Futureweb

Software

BeamFS is a file-system proposal. Linux has not merged it.

BeamFS is an October 19 file-system proposal. It corrects bytes on one device and fails a read it cannot vouch for. Linux has not merged it.

The short version

Aurelien Desbrieres posted an 18-patch request for comments for BeamFS, a single-device file system that decodes Reed-Solomon parity on read and returns an error if it cannot vouch for the bytes. Phoronix reports the series the same morning. It is a proposal, not a merged file system.

Eighteen patches, still a request

The series tracker dates the mail to October 19, 2026 and lists 18 patches from Aurelien Desbrieres: on-disk format, a Reed-Solomon layer, allocators, inode and directory operations, a scrubber, documentation, Kconfig, and a MAINTAINERS entry. The tracker's status for the session is in progress. An in-progress CI row is a robot looking at a patch set. It is not a maintainer's merge. Asked directly, What is the October 9 report, relative to the patches? A Zenodo record of an earlier measurement campaign on beamfs 0.1.26, run by the project on Linux 7.3-rc5. It is not the kernel's acceptance of the October 19 series.memory safety in the release notes.

What the cover letter promises a read

Phoronix quotes the letter. BeamFS is a block-device file system that corrects corrupted bytes in place on a single device, with no second copy to repair from. Every data block is a capsule of Reed-Solomon codewords. Every read decodes. A read the decoder cannot vouch for fails with an input-output error instead of returning wrong bytes. Directory blocks, inodes, the bitmap, and the superblock carry parity too. In the default layout, 288 bytes of every 4,096 are overhead, which the letter calls 7 percent.the software desk.

It says it is a new format, not a renamed April post

The same quotation says BeamFS descends from an April request for comments called FTRFS, and that the earlier design's limits were reason to start a new file system with its own on-disk format and a new thread. The environments named are ones where data sits on one device and radiation is a concern, including a lineage that began with satellite computers. That lineage is the author's framing. It is not a flight log. Asked directly, Has Linux merged BeamFS? No. Phoronix presents the morning's mail as a request for comments, and it notes that new file systems face guidance the kernel adopted after earlier proposals. A mailing-list post is not a tarball on kernel.org.the October 12 prepatch.

An earlier report is the project's own campaign

A Zenodo record published October 9, 2026 describes measurements of beamfs 0.1.26 on Linux 7.3-rc5. The abstract says that, in the campaign, the file system did not return wrong data under the single-bit upsets the project injected, and it sets the scope as a family of stochastic bit flips rather than a claim about every fault. Those numbers are the project's. They predate the October 19 mail, and they do not merge the mail.Futureweb.

Phoronix is the morning's second account

Phoronix, on October 19, calls the series yet another file-system attempt and an initial request for comments. It repeats the cover letter and notes that kernel developers have guidance on admitting new file systems, adopted after FTRFS and other proposals. The article does not claim a performance win, and it does not claim a merge. Six comments on a news post are not review. The stable tarball filed beside this proposal is a different document.Linux 7.3, the release.

What this page will not treat as upstream

It will not say BeamFS is in Linux 7.3. It will not copy a fault-injection procedure, a scrubber command, or a mount option. It will not treat the author's assisted-by note in a patch header as a quality finding. The record is a proposal, a news account, and an earlier measurement the project published itself. Asked directly, What overhead does that letter name? 288 bytes of every 4,096 in the default layout, which the letter calls 7 percent. That figure is the proposal's, for the default layout, not a benchmark of a merged driver.

What a merge would look like

A maintainer's pull, a kernel.org tree that contains fs/beamfs, or a release note that names the file system would replace this brief's status line. A second laboratory repeating the October 9 campaign would be a different document from the author's report. Until one of those exists, BeamFS is mail. Asked directly, Who posted the series, and when? Aurelien Desbrieres, on October 19, 2026. The series tracker lists 18 patches, from an on-disk format through a MAINTAINERS entry, and marks the run as in progress.

Guidance is a gate, not a verdict

Phoronix says new file systems now face written guidance. The guidance can slow a proposal without rejecting this one. The series tracker still says in progress. Both can be true on the morning of October 19: the mail is public, and the kernel has not taken it. The short answer this page will stand on is this: Aurelien Desbrieres posted an 18-patch request for comments for BeamFS, a single-device file system that decodes Reed-Solomon parity on read and returns an error if it cannot vouch for the bytes. Phoronix reports the series the same morning. It is a proposal, not a merged file system.

What stays on the record

October 19, 18 patches, one device, a decode on read, an error when the decoder will not vouch, 7 percent overhead in the default layout the letter describes, and an October 9 report that is the project's own. Linux 7.3, filed separately, does not list this file system as merged. Asked directly, What does the cover letter say a read does? Phoronix quotes it: every data block is checked on read, and a read the decoder cannot vouch for fails with an error instead of returning bytes the decoder rejects. The cover letter also says there is no second copy to repair from.

Sources

The reports this brief is filing. Futureweb did not republish them.

  1. BeamFS RFC series
  2. Phoronix
  3. beamfs v3 report

Questions

Who posted the series, and when?

Aurelien Desbrieres, on October 19, 2026. The series tracker lists 18 patches, from an on-disk format through a MAINTAINERS entry, and marks the run as in progress.

What does the cover letter say a read does?

Phoronix quotes it: every data block is checked on read, and a read the decoder cannot vouch for fails with an error instead of returning bytes the decoder rejects. The cover letter also says there is no second copy to repair from.

What overhead does that letter name?

288 bytes of every 4,096 in the default layout, which the letter calls 7 percent. That figure is the proposal's, for the default layout, not a benchmark of a merged driver.

What is the October 9 report, relative to the patches?

A Zenodo record of an earlier measurement campaign on beamfs 0.1.26, run by the project on Linux 7.3-rc5. It is not the kernel's acceptance of the October 19 series.

Has Linux merged BeamFS?

No. Phoronix presents the morning's mail as a request for comments, and it notes that new file systems face guidance the kernel adopted after earlier proposals. A mailing-list post is not a tarball on kernel.org.