- cross-posted to:
- linux@programming.dev
- linux@programming.dev
- cross-posted to:
- linux@programming.dev
- linux@programming.dev
You must log in or # to comment.
Kent is such a dumbass. It’s a pitty because it’s a great filesystem.
Definitely not going anywhere near the comments section on that phoronix article. It’s guaranteed to be pure poison.
Tbh, what phoronix comment section isn’t pure neurological poison?
Should have never been merged into the kernel lol
I’ve never used bcachefs and only recently read about some of the drama. I wish the project the best but at this point it is hard to beat zfs
zfs is confusing as hell for noobs like me. I only really recently learned how to use btrfs. Is there any real reason to use zfs over btrfs on Linux anyway?
No unless your doing raid 5 or 6 not there isn’t.
On top of being confusing, I had my whole proxmox node crash because the ZFS pool randomly crashed out multiple times 🤷♂️
Probably due to the consumer grade nvme I was using it on but… Still why?
Also used a lot of extra ram just to function
I think it’s just hardware optimization. You get a ton more pain and risk replacing a drive in zfs vs raid10, but it’s more space efficient and flexible to use zfs. This is all academic, because the goal of these systems is a certain level of performance, availability, and data integrity, but not data safety. You need backups (preferably off-site and even off line) backups for that.
To be clear, I didn’t lose any data, ended up moving everything off the ZFS pool and went back to ext, the crashes I had just made the ZFS unavailable until I rebooted the machine.
There are some niche features, but if you’re not aware of them then no. It’s just licence encumbered btrfs for the majority of us.
I have used btrfs exactly once because it was the default on openSUSE, and the filesystem eventually became corrupted and unrecoverable.
They don’t really compete on the same features, but I get what you mean.
Good 👍
Why?
The developer of bcachefs, Kent Overstreet, has repeatedly failed to abide by the expectations of kernel release schedules, particularly the rc (release candidate) stage, which is supposed to freeze new features until next release.
Kent has open-air arguments with Linus Torvalds about not being able to develop the way he wants to, Linus Torvalds does not like wasting time discussing it with Kent.
IMO, Kent created this situation himself. He’ll be happier developing outside upstream anyway.
It should be noted that while some folks have commented that bcachefs was not ready for upstream, several kernel devs have a lot of respect for the technical quality of Kent’s work, so I think the argument of whether bcachefs is good or not good is separate from Kent’s behaviour as a kernel contributor.
Problem is if it isn’t in the kernel it well be used by a lot less people.
Additionally, Kent got most of his kernel changes needed for bcachefs merged already, so a dkms should be easier to manage now.
So if I’m reading this correctly, the program can operate fine as an external module because the kernel itself supports it well with those changes?
To be more clear, before he got his code mainlined, you needed to run h8s full fork of the kernel, with changes made not just to the cache code itself, but also to other parts.
Not all of his changes went in though; but the differences got sorted out enough that the vast majority of his newer changes were driver only.
That said, he was still ruffling feathers about wanting some fast moving kernel changes.
It was nowhere close to be mature enough to be in the kernel. The developer is nowhere close to be mature enough to be involved in the kernel. It’s better for everybody if it is developed separately and maybe integrated again at a later stage when the file system and tooling are considered stable and changes are smaller and less sensitive. CacheFS being in the kernel might mislead people to rely on a filesystem that is still experimental and under heavy development. Personally I am looking forward to see it mature because I’d love to run it on my file storage home server when it is stable enough.
I have been relying on the filesystem. It is excellent. It is mature enough.
Sadly, the lead dev for bcachefs is not mature enough.
I agree, on both statements.
It is easily stable enough for experimental use.
Not really why op said, though. It’s stable enough for
exeprimentaluseIt’s marked as experimental, hence the “experimental”
It was nowhere close to be mature enough to be in the kernel. The developer is nowhere close to be mature enough to be involved in the kernel
what independently verifiable condition(s) will satisfy these requirements?
