backpointer-based-consistency · research · also NoFS consistency

Backpointer-based consistency

Backpointer-based consistency checks agreement between a filesystem reference and ownership information embedded in its target.

Related Observatory article: Backpointer-based consistency on hesela.com

NoFS data-block backpointers encode owner inode number and logical block offset; inode backlinks identify parent directories. Allocation information is reconstructed online rather than trusted from persistent bitmaps. [1]

Following a crash, pathname lookup may reject an entry still visible in a directory listing. Orphan recovery is possible when its data blocks validate, but neither atomic transaction recovery nor restoration of overwritten contents is guaranteed. [1]

In the prototype, mounting does not finish verification. Calling stat() may inspect all of a file's blocks, the first overwrite of an unverified block requires an ownership read, and allocation can wait for a scan to identify usable free space. [1]

Relations

related

Soft updates, Metadata update dependency, Optimistic crash consistency, Crash consistency

Advantages

• Ownership checks can expose stale references to reallocated blocks. [1]

Limitations

• The design assumes atomic block-plus-backpointer writes. Matching identity permits stale versions; detection does not guarantee restoration of lost or overwritten contents. [1]

Citations

1. Chidambaram et al. Consistency Without Ordering. FAST 2012, sections 2.1, 3.2-3.3 and 4.2-4.4 (allocation reconstruction: 4.3). (accessed 2026-11-13)

JSON: /data/terms/backpointer-based-consistency.json