A new experimental filesystem, BeamFS, has been proposed for inclusion in the Linux kernel. Developed by Aurelien Desbrieres, the project focuses on protecting stored data from corruption caused by radiation, electromagnetic interference, and other hardware errors.
Unlike traditional filesystems that rely on checksums to detect corruption, BeamFS uses Reed–Solomon forward error correction to detect and repair damaged data directly on a storage device, even with only a single drive.
The approach is especially useful for systems operating in radiation-heavy environments such as satellites, spacecraft computers, and medical equipment used in radiation therapy. In these cases, storage errors can occur and adding redundant hardware may not be practical.
BeamFS organizes each 4 KB data block into 16 interleaved Reed–Solomon RS(255,239) codewords. This lets it correct up to eight corrupted bytes per codeword while requiring about 7% extra storage for error-correction information.
The same protection extends to filesystem metadata, including superblocks, inodes, directory structures, and allocation bitmaps. When data is read, BeamFS checks its integrity and tries to repair any detected errors before passing it to the operating system.
If recovery fails, the filesystem returns an I/O error instead of silently providing potentially corrupted data. The idea is simple: returning an error is better than delivering incorrect information without warning.
Moreover, BeamFS does not require multiple storage devices or mirrored copies to perform corrections. Instead, it uses redundancy stored alongside the data, although the damage it can repair is limited by the error-correction scheme.
The developer published experimental results from testing on x86-64 and ARM64 systems running Linux 7.3-rc5. During simulated single-bit corruption tests, BeamFS recovered the original data in nearly every case, returning an I/O error in the remaining one instead of supplying incorrect content.
In the same tests, ext3 and ext4 sometimes returned corrupted data without detecting the problem, while Btrfs detected the corruption but could not repair it. These findings are limited to the tested scenarios and should not be seen as a general comparison of filesystem reliability.
However, BeamFS has some important limitations. It currently lacks journaling and copy-on-write support, meaning an unexpected shutdown can require a full filesystem consistency check using fsck.beamfs.
More importantly, the project remains highly experimental. Its documentation explicitly warns about known bugs that can cause data loss, security vulnerabilities, and kernel crashes. Some issues have already been addressed in the development branch, but the filesystem is not considered safe for production workloads.
For now, the submission is an initial request for review and discussion among Linux filesystem developers. There is no indication yet of when or whether BeamFS might be accepted into the mainline Linux kernel.
