentry, mopts *MountO
 	return clone
 }
 
-type cloneTreeNode struct {
-	prevMount   *Mount
-	parentMount *Mount
-}
-
-// cloneMountTree creates a copy of mnt's tree with the specified root
-// dentry at root. The new descendents are added to mnt's pending mount list.
-//
-// +checklocks:vfs.mountMu
-func (vfs *VirtualFilesystem) cloneMountTree(ctx context.Context, mnt *Mount, root *Dentry) (*Mount, error) {
-	clone := vfs.cloneMount(mnt, root, nil)
-	queue := []cloneTreeNode{{mnt, clone}}
-	for len(queue) != 0 {
-		p := queue[len(queue)-1]
-		queue = queue[:len(queue)-1]
-		for c := range p.prevMount.children {
-			m := vfs.cloneMount(c, c.root, nil)
-			vfs.delayDecRef(m)
-			mp := VirtualDentry{
-				mount:  p.parentMount,
-				dentry: c.point(),
-			}
-			mp.IncRef()
-			m.setKey(mp)
-			p.parentMount.pendingChildren = append(p.parentMount.pendingChildren, m)
-			if len(c.children) != 0 {
-				queue = append(queue, cloneTreeNode{c, m})
-			}
-		}
-	}
-	return clone, nil
-}
-
 // BindAt creates a clone of the source path's parent mount and mounts it at
 // the target path. The new mount's root dentry is one pointed to by the source
 // path.
diff --git a/pkg/sentry/vfs/namespace.go b/pkg/sentry/vfs/namespace.go
index c11d09063..131193fff 100644
--- a/pkg/sentry/vfs/namespace.go
+++ b/pkg/sentry/vfs/namespace.go
@@ -125,6 +125,11 @@ func (vfs *VirtualFilesystem) NewMountNamespaceFrom(
 	return mntns
 }
 
+type cloneEntry struct {
+	prevMount   *Mount
+	parentMount *Mount
+}
+
 // +checklocks:vfs.mountMu
 func (vfs *VirtualFilesystem) updateRootAndCWD(ctx context.Context, root *VirtualDentry, cwd *VirtualDentry, src *Mount, dst *Mount) {
 	if root.mount == src {
@@ -137,10 +142,6 @@ func (vfs *VirtualFilesystem) updateRootAndCWD(ctx context.Context, root *Virtua
 		cwd.mount = dst
 		cwd.mount.IncRef()
 	}
-	for srcChild := range src.children {
-		dstChild := vfs.mounts.Lookup(dst, srcChild.point())
-		vfs.updateRootAndCWD(ctx, root, cwd, srcChild, dstChild)
-	}
 }
 
 // NamespaceInodeGetter is an interface that provides the GetNamespaceInode method.
@@ -169,16 +170,38 @@ func (vfs *VirtualFilesystem) CloneMountNamespace(
 	vfs.lockMounts()
 	defer vfs.unlockMounts(ctx)
 
-	newRoot, err := vfs.cloneMountTree(ctx, ns.root, ns.root.root)
-	if err != nil {
-		newns.DecRef(ctx)
-		vfs.abortTree(ctx, newRoot)
-		return nil, err
+	ns.root.root.IncRef()
+	ns.root.fs.IncRef()
+	newns.root = newMount(vfs, ns.root.fs, ns.root.root, newns, &MountOptions{Flags: ns.root.Flags, ReadOnly: ns.root.ReadOnly()})
+	if ns.root.isShared {
+		vfs.addPeer(ns.root, newns.root)
 	}
-	newns.root = newRoot
-	newns.root.ns = newns
-	vfs.commitTree(ctx, newRoot)
 	vfs.updateRootAndCWD(ctx, root, cwd, ns.root, newns.root)
+
+	queue := []cloneEntry{cloneEntry{ns.root, newns.root}}
+	for len(queue) != 0 {
+		p := queue[0]
+		queue = queue[1:]
+		for c := range p.prevMount.children {
+			m := vfs.cloneMount(c, c.root, nil)
+			vd := VirtualDentry{
+				mount:  p.parentMount,
+				dentry: c.point(),
+			}
+			vd.IncRef()
+
+			err := vfs.connectMountAtLocked(ctx, m, vd)
+			vfs.delayDecRef(m)
+			if err != nil {
+				newns.DecRef(ctx)
+				return nil, err
+			}
+			vfs.updateRootAndCWD(ctx, root, cwd, c, m)
+			if len(c.children) != 0 {
+				queue = append(queue, cloneEntry{c, m})
+			}
+		}
+	}
 	return newns, nil
 }
 
diff --git a/pkg/sentry/vfs/propagation.go b/pkg/sentry/vfs/propagation.go
index 8f70af2b9..1d34d9740 100644
--- a/pkg/sentry/vfs/propagation.go
+++ b/pkg/sentry/vfs/propagation.go
@@ -153,43 +153,6 @@ func (vfs *VirtualFilesystem) abortPropagationTree(ctx context.Context, tree map
 	}
 }
 
-// +checklocks:vfs.mountMu
-func (vfs *VirtualFilesystem) commitTree(ctx context.Context, mnt *Mount) {
-	vfs.mounts.seq.BeginWrite()
-	for _, c := range mnt.pendingChildren {
-		vfs.commitTreeSeqed(ctx, c)
-	}
-	mnt.pendingChildren = nil
-	vfs.mounts.seq.EndWrite()
-}
-
-// +checklocks:vfs.mountMu
-func (vfs *VirtualFilesystem) commitTreeSeqed(ctx context.Context, mnt *Mount) {
-	mp := mnt.getKey()
-	mp.dentry.mu.Lock()
-	vfs.connectLocked(mnt, mp, mp.mount.ns)
-	mp.dentry.mu.Unlock()
-	for _, c := range mnt.pendingChildren {
-		vfs.commitTreeSeqed(ctx, c)
-	}
-	mnt.pendingChildren = nil
-}
-
-// abortTree releases references on a pending mount and all its pending
-// descendants.
-//
-// +checklocks:vfs.mountMu
-func (vfs *VirtualFilesystem) abortTree(ctx context.Context, mnt *Mount) {
-	mp := mnt.getKey()
-	vfs.delayDecRef(mnt)
-	vfs.delayDecRef(mp.dentry)
-	vfs.setPropagation(mnt, linux.MS_PRIVATE)
-	for _, c := range mnt.pendingChildren {
-		vfs.abortTree(ctx, c)
-	}
-	mnt.pendingChildren = nil
-}
-
 // SetMountPropagationAt changes the propagation type of the mount pointed to by
 // pop.
 func (vfs *VirtualFilesystem) SetMountPropagationAt(ctx context.Context, creds *auth.Credentials, pop *PathOperation, propFlags uint32) error {