That the developer himself finds it absolutely necessary to push new code outside the window for upcoming versions of the kernel is a pretty good indication.
That is a personality issue, not a code emergency.
There were two dozen patches submitted for 6.17 that were never merged. What has the fall-out been? Where are all the stories about data loss? I am sure they would hit the front page.
The file system can improve but it is already fine.
That is a personality issue, not a code emergency.
True, but it is an indication that the developer cannot follow a common rules. Simply Torvalds was tired of how he behaved.
There were two dozen patches submitted for 6.17 that were never merged. What has the fall-out been? Where are all the stories about data loss? I am sure they would hit the front page.
And so ? A patch can be submitted but never merged, for whatever reason. Problem is: these two dozen patches were submitted during the -RC cycle ?
The file system can improve but it is already fine.
Good. Now it it the developer that need to improve his attitude to work in teams.
The fallout for people knowingly risking their data beta testing a filesystem that is still in experimental and some users running into issues and possibly corruption?
There are no stories because it is not a story when a test environment for finding bugs fails and the bugs get fixed. Nobody with data they can not lose are putting it on bcachefs because why would they.
Thanks for running a test environment though. Please take backups of anything important, just in case.
They want you to use Oracle ZFS instead, they have a lot of money riding on this.
Who Torvalds? No the developer likes to do his own thing so he can now do his own thing.
Seriously?
No, comment is not true. You can use ZFS or BTFS, both of which are open source. ZFS just happens to be historically funded by Oracle, which is a good thing.
The reason is bcachefs has
majorstability problems (that don’t allow it to meet kernel release schedules). https://hackaday.com/2025/06/10/the-ongoing-bcachefs-filesystem-stability-controversy/Oracle funding doesn’t sound like a good thing at all since they’re basically CIA cloud ran by one of the most influential Zionists. But an unstable filesystem sounds even worse.
Open ZFS is now the main branch as far as I remember.
Everyone always says “Companies should fund FOSS instead of spending money on big corpos!”, yet then this.
It’s FOSS. It’s auditable. Funding is a good thing.
That’s true, but we also know that funding can come with stipulations. Oracle is an especially sketchy company.
But that counts for all big tech I guess.
So not using Linux at all then? Most of the development is paid for by big tech.
In this situation it works well, IMO. For some more context, ZFS was created by Sun (FOSS). Oacle bought them and built Oracle ZFS out of it. OpenZFS forked at that point from Sun code, and that’s what we use in Linux/etc. The Oracle variant supplies support to the FOSS variant. So Oracle has no control over OpenZFS.
“Major stability problems”. Hilarious. Get one person that actually uses bcachefs to confirm that. Good luck.
Kent has stability problems and drives me crazy but that is a baseless hit piece. Bcachefs is a solid fs.
Linus has not pushed back on the quality of bcachefs other than to say it is too unproven to rely on (too new). What Linus objected to was the process violations and the attitude of the lead developer.
Given how much Linus and the other LKML devs wanted to get rid of Kent, the fact that it took so long to dump him tells us that they really wanted to keep bcachefs (the technology).
There may have been more data loss bugs in btrfs and even OpenZFS than bcachefs since it was added to the kernel. I have many bcachefs systems. I have one btrfs system. Guess which one has caused me problems.
Fair enough on “major”. Edited that. But it has stability issues that aren’t handled well enough for RCs, so it’s not a hit piece to state that fact. Those stability issues may come from it being new, but it’s still an issue. Saying it’s because they want to “get rid of Kent” is just as much of a hit piece, too.
The dev is unstable. And he made the kernel process chaotic. But the filesystem itself is pretty solid. Do you have a link to stability issues?
My understanding is the stability risks come from active development additions vs “fixes” during that stage of the development cycle.
https://linuxiac.com/torvalds-expresses-regret-over-merging-bcachefs-into-kernel/
Simply put, only small bug fixes are allowed after the post-merge phase to integrate changes into the current kernel cycle. However, Overstreet’s PR included more than just fixes; it continued to develop new features, which always carry risks. That’s why Torvalds was unhappy with it. As a result, the changes were rejected.
