Unconventional Linux Filesystems: OverlayFS, IncFS, ZoneFS, LTFS, and StegFS

Linux filesystems are diverse, with the "everything is a file" philosophy generating numerous different filesystem types. Some lack persistent storage (like tmpfs using memory), while others don't store real files at all (like procfs and sysfs providing system status information). This article explores several seemingly conventional yet innovative filesystems with unique approaches.

OverlayFS allows overlaying contents of one directory onto another, with source directories potentially on different volumes or filesystems. It creates a mechanism for temporary modifications to read-only filesystems. The "upper" directory is writable while the "lower" directory remains read-only. Modifications to files in the "lower" directory create copies in the "upper" directory. Deletions are handled through whiteouts and opaque directory mechanisms. Originally used in Live CD systems and Android development, its most significant application is in cloud platform container images, enabling layered container image distribution.

Incremental FS (IncFS) , proposed by Google in 2019 for Android app download updates, does not allow direct file writes, and appended content never changes. It enables running large Android applications before binary files are fully loaded. IncFS works atop other filesystems (EXT4, F2FS), passing most operations through while handling incomplete data loading by notifying applications to wait. Write operations record data blocks and update indexes; data blocks are compressed and decompressed on read. Read operations are logged for optimizing download order in future updates.

ZoneFS , developed by Western Digital, exposes each sequential zone of a zoned block device as a file. Zoned block devices divide address space into zones—conventional zones allow any read/write, while sequential zones accept random reads but require sequential writes with a write pointer maintained by the device. ZoneFS is a minimal filesystem that doesn't hide these constraints from users, making it closer to raw block device access than a full POSIX filesystem. It's useful for KV stores like LevelDB to store SSTables in sequential zones, implementing LSM Tree structures on zoned block devices.

LTFS (Linear Tape File System) brings filesystem capabilities to magnetic tape storage, which has been used for over 50 years without its own filesystem. Traditionally, tapes required external databases for file metadata, making it impossible to browse tape contents directly. LTFS stores the entire filesystem metadata at the beginning of the tape, enabling file定位 and reading. It supports file modifications by writing new content at the current tape position, with old data becoming obsolete. LTFS can mount multiple tape drives or span across networks to form unified filesystem partitions. Originally developed by IBM in 2008-2009, it's credited with revitalizing tape storage, with the European Extremely Large Telescope set to use LTFS for PB-scale daily data storage starting in 2025.

StegFS is a steganographic filesystem that uses steganography—the technique of hiding information existence—within the filesystem layer. All available space on the partition is filled with random encrypted data, with file data encrypted and randomly distributed across the partition. There's no easy way to distinguish meaningless garbage from steganographic data; file locations are derived from file keys. Built on EXT2, when StegFS is not mounted, it appears as a normal EXT2 partition with all steganographic files invisible. When mounted, a "stegfs" directory appears with hidden files in corresponding subdirectories. StegFS provides the unique capability of making the existence of hidden files unknown to others, offering a different approach from encryption.