Ori advances GEFS filesystem for OpenBSD with improved error handling
Developer Ori pushes forward the Good Enough File System toward production readiness on OpenBSD.
What to know
- Ori released a major GEFS update with improved error handling, bugfixes from 9front, and new userspace tools—moving the filesystem closer to general use on OpenBSD.
- The consistency protocol remains the biggest missing piece; Ori is exploring write ordering barriers and synchronous I/O approaches to ensure data safety on crash.
- GEFS aims to address long-standing problems in Plan 9 filesystems: crash safety, corruption detection, and poor performance in large directories via copy-on-write Bε trees.
- The project is not yet in the official OpenBSD tree but has attracted early testers and contributors; community enthusiasm suggests strong interest in a modern filesystem alternative.
GEFS fills an important gap and represents promising progress toward a modern, production-ready filesystem for OpenBSD.
-
“A modern filesystem has been missing in OpenBSD for quite a while now. GEFS seems to fit the general mindset and is developed inside the tree already, so I'm really excited about it!”
Undeadly commenter · Undeadly ↗
The project remains early and is not yet ready for production; significant work on consistency protocols and snapshot control is still needed.
-
“Whoops, it's not in the tree yet, but I can't edit my post. Exciting nonetheless!”
Undeadly commenter · Undeadly ↗
How it unfolded 1 development · click the chart to see its coverage posts
-
1
Community discussion emerges on Hacker News and Lobsters
The GEFS update gains traction on tech discussion platforms, with commenters expressing enthusiasm for a modern filesystem on OpenBSD while noting the project is not yet in the official tree.
“A modern filesystem has been missing in OpenBSD for quite a while now. GEFS seems to fit the general mindset and is developed inside the tree already, so I'm really excited about it!”
— Undeadly commenter -
I spent the morning reading design document on GEFS. It’s quite interesting!
2 more of the top 3 · 13 posts in this stretch
-
How does GEFS's B-epsilon tree balance write gains against read latency on actual filesystem workloads?
-
Sadly, the incentives often work the other way. If a paper explains its core ideas well enough, reviewers often say that it is too obvious to merit being a contribution at a top-tier venue. I tend to have the opposite view: if you have explained the problem so well that your solution is obvious after reading that part of the paper (but not before…
-
-
background
Undeadly publishes coverage of GEFS progress — Undeadly, an OpenBSD news site, publishes an article by Peter N. M. Hansteen covering Ori's latest GEFS update, highlighting the filesystem's progress toward general use.
-
background
Ori publishes major GEFS update with error handling improvements — Ori sends a detailed technical update to the OpenBSD tech mailing list describing progress on GEFS. The update includes improvements to error handling, forward-ported bugfixes from 9front, userspace tools for filesystem creation, and plans for a userspace 'gefs' program to handle snapshot control.
What people are saying 9 voices from 1 site · best of 13 · verbatim
- Nov 2
-
Agreed! This is what I spend most of my time doing these days...
-
> Unfortunately I don't think that "just write better" is a relatively easy fix The most worthwhile fixes are rarely the easiest. I don't think there's wrong with doing something hard.
-
We've been doing something like the first-pass-reject round mentioned above more in the networking world (originally HotNets, now NSDI), and I really like it. It is a bummer to have your paper rejected at that round since it means you get less feedback, but at least you get a result fast, and hopefully the proceedings will be more fun and useful…
-
I essentially became a fan through Myrddin which still looks like such a cool programming language. Barely did anything with it to be honest but secretly hope for it one day taking over the world. ;) But now I hope GEFS takes over the world. Hehe.
-
That's part of what I meant with self-obsessed. The whole "too obvious" feels silly in a world where people have trouble reproducing findings and then not being sure if it's because they did something that wasn't explicitly mentioned. Leaving things out feels like the opposite of science. And terminology as well often times varies a bit, enough…
-
+1, I overlapped with him at the recurse center, greatly enjoyed chatting and enjoy whenever I see him around here tbh
-
Ori is also a terrific guy in person. I had the honour of meeting him around 10 years ago and he is extremely humble and wise.
-
I too really wanna recommend that read to everyone, even when not interested in GEFS or file systems. This is what in my opinion technical writing should look like. Really great clear non-self-obsessed writing. [off topic rambling on scientific writing] I think scientific paper authors can learn something from this as well and hopefully make it…
- Nov 1
-
This is so exciting to see! I can't wait until it reaches testing.