…
Currently, the file system is being actively developed. Although it shows great potential with impressive features and strong data reliability, it’s not yet stable enough to be adopted by major Linux distributions as a proven and reliable solution.
YMMV, but my production systems will stick with ZFS since it’s kernel release updates are clear when there are “upgrades” vs “updates”, as you do those manually when it alerts you.
“Stable” in this context doesnt mean “your PC will definately crash and you will lose data!”, bcachefs is well past that. It means that the development is too active to be considered production ready since the code changes are too large to confirm the scary bit won’t happen (as much as can be).
Even JC threw in the towel on
bcachefs-toolsdue to this: https://www.phoronix.com/news/Debian-Orphans-Bcachefs-Tools
Thank You!!
This is a bot lollllll
Not all of us know what this is. Can you expand on your thoughts?
Agreed!
I don’t know what this is but it is a topic I am somewhat familiar with and it is somehow significant enough to be a headline.
Curious minds would like to know more.
Thanks!
Yes I’m asking for the reason why you think this development is good. It seemed to me like it could have worked out if they talked it out and could have added something of value to the OS
Very easy to search, but you’re lazy so:https://hackaday.com/2025/06/10/the-ongoing-bcachefs-filesystem-stability-controversy/
There’s no reason to be rude and insulting. It doesn’t make the other person look lazy; it just makes you look bad, especially when you end up being wrong because you didn’t do any research either. The article is garbage. It’s obviously written by someone who wants to talk about why they don’t like bcachefs, which would be fine, but they make it look like that’s why Linus wanted to remove bcachefs, which is a blatant lie.
Despite this, it has become clear that BcacheFS is rather unstable, with frequent and extensive patches being submitted to the point where [Linus Torvalds] in August of last year pushed back against it, as well as expressing regret for merging BcacheFS into mainline Linux.
But if we click on the article’s own source in the quote we see the message (emphasis mine):
Yeah, no, enough is enough. The last pull was already big.
This is too big, it touches non-bcachefs stuff, and it’s not even remotely some kind of regression.
At some point “fix something” just turns into development, and this is that point.
Nobody sane uses bcachefs and expects it to be stable, so every single user is an experimental site.
The bcachefs patches have become these kinds of "lots of development during the release cycles rather than before it", to the point where I’m starting to regret merging bcachefs.
If bcachefs can’t work sanely within the normal upstream kernel release schedule, maybe it shouldn’t be in the normal upstream kernel.
This is getting beyond ridiculous.
Stability has absolutely nothing to do with it. On the contrary, bcachefs is explicitly expected to be unstable. The entire thing is about the developer, Kent Overstreet, refusing to follow the linux development schedule and pushing features during a period where strictly bug fixes are allowed. This point is reiterated in the rest of the thread if anyone is having doubts about whether it is stated clearly enough in the above message alone.
Ohhhh nooooo 🙀
It’s not lazy to ask someone who seems to know something about the topic within a discussion thread about said topic. You know more than I do on this.
I understand how you may not want to take the time to answer someone’s question but also you could have replied with the link you eventually did instead of saying “Seriously?” Within the context of calling others lazy you could also qualify under the same term since you took the time to respond but not with the answer.
With search being what it is nowadays I wouldn’t know if I am getting a good result to find out the answer since it is of a technical and specific nature I may or may not even know if I am familiar with to begin with. It could take me much longer to figure it out, or I will give up and not be interested in finding out more about a field you seem to have an interest and knowledge about and I am demonstrating I want to know more about.
I think it is fair to ask for more information from someone who shows more expertise in the topic before searching.
It is
no u
I’ve heard about this and wanted to hear your opinion on it because you seemed to have gotten to another conclusion than I have. But it seems that you’re not interested in discussing so I’m no longer interested
