dm-flakey · shipping · also Device-mapper flakey target

dm-flakey

dm-flakey is a Linux device-mapper target that alternates normal block I/O with configured error, write-drop or data-corruption behavior.

The up and down intervals select time windows, not an exact number of failed operations. During a down window, error_writes reports a write error, while drop_writes discards writes without reporting that failure. These modes test different recovery assumptions. [1]

Linux v6.12 modifies matching write-data bios through a private clone. If clone allocation fails, flakey_map falls through to remapping the original bio without the requested byte corruption. Fixed-byte replacement also has no effect when its position lies beyond the bio data or the existing byte already equals the replacement. A configured corruption attempt therefore does not establish altered bytes. [2]

Rebello et al. used a custom, targeted block injector for filesystem tests, then CuttleFS for application tests. A periodic dm-flakey configuration is not the same experimental protocol. [3]

Relations

related

Fault injection, Fault model, Lost write, Writeback error, Crash-test oracle

Limitations

• Use only disposable, isolated storage: an injector can damage the filesystem and application data being tested. [1]
• The reviewed v6.12 path passes zone-management operations through; it does not inject every command class. [2]

Citations

1. Linux v6.12: dm-flakey target parameters and feature behavior (accessed 2026-11-07)
2. Linux v6.12: dm-flakey byte replacement, cloning and I/O mapping (accessed 2026-11-07)
3. Rebello et al., Can Applications Recover from fsync Failures? USENIX ATC 2020, sections 3.2 and 4 (accessed 2026-11-07)

JSON: /data/terms/dm-flakey.json