[vfs] kernfs: Fix inode memory leak issue.

This change aims to fix the memory leak issue reported in #3933.

Background:
VFS2 kernfs kept accumulating invalid dentries if those dentries were not
walked on. After substantial consideration of the problem by our team, we
decided to have an LRU cache solution. This change is the first part to that
solution, where we don't cache anything. The LRU cache can be added on top of
this.

What has changed:
- Introduced the concept of an inode tree in kernfs.OrderedChildren.
  This is helpful is cases where the lifecycle of an inode is different from
  that of a dentry.
    - OrderedChildren now deals with initialized inodes instead of initialized
      dentries. It now implements Lookup() where it constructs a new dentry
      using the inode.
    - OrderedChildren holds a ref on all its children inodes. With this change,
      now an inode can "outlive" a dentry pointing to it. See comments in
      kernfs.OrderedChildren.
- The kernfs dentry tree is solely maintained by kernfs only. Inode
  implementations can not modify the dentry tree.
- Dentries that reach ref count 0 are removed from the dentry tree.
- revalidateChildLocked now defer-DecRefs the newly created dentry from
  Inode.Lookup(), limiting its life to the current filesystem operation. If
  refs are picked on the dentry during the FS op (via an FD or something),
  then it will stick around and will be removed when the FD is closed. So there
  is essentially _no caching_ for Look()ed up dentries.
- kernfs.DecRef does not have the precondition that fs.mu must be locked.

Fixes #3933

PiperOrigin-RevId: 336768576

1 files changed, 12 insertions, 22 deletions

diff --git a/pkg/sentry/fsimpl/proc/task_files.go b/pkg/sentry/fsimpl/proc/task_files.go
index 3fbf081a6..79f8b7e9f 100644
--- a/pkg/sentry/fsimpl/proc/task_files.go
+++ b/pkg/sentry/fsimpl/proc/task_files.go
@@ -247,13 +247,10 @@ type commInode struct {
 	task *kernel.Task
 }
 
-func (fs *filesystem) newComm(task *kernel.Task, ino uint64, perm linux.FileMode) *kernfs.Dentry {
+func (fs *filesystem) newComm(task *kernel.Task, ino uint64, perm linux.FileMode) kernfs.Inode {
 	inode := &commInode{task: task}
 	inode.DynamicBytesFile.Init(task.Credentials(), linux.UNNAMED_MAJOR, fs.devMinor, ino, &commData{task: task}, perm)
-
-	d := &kernfs.Dentry{}
-	d.Init(inode)
-	return d
+	return inode
 }
 
 func (i *commInode) CheckPermissions(ctx context.Context, creds *auth.Credentials, ats vfs.AccessTypes) error {
@@ -658,13 +655,10 @@ type exeSymlink struct {
 
 var _ kernfs.Inode = (*exeSymlink)(nil)
 
-func (fs *filesystem) newExeSymlink(task *kernel.Task, ino uint64) *kernfs.Dentry {
+func (fs *filesystem) newExeSymlink(task *kernel.Task, ino uint64) kernfs.Inode {
 	inode := &exeSymlink{task: task}
 	inode.Init(task.Credentials(), linux.UNNAMED_MAJOR, fs.devMinor, ino, linux.ModeSymlink|0777)
-
-	d := &kernfs.Dentry{}
-	d.Init(inode)
-	return d
+	return inode
 }
 
 // Readlink implements kernfs.Inode.Readlink.
@@ -737,13 +731,10 @@ type cwdSymlink struct {
 
 var _ kernfs.Inode = (*cwdSymlink)(nil)
 
-func (fs *filesystem) newCwdSymlink(task *kernel.Task, ino uint64) *kernfs.Dentry {
+func (fs *filesystem) newCwdSymlink(task *kernel.Task, ino uint64) kernfs.Inode {
 	inode := &cwdSymlink{task: task}
 	inode.Init(task.Credentials(), linux.UNNAMED_MAJOR, fs.devMinor, ino, linux.ModeSymlink|0777)
-
-	d := &kernfs.Dentry{}
-	d.Init(inode)
-	return d
+	return inode
 }
 
 // Readlink implements kernfs.Inode.Readlink.
@@ -851,7 +842,7 @@ type namespaceSymlink struct {
 	task *kernel.Task
 }
 
-func (fs *filesystem) newNamespaceSymlink(task *kernel.Task, ino uint64, ns string) *kernfs.Dentry {
+func (fs *filesystem) newNamespaceSymlink(task *kernel.Task, ino uint64, ns string) kernfs.Inode {
 	// Namespace symlinks should contain the namespace name and the inode number
 	// for the namespace instance, so for example user:[123456]. We currently fake
 	// the inode number by sticking the symlink inode in its place.
@@ -862,9 +853,7 @@ func (fs *filesystem) newNamespaceSymlink(task *kernel.Task, ino uint64, ns stri
 	inode.Init(task.Credentials(), linux.UNNAMED_MAJOR, fs.devMinor, ino, target)
 
 	taskInode := &taskOwnedInode{Inode: inode, owner: task}
-	d := &kernfs.Dentry{}
-	d.Init(taskInode)
-	return d
+	return taskInode
 }
 
 // Readlink implements kernfs.Inode.Readlink.
@@ -882,11 +871,12 @@ func (s *namespaceSymlink) Getlink(ctx context.Context, mnt *vfs.Mount) (vfs.Vir
 	}
 
 	// Create a synthetic inode to represent the namespace.
+	fs := mnt.Filesystem().Impl().(*filesystem)
 	dentry := &kernfs.Dentry{}
-	dentry.Init(&namespaceInode{})
+	dentry.Init(&fs.Filesystem, &namespaceInode{})
 	vd := vfs.MakeVirtualDentry(mnt, dentry.VFSDentry())
-	vd.IncRef()
-	dentry.DecRef(ctx)
+	// Only IncRef vd.Mount() because vd.Dentry() already holds a ref of 1.
+	mnt.IncRef()
 	return vd, "", nil
 }