vfs/cache: Fix promoteToFast race with WriteCloser Close

promoteToFast read FileInfo.Size from Stat after the VFS lock was
released, while WriteCloser.Close updated Size on the same live
object. Concurrent Create/Open on overlapping keys tripped the race
detector.

Stat now returns a FileInfo snapshot, and promotion uses the
ReadAll length for the fast-tier Create size. Promotion stays
best-effort.

#21

vfs/disk/disk.go

@@ -612,7 +612,9 @@ func (d *DiskFS) Delete(key string) error {
 	return nil
 }
 
-// Stat returns file information with lazy discovery
+// Stat returns a snapshot of file information with lazy discovery.
+// The returned *FileInfo is not the live cache entry; Close may update Size
+// on the in-map object under d.mu.
 func (d *DiskFS) Stat(key string) (*vfs.FileInfo, error) {
 	if key == "" {
 		return nil, vfserror.ErrInvalidKey
@@ -627,9 +629,10 @@ func (d *DiskFS) Stat(key string) (*vfs.FileInfo, error) {
 	keyMu.RLock()
 	d.mu.RLock()
 	if fi, ok := d.info[key]; ok {
+		snap := fi.Clone()
 		d.mu.RUnlock()
 		keyMu.RUnlock()
-		return fi, nil
+		return snap, nil
 	}
 	d.mu.RUnlock()
 	keyMu.RUnlock()
@@ -649,8 +652,9 @@ func (d *DiskFS) Stat(key string) (*vfs.FileInfo, error) {
 	// Double-check after acquiring write lock
 	d.mu.Lock()
 	if fi, ok := d.info[key]; ok {
+		snap := fi.Clone()
 		d.mu.Unlock()
-		return fi, nil
+		return snap, nil
 	}
 
 	// Re-verify the file still exists on disk under the lock before inserting.
@@ -669,9 +673,10 @@ func (d *DiskFS) Stat(key string) (*vfs.FileInfo, error) {
 	fi.UpdateAccessBatched(d.timeUpdater)
 	// Note: size not updated on lazy discovery (preserves prior behavior; initial on-disk accounted via bg populate at New time,
 	// subsequent files come via Create which accounts size).
+	snap := fi.Clone()
 	d.mu.Unlock()
 
-	return fi, nil
+	return snap, nil
 }
 
 // EvictLRU evicts the least recently used files to free